Influence of inter-observer delineation variability on radiomics stability in different tumor sites

Effects of Interobserver Segmentation Variability and Intensity Discretization on MRI-Based Radiomic Feature Reproducibility of Lipoma and Atypical Lipomatous Tumor

Segmentation and image intensity discretization impact on radiomics workflow. The aim of this study is to investigate the influence of interobserver segmentation variability and intensity discretization methods on the reproducibility of MRI-based radiomic features in lipoma and atypical lipomatous tumor (ALT). Thirty patients with lipoma or ALT were retrospectively included. Three readers independently performed manual contour-focused segmentation on T1-weighted and T2-weighted sequences, including the whole tumor volume. Additionally, a marginal erosion was applied to segmentations to evaluate its influence on feature reproducibility. After image pre-processing, with included intensity discretization employing both fixed bin number and width approaches, 1106 radiomic features were extracted from each sequence. Intraclass correlation coefficient (ICC) 95% confidence interval lower bound ≥ 0.75 defined feature stability. In contour-focused vs. margin shrinkage segmentation, the rates of stable features extracted from T1-weighted and T2-weighted images ranged from 92.68 to 95.21% vs. 90.69 to 95.66% after fixed bin number discretization and from 95.75 to 97.65% vs. 95.39 to 96.47% after fixed bin width discretization, respectively, with no difference between the two segmentation approaches (p ≥ 0.175). Higher stable feature rates and higher feature ICC values were found when implementing discretization with fixed bin width compared to fixed bin number, regardless of the segmentation approach (p < 0.001). In conclusion, MRI radiomic features of lipoma and ALT are reproducible regardless of the segmentation approach and intensity discretization method, although a certain degree of interobserver variability highlights the need for a preliminary reliability analysis in future studies.

Development and benchmarking of a Deep Learning-based MRI-guided gross tumor segmentation algorithm for Radiomics analyses in extremity soft tissue sarcomas

BACKGROUND

Volume of interest (VOI) segmentation is a crucial step for Radiomics analyses and radiotherapy (RT) treatment planning. Because it can be time-consuming and subject to inter-observer variability, we developed and tested a Deep Learning-based automatic segmentation (DLBAS) algorithm to reproducibly predict the primary gross tumor as VOI for Radiomics analyses in extremity soft tissue sarcomas (STS).

METHODS

A DLBAS algorithm was trained on a cohort of 157 patients and externally tested on an independent cohort of 87 patients using contrast-enhanced MRI. Manual tumor delineations by a radiation oncologist served as ground truths (GTs). A benchmark study with 20 cases from the test cohort compared the DLBAS predictions against manual VOI segmentations of two residents (ERs) and clinical delineations of two radiation oncologists (ROs). The ROs rated DLBAS predictions regarding their direct applicability.

RESULTS

The DLBAS achieved a median dice similarity coefficient (DSC) of 0.88 against the GTs in the entire test cohort (interquartile range (IQR): 0.11) and a median DSC of 0.89 (IQR 0.07) and 0.82 (IQR 0.10) in comparison to ERs and ROs, respectively. Radiomics feature stability was high with a median intraclass correlation coefficient of 0.97, 0.95 and 0.94 for GTs, ERs, and ROs, respectively. DLBAS predictions were deemed clinically suitable by the two ROs in 35% and 20% of cases, respectively.

CONCLUSION

The results demonstrate that the DLBAS algorithm provides reproducible VOI predictions for radiomics feature extraction. Variability remains regarding direct clinical applicability of predictions for RT treatment planning.

The effect of harmonization on the variability of PET radiomic features extracted using various segmentation methods

PURPOSE

This study aimed to examine the robustness of positron emission tomography (PET) radiomic features extracted via different segmentation methods before and after ComBat harmonization in patients with non-small cell lung cancer (NSCLC).

METHODS

We included 120 patients (positive recurrence = 46 and negative recurrence = 74) referred for PET scanning as a routine part of their care. All patients had a biopsy-proven NSCLC. Nine segmentation methods were applied to each image, including manual delineation, K-means (KM), watershed, fuzzy-C-mean, region-growing, local active contour (LAC), and iterative thresholding (IT) with 40, 45, and 50% thresholds. Diverse image discretizations, both without a filter and with different wavelet decompositions, were applied to PET images. Overall, 6741 radiomic features were extracted from each image (749 radiomic features from each segmented area). Non-parametric empirical Bayes (NPEB) ComBat harmonization was used to harmonize the features. Linear Support Vector Classifier (LinearSVC) with L1 regularization For feature selection and Support Vector Machine classifier (SVM) with fivefold nested cross-validation was performed using StratifiedKFold with 'n_splits' set to 5 to predict recurrence in NSCLC patients and assess the impact of ComBat harmonization on the outcome.

RESULTS

From 749 extracted radiomic features, 206 (27%) and 389 (51%) features showed excellent reliability (ICC ≥ 0.90) against segmentation method variation before and after NPEB ComBat harmonization, respectively. Among all, 39 features demonstrated poor reliability, which declined to 10 after ComBat harmonization. The 64 fixed bin widths (without any filter) and wavelets (LLL)-based radiomic features set achieved the best performance in terms of robustness against diverse segmentation techniques before and after ComBat harmonization. The first-order and GLRLM and also first-order and NGTDM feature families showed the largest number of robust features before and after ComBat harmonization, respectively. In terms of predicting recurrence in NSCLC, our findings indicate that using ComBat harmonization can significantly enhance machine learning outcomes, particularly improving the accuracy of watershed segmentation, which initially had fewer reliable features than manual contouring. Following the application of ComBat harmonization, the majority of cases saw substantial increase in sensitivity and specificity.

CONCLUSION

Radiomic features are vulnerable to different segmentation methods. ComBat harmonization might be considered a solution to overcome the poor reliability of radiomic features.

Characterization of renal masses with MRI-based radiomics: assessment of inter-package and inter-observer reproducibility in a prospective pilot study

OBJECTIVES

To evaluate radiomics features' reproducibility using inter-package/inter-observer measurement analysis in renal masses (RMs) based on MRI and to employ machine learning (ML) models for RM characterization.

METHODS

32 Patients (23M/9F; age 61.8 ± 10.6 years) with RMs (25 renal cell carcinomas (RCC)/7 benign masses; mean size, 3.43 ± 1.73 cm) undergoing resection were prospectively recruited. All patients underwent 1.5 T MRI with T2-weighted (T2-WI), diffusion-weighted (DWI)/apparent diffusion coefficient (ADC), and pre-/post-contrast-enhanced T1-weighted imaging (T1-WI). RMs were manually segmented using volume of interest (VOI) on T2-WI, DWI/ADC, and T1-WI pre-/post-contrast imaging (1-min, 3-min post-injection) by two independent observers using two radiomics software packages for inter-package and inter-observer assessments of shape/histogram/texture features common to both packages (104 features; n = 26 patients). Intra-class correlation coefficients (ICCs) were calculated to assess inter-observer and inter-package reproducibility of radiomics measurements [good (ICC ≥ 0.8)/moderate (ICC = 0.5-0.8)/poor (ICC < 0.5)]. ML models were employed using reproducible features (between observers and packages, ICC > 0.8) to distinguish RCC from benign RM.

RESULTS

Inter-package comparisons demonstrated that radiomics features from T1-WI-post-contrast had the highest proportion of good/moderate ICCs (54.8-58.6% for T1-WI-1 min), while most features extracted from T2-WI, T1-WI-pre-contrast, and ADC exhibited poor ICCs. Inter-observer comparisons found that radiomics measurements from T1-WI pre/post-contrast and T2-WI had the greatest proportion of features with good/moderate ICCs (95.3-99.1% T1-WI-post-contrast 1-min), while ADC measurements yielded mostly poor ICCs. ML models generated an AUC of 0.71 [95% confidence interval = 0.67-0.75] for diagnosis of RCC vs. benign RM.

CONCLUSION

Radiomics features extracted from T1-WI-post-contrast demonstrated greater inter-package and inter-observer reproducibility compared to ADC, with fair accuracy for distinguishing RCC from benign RM.

CLINICAL RELEVANCE

Knowledge of reproducibility of MRI radiomics features obtained on renal masses will aid in future study design and may enhance the diagnostic utility of radiomics models for renal mass characterization.

Predicting tumor deposits in rectal cancer: a combined deep learning model using T2-MR imaging and clinical features

BACKGROUND

Tumor deposits (TDs) are associated with poor prognosis in rectal cancer (RC). This study aims to develop and validate a deep learning (DL) model incorporating T2-MR image and clinical factors for the preoperative prediction of TDs in RC patients.

METHODS AND METHODS

A total of 327 RC patients with pathologically confirmed TDs status from January 2016 to December 2019 were retrospectively recruited, and the T2-MR images and clinical variables were collected. Patients were randomly split into a development dataset (n = 246) and an independent testing dataset (n = 81). A single-channel DL model, a multi-channel DL model, a hybrid DL model, and a clinical model were constructed. The performance of these predictive models was assessed by using receiver operating characteristics (ROC) analysis and decision curve analysis (DCA).

RESULTS

The areas under the curves (AUCs) of the clinical, single-DL, multi-DL, and hybrid-DL models were 0.734 (95% CI, 0.674-0.788), 0.710 (95% CI, 0.649-0.766), 0.767 (95% CI, 0.710-0.819), and 0.857 (95% CI, 0.807-0.898) in the development dataset. The AUC of the hybrid-DL model was significantly higher than the single-DL and multi-DL models (both p < 0.001) in the development dataset, and the single-DL model (p = 0.028) in the testing dataset. Decision curve analysis demonstrated the hybrid-DL model had higher net benefit than other models across the majority range of threshold probabilities.

CONCLUSIONS

The proposed hybrid-DL model achieved good predictive efficacy and could be used to predict tumor deposits in rectal cancer.

CRITICAL RELEVANCE STATEMENT

The proposed hybrid-DL model achieved good predictive efficacy and could be used to predict tumor deposits in rectal cancer.

KEY POINTS

• Preoperative non-invasive identification of TDs is of great clinical significance. • The combined hybrid-DL model achieved good predictive efficacy and could be used to predict tumor deposits in rectal cancer. • A preoperative nomogram provides gastroenterologist with an accurate and effective tool.

Enhancing Interoperability and Harmonisation of Nuclear Medicine Image Data and Associated Clinical Data

Nuclear imaging techniques such as positron emission tomography (PET) and single photon emission computed tomography (SPECT) in combination with computed tomography (CT) are established imaging modalities in clinical practice, particularly for oncological problems. Due to a multitude of manufacturers, different measurement protocols, local demographic or clinical workflow variations as well as various available reconstruction and analysis software, very heterogeneous datasets are generated. This review article examines the current state of interoperability and harmonisation of image data and related clinical data in the field of nuclear medicine. Various approaches and standards to improve data compatibility and integration are discussed. These include, for example, structured clinical history, standardisation of image acquisition and reconstruction as well as standardised preparation of image data for evaluation. Approaches to improve data acquisition, storage and analysis will be presented. Furthermore, approaches are presented to prepare the datasets in such a way that they become usable for projects applying artificial intelligence (AI) (machine learning, deep learning, etc.). This review article concludes with an outlook on future developments and trends related to AI in nuclear medicine, including a brief research of commercial solutions.

Dual adversarial deconfounding autoencoder for joint batch-effects removal from multi-center and multi-scanner radiomics data

Medical imaging represents the primary tool for investigating and monitoring several diseases, including cancer. The advances in quantitative image analysis have developed towards the extraction of biomarkers able to support clinical decisions. To produce robust results, multi-center studies are often set up. However, the imaging information must be denoised from confounding factors-known as batch-effect-like scanner-specific and center-specific influences. Moreover, in non-solid cancers, like lymphomas, effective biomarkers require an imaging-based representation of the disease that accounts for its multi-site spreading over the patient's body. In this work, we address the dual-factor deconfusion problem and we propose a deconfusion algorithm to harmonize the imaging information of patients affected by Hodgkin Lymphoma in a multi-center setting. We show that the proposed model successfully denoises data from domain-specific variability (p-value < 0.001) while it coherently preserves the spatial relationship between imaging descriptions of peer lesions (p-value = 0), which is a strong prognostic biomarker for tumor heterogeneity assessment. This harmonization step allows to significantly improve the performance in prognostic models with respect to state-of-the-art methods, enabling building exhaustive patient representations and delivering more accurate analyses (p-values < 0.001 in training, p-values < 0.05 in testing). This work lays the groundwork for performing large-scale and reproducible analyses on multi-center data that are urgently needed to convey the translation of imaging-based biomarkers into the clinical practice as effective prognostic tools. The code is available on GitHub at this https://github.com/LaraCavinato/Dual-ADAE .

Segmentation methods applied to MRI-derived radiomic analysis for the prediction of placenta accreta spectrum in patients with placenta previa

PURPOSE

To retrospectively evaluate the performance of different manual segmentation methods of placenta MR images for predicting Placenta Accreta Spectrum (PAS) disorders in patients with placenta previa (PP) using a Machine Learning (ML) Radiomics analysis.

METHODS

64 patients (n=41 with PAS and n= 23 without PAS) with PP who underwent MRI examination for suspicion of PAS were retrospectively selected. All MRI examinations were acquired on a 1.5 T using T2-weighted (T2w) sequences on axial, sagittal and coronal planes. Ten different manual segmentation methods were performed on sagittal placental T2-weighted images obtaining five sets of 2D regions of interest (ROIs) and five sets of 3D volumes of interest (VOIs) from each patient. In detail, ROIs and VOIs were positioned on the following areas: placental tissue, retroplacental myometrium, cervix, placenta with underneath myometrium, placenta with underneath myometrium and cervix. For feature stability testing, the same process was repeated on 30 randomly selected placental MRI examinations by two additional radiologists, working independently and blinded to the original segmentation. Radiomic features were extracted from all available ROIs and VOIs. 100 iterations of 5-fold cross-validation with nested feature selection, based on recursive feature elimination, were subsequently run on each ROI/VOI to identify the best-performing method to classify instances correctly.

RESULTS

Among the segmentation methods, the best performance in predicting PAS was obtained by the VOIs covering the retroplacental myometrium (Mean validation score: 0.761, standard deviation: 0.116).

CONCLUSION

Our preliminary results show that the VOI including the retroplacental myometrium using T2w images seems to be the best method when segmenting images for the development of ML radiomics predictive models to identify PAS in patients with PP.

Robustness of pulmonary nodule radiomic features on computed tomography as a function of varying radiation dose levels-a multi-dose in vivo patient study

OBJECTIVE

Analysis of textural features of pulmonary nodules in chest CT, also known as radiomics, has several potential clinical applications, such as diagnosis, prognostication, and treatment response monitoring. For clinical use, it is essential that these features provide robust measurements. Studies with phantoms and simulated lower dose levels have demonstrated that radiomic features can vary with different radiation dose levels. This study presents an in vivo stability analysis of radiomic features for pulmonary nodules against varying radiation dose levels.

METHODS

Nineteen patients with a total of thirty-five pulmonary nodules underwent four chest CT scans at different radiation dose levels (60, 33, 24, and 15 mAs) in a single session. The nodules were manually delineated. To assess the robustness of features, we calculated the intra-class correlation coefficient (ICC). To visualize the effect of milliampere-second variation on groups of features, a linear model was fitted to each feature. We calculated bias and calculated the R2 value as a measure of goodness of fit.

RESULTS

A small minority of 15/100 (15%) radiomic features were considered stable (ICC > 0.9). Bias increased and R2 decreased at lower dose, but shape features seemed to be more robust to milliampere-second variations than other feature classes.

CONCLUSION

A large majority of pulmonary nodule radiomic features were not inherently robust to radiation dose level variations. For a subset of features, it was possible to correct this variability by a simple linear model. However, the correction became increasingly less accurate at lower radiation dose levels.

CLINICAL RELEVANCE STATEMENT

Radiomic features provide a quantitative description of a tumor based on medical imaging such as computed tomography (CT). These features are potentially useful in several clinical tasks such as diagnosis, prognosis prediction, treatment effect monitoring, and treatment effect estimation.

KEY POINTS

• The vast majority of commonly used radiomic features are strongly influenced by variations in radiation dose level. • A small minority of radiomic features, notably the shape feature class, are robust against dose-level variations according to ICC calculations. • A large subset of radiomic features can be corrected by a linear model taking into account only the radiation dose level.

Morphometry-based radiomics for predicting therapeutic response in patients with gliomas following radiotherapy

Introduction

Gliomas are still considered as challenging in oncologic management despite the developments in treatment approaches. The complete elimination of a glioma might not be possible even after a treatment and assessment of therapeutic response is important to determine the future course of actions for patients with such cancers. In the recent years radiomics has emerged as a promising solution with potential applications including prediction of therapeutic response. Hence, this study was focused on investigating whether morphometry-based radiomics signature could be used to predict therapeutic response in patients with gliomas following radiotherapy.

Methods

105 magnetic resonance (MR) images including segmented and non-segmented images were used to extract morphometric features and develop a morphometry-based radiomics signature. After determining the appropriate machine learning algorithm, a prediction model was developed to predict the therapeutic response eliminating the highly correlated features as well as without eliminating the highly correlated features. Then the model performance was evaluated.

Results

Tumor grade had the highest contribution to develop the morphometry-based signature. Random forest provided the highest accuracy to train the prediction model derived from the morphometry-based radiomics signature. An accuracy of 86% and area under the curve (AUC) value of 0.91 were achieved for the prediction model evaluated without eliminating the highly correlated features whereas accuracy and AUC value were 84% and 0.92 respectively for the prediction model evaluated after eliminating the highly correlated features.

Discussion

Nonetheless, the developed morphometry-based radiomics signature could be utilized as a noninvasive biomarker for therapeutic response in patients with gliomas following radiotherapy.

REliability of consensus-based segMentatIoN in raDiomic feature reproducibility (REMIND): A word of caution

OBJECTIVE

To evaluate the reliability of consensus-based segmentation in terms of reproducibility of radiomic features.

METHODS

In this retrospective study, three tumor data sets were investigated: breast cancer (n = 30), renal cell carcinoma (n = 30), and pituitary macroadenoma (n = 30). MRI was utilized for breast and pituitary data sets, while CT was used for renal data set. 12 readers participated in the segmentation process. Consensus segmentation was created by making corrections on a previous region or volume of interest. Four experiments were designed to evaluate the reproducibility of radiomic features. Reliability was assessed with intraclass correlation coefficient (ICC) with two cut-off values: 0.75 and 0.9.

RESULTS

Considering the lower bound of the 95% confidence interval and the ICC threshold of 0.90, at least 61% of the radiomic features were not reproducible in the inter-consensus analysis. In the susceptibility experiment, at least half (54%) became non-reproducible when the first reader is replaced with a different reader. In the intra-consensus analysis, at least about one-third (32%) were non-reproducible when the same second reader segmented the image over the same first reader two weeks later. Compared to inter-reader analysis based on independent single readers, the inter-consensus analysis did not statistically significantly improve the rates of reproducible features in all data sets and analyses.

CONCLUSIONS

Despite the positive connotation of the word "consensus", it is essential to REMIND that consensus-based segmentation has significant reproducibility issues. Therefore, the usage of consensus-based segmentation alone should be avoided unless a reliability analysis is performed, even if it is not practical in clinical settings.

Artificial intelligence-driven radiomics study in cancer: the role of feature engineering and modeling

Modern medicine is reliant on various medical imaging technologies for non-invasively observing patients' anatomy. However, the interpretation of medical images can be highly subjective and dependent on the expertise of clinicians. Moreover, some potentially useful quantitative information in medical images, especially that which is not visible to the naked eye, is often ignored during clinical practice. In contrast, radiomics performs high-throughput feature extraction from medical images, which enables quantitative analysis of medical images and prediction of various clinical endpoints. Studies have reported that radiomics exhibits promising performance in diagnosis and predicting treatment responses and prognosis, demonstrating its potential to be a non-invasive auxiliary tool for personalized medicine. However, radiomics remains in a developmental phase as numerous technical challenges have yet to be solved, especially in feature engineering and statistical modeling. In this review, we introduce the current utility of radiomics by summarizing research on its application in the diagnosis, prognosis, and prediction of treatment responses in patients with cancer. We focus on machine learning approaches, for feature extraction and selection during feature engineering and for imbalanced datasets and multi-modality fusion during statistical modeling. Furthermore, we introduce the stability, reproducibility, and interpretability of features, and the generalizability and interpretability of models. Finally, we offer possible solutions to current challenges in radiomics research.

Assessment of Variabilities in Lung-Contouring Methods on CBCT Preclinical Radiomics Outputs

Simple Summary

This study is the first to evaluate the impact of contouring differences on radiomics analysis in preclinical CBCT scans. We found that the variation in quantitative image readouts was greater between segmentation tools than between observers.

Abstract

Radiomics image analysis has the potential to uncover disease characteristics for the development of predictive signatures and personalised radiotherapy treatment. Inter-observer and inter-software delineation variabilities are known to have downstream effects on radiomics features, reducing the reliability of the analysis. The purpose of this study was to investigate the impact of these variabilities on radiomics outputs from preclinical cone-beam computed tomography (CBCT) scans. Inter-observer variabilities were assessed using manual and semi-automated contours of mouse lungs (n = 16). Inter-software variabilities were determined between two tools (3D Slicer and ITK-SNAP). The contours were compared using Dice similarity coefficient (DSC) scores and the 95th percentile of the Hausdorff distance (HD_95p) metrics. The good reliability of the radiomics outputs was defined using intraclass correlation coefficients (ICC) and their 95% confidence intervals. The median DSC scores were high (0.82–0.94), and the HD_95p metrics were within the submillimetre range for all comparisons. the shape and NGTDM features were impacted the most. Manual contours had the most reliable features (73%), followed by semi-automated (66%) and inter-software (51%) variabilities. From a total of 842 features, 314 robust features overlapped across all contouring methodologies. In addition, our results have a 70% overlap with features identified from clinical inter-observer studies.

Radiomics combined with transcriptomics to predict response to immunotherapy from patients treated with PD-1/PD-L1 inhibitors for advanced NSCLC

Introduction

In this study, we aim to build radiomics and multiomics models based on transcriptomics and radiomics to predict the response from patients treated with the PD-L1 inhibitor.

Materials and methods

One hundred and ninety-five patients treated with PD-1/PD-L1 inhibitors were included. For all patients, 342 radiomic features were extracted from pretreatment computed tomography scans. The training set was built with 110 patients treated at the Léon Bérard Cancer Center. An independent validation cohort was built with the 85 patients treated in Dijon. The two sets were dichotomized into two classes, patients with disease control and those considered non-responders, in order to predict the disease control at 3 months. Various models were trained with different feature selection methods, and different classifiers were evaluated to build the models. In a second exploratory step, we used transcriptomics to enrich the database and develop a multiomic signature of response to immunotherapy in a 54-patient subgroup. Finally, we considered the HOT/COLD status. We first trained a radiomic model to predict the HOT/COLD status and then prototyped a hybrid model integrating radiomics and the HOT/COLD status to predict the response to immunotherapy.

Results

Radiomic signature for 3 months' progression-free survival (PFS) classification: The most predictive model had an area under the receiver operating characteristic curve (AUROC) of 0.94 on the training set and 0.65 on the external validation set. This model was obtained with the t-test selection method and with a support vector machine (SVM) classifier. Multiomic signature for PFS classification: The most predictive model had an AUROC of 0.95 on the training set and 0.99 on the validation set. Radiomic model to predict the HOT/COLD status: the most predictive model had an AUROC of 0.93 on the training set and 0.86 on the validation set. HOT/COLD radiomic hybrid model for PFS classification: the most predictive model had an AUROC of 0.93 on the training set and 0.90 on the validation set.

Conclusion

In conclusion, radiomics could be used to predict the response to immunotherapy in non-small-cell lung cancer patients. The use of transcriptomics or the HOT/COLD status, together with radiomics, may improve the working of the prediction models.

CT texture analysis reliability in pulmonary lesions: the influence of 3D vs. 2D lesion segmentation and volume definition by a Hounsfield-unit threshold

OBJECTIVE

Reproducibility problems are a known limitation of radiomics. The segmentation of the target lesion plays a critical role in texture analysis variability. This study's aim was to compare the interobserver reliability of manual 2D vs. 3D lung lesion segmentation with and without pre-definition of the volume using a threshold of - 50 HU.

METHODS

Seventy-five patients with histopathologically proven lung lesions (15 patients each with adenocarcinoma, squamous cell carcinoma, small cell lung cancer, carcinoid, and organizing pneumonia) who underwent an unenhanced CT scan of the chest were included. Three radiologists independently segmented each lesion manually in 3D and 2D with and without pre-segmentation volume definition by a HU threshold, and shape parameters and original, Laplacian of Gaussian-filtered, and wavelet-based texture features were derived. To assess interobserver reliability and identify the most robust texture features, intraclass correlation coefficients (ICCs) for different segmentation settings were calculated.

RESULTS

Shape parameters had high reliability (64-79% had excellent and good ICCs). Texture features had weak reliability levels, with the highest ICCs (38% excellent or good) found for original features in 3D segmentation without the use of a HU threshold. A small proportion (4.3-11.5%) of texture features had excellent or good ICC values at all segmentation settings.

CONCLUSION

Interobserver reliability of texture features from CT scans of a heterogeneous collection of manually segmented lung lesions was low with a small proportion of features demonstrating high reliability independent of the segmentation settings. These results indicate a limited applicability of texture analysis and the need to define robust texture features in patients with lung lesions.

KEY POINTS

• Our study showed a low reproducibility of texture features when 3 radiologists independently segmented lung lesions in CT images, which highlights a serious limitation of texture analysis. • Interobserver reliability of texture features was low regardless of whether the lesion was segmented in 2D and 3D with or without a HU threshold. • In contrast to texture features, shape parameters showed a high interobserver reliability when lesions were segmented in 2D vs. 3D with and without a HU threshold of - 50.

Ultrasound-based radiomics in the diagnosis of carpal tunnel syndrome: The influence of regions of interest delineation method on mode

BACKGROUND

In the recent years, artificial intelligence (AI) algorithms have been used to accurately diagnose musculoskeletal diseases. However, it is not known whether the particular regions of interest (ROI) delineation method would affect the performance of the AI algorithm.

PURPOSE

The purpose of this study was to investigate the influence of ROI delineation methods on model performance and observer consistency.

METHODS

In this retrospective analysis, ultrasound (US) measures of median nerves affected with carpal tunnel syndrome (CTS) were compared to median nerves in a control group without CTS. Two methods were used for delineation of the ROI: (1) the ROI along the hyperechoic medial edge of the median nerve but not including the epineurium (MN) (ROI1); and (2) the ROI including the hyperechoic epineurium (ROI2), respectively. The intra group correlation coefficient (ICC) was used to compare the observer consistency of ROI features (i.e. the corresponding radiomics parameters). Parameters α₁ and α₂ were obtained based on the ICC of ROI₁ features and ROI₂ features. The ROC analysis was used to determine the area under the curve (AUC) and evaluate the performance of the radiologists and network. In addition, four indices, namely sensitivity, specificity, positive prediction and negative prediction were analyzed too.

RESULTS

A total of 136 wrists of 77 CTS group and 136 wrists of 74 control group were included in the study. Control group was matched to CTS group according to the age and sex. The observer consistency of ROI features delineated by the two schemes was different, and the consistency of ROI₁ features was higher (α1 ˃ α2). The intra-observer consistency was higher than the inter-observer consistency regardless of the scheme, and the intra-observer consistency was higher when chose scheme one. The performances of models based on the two ROI features were different, although the AUC of each model was greater than 0.8.The model performed better when the MN epineurium was included in the ROI. Among five artificial intelligence algorithms, the Forest models (model₁ achieved an AUC of 0.921 in training datasets and 0.830 in testing datasets; model₂ achieved an AUC of 0.967 in training datasets and 0.872 in testing datasets.) obtained the highest performance, followed by the support vector machine (SVM) models and the Logistic models. The performances of the models were significantly better than the inexperienced radiologist (Dr. B. Z. achieved an AUC of 0.702).

CONCLUSION

Different ROI delineation methods may affect the performance of the model and the consistency of observers. Model performance was better when the ROI contained the MN epineurium, and observer consistency was higher when the ROI was delineated along the hyperechoic medial border of the MN.

Brain metastases from NSCLC treated with stereotactic radiotherapy: prediction mismatch between two different radiomic platforms

BACKGROUND AND PURPOSE

Radiomics enables the mining of quantitative features from medical images. The influence of the radiomic feature extraction software on the final performance of models is still a poorly understood topic. This study aimed to investigate the ability of radiomic features extracted by two different radiomic platforms to predict clinical outcomes in patients treated with radiosurgery for brain metastases from non-small cell lung cancer. We developed models integrating pre-treatment magnetic resonance imaging (MRI)-derived radiomic features and clinical data.

MATERIALS AND METHODS

Pre-radiotherapy gadolinium enhanced axial T1-weighted MRI scans were used. MRI images were re-sampled, intensity-shifted, and histogram-matched before radiomic extraction by means of two different platforms (PyRadiomics and SOPHiA Radiomics). We adopted LASSO Cox regression models for multivariable analyses by creating radiomic, clinical, and combined models using three survival clinical endpoints (local control, distant progression, and overall survival). The statistical analysis was repeated 50 times with different random seeds and the median concordance index was used as performance metric of the models.

RESULTS

We analysed 276 metastases from 148 patients. The use of the two platforms resulted in differences in both the quality and the number of extractable features. That led to mismatches in terms of end-to-end performance, statistical significance of radiomic scores, and clinical covariates found significant in combined models.

CONCLUSION

This study shed new light on how extracting radiomic features from the same images using two different platforms could yield several discrepancies. That may lead to acute consequences on drawing conclusions, comparing results across the literature, and translating radiomics into clinical practice.

External validation of a CT-based radiomics signature in oropharyngeal cancer: Assessing sources of variation

BACKGROUND AND PURPOSE

Radiomics is a high-throughput approach that allows for quantitative analysis of imaging data for prognostic applications. Medical images are used in oropharyngeal cancer (OPC) diagnosis and treatment planning and these images may contain prognostic information allowing for treatment personalization. However, the lack of validated models has been a barrier to the translation of radiomic research to the clinic. We hypothesize that a previously developed radiomics model for risk stratification in OPC can be validated in a local dataset.

MATERIALS AND METHODS

The radiomics signature predicting overall survival incorporates features derived from the primary gross tumor volume of OPC patients treated with radiation +/- chemotherapy at a single institution (n = 343). Model fit, calibration, discrimination, and utility were evaluated. The signature was compared with a clinical model using overall stage and a model incorporating both radiomics and clinical data. A model detecting dental artifacts on computed tomography images was also validated.

RESULTS

The radiomics signature had a Concordance index (C-index) of 0.66 comparable to the clinical model's C-index of 0.65. The combined model significantly outperformed (C-index of 0.69, p = 0.024) the clinical model, suggesting that radiomics provides added value. The dental artifact model demonstrated strong ability in detecting dental artifacts with an area under the curve of 0.87.

CONCLUSION

This work demonstrates model performance comparable to previous validation work and provides a framework for future independent and multi-center validation efforts. With sufficient validation, radiomic models have the potential to improve traditional systems of risk stratification, treatment personalization and patient outcomes.

Robustness of Radiomics in Pre-Surgical Computer Tomography of Non-Small-Cell Lung Cancer

Background: Radiomic features are increasingly used in CT of NSCLC. However, their robustness with respect to segmentation variability has not yet been demonstrated. The aim of this study was to assess radiomic features agreement across three kinds of segmentation. Methods: We retrospectively included 48 patients suffering from NSCLC who underwent pre-surgery CT. Two expert radiologists in consensus manually delineated three 3D-ROIs on each patient. To assess robustness for each feature, the intra-class correlation coefficient (ICC) across segmentations was evaluated. The ‘sensitivity’ of ICC upon some parameters affecting features computation (such as bin-width for first-order features and pixel-distances for second-order features) was also evaluated. Moreover, an assessment with respect to interpolator and isotropic resolution was also performed. Results: Our results indicate that ‘shape’ features tend to have excellent agreement (ICC > 0.9) across segmentations; moreover, they have approximately zero sensitivity to other parameters. ‘First-order’ features are in general sensitive to parameters variation; however, a few of them showed excellent agreement and low sensitivity (below 0.1) with respect to bin-width and pixel-distance. Similarly, a few second-order features showed excellent agreement and low sensitivity. Conclusions: Our results suggest that a limited number of radiomic features can achieve a high level of reproducibility in CT of NSCLC.

Application of nnU-Net for Automatic Segmentation of Lung Lesions on CT Images and Its Implication for Radiomic Models

Radiomics investigates the predictive role of quantitative parameters calculated from radiological images. In oncology, tumour segmentation constitutes a crucial step of the radiomic workflow. Manual segmentation is time-consuming and prone to inter-observer variability. In this study, a state-of-the-art deep-learning network for automatic segmentation (nnU-Net) was applied to computed tomography images of lung tumour patients, and its impact on the performance of survival radiomic models was assessed. In total, 899 patients were included, from two proprietary and one public datasets. Different network architectures (2D, 3D) were trained and tested on different combinations of the datasets. Automatic segmentations were compared to reference manual segmentations performed by physicians using the DICE similarity coefficient. Subsequently, the accuracy of radiomic models for survival classification based on either manual or automatic segmentations were compared, considering both hand-crafted and deep-learning features. The best agreement between automatic and manual contours (DICE = 0.78 ± 0.12) was achieved averaging 2D and 3D predictions and applying customised post-processing. The accuracy of the survival classifier (ranging between 0.65 and 0.78) was not statistically different when using manual versus automatic contours, both with hand-crafted and deep features. These results support the promising role nnU-Net can play in automatic segmentation, accelerating the radiomic workflow without impairing the models' accuracy. Further investigations on different clinical endpoints and populations are encouraged to confirm and generalise these findings.

Exploring Texture Analysis to Optimize Bladder Preservation in Muscle Invasive Bladder Cancer

PURPOSE

To explore if texture analysis of Muscle Invasive Bladder Cancer (MIBC) can aid in better patient selection for bladder preservation.

METHODS

Pretreatment noncontrast CT images of 41 patients of MIBC treated with bladder preservation were included. The visible tumor was contoured on all slices by a single observer. The primary endpoint was to identify texture parameters associated with disease recurrence posttreatment. The secondary endpoints included intra and interobserver variability, single and multislice analysis, and differentiating the texture features of normal bladder and tumor. For interobserver variability of bladder tumor texture features, 3 observers contoured the visible tumor on all slices independently. Observer 1 contoured again at an interval of 1 month for intraobserver variability.

RESULTS

The median follow-up was 30 months with 12 patients having a recurrence. In the primary endpoint analysis, the mean of the pixels at Spatial Scaling Filter (SSF) 2 for the no recurrence group and recurrence group was 6.44 v 13.73 respectively (P = .031) and the same at SSF-3 was 11.95 and 22.32 respectively (P = .034). The texture features that could significantly differentiate tumor and normal bladder were mean, standard deviation and kurtosis of the pixels at SSF-2 and entropy and kurtosis of the pixels at SSF-3. Overall, there was an excellent intra and interobserver concordance in texture features. Only multislice analysis and not single-slice could differentiate recurrence and no recurrence posttreatment.

CONCLUSIONS

Texture analysis can be explored as a modality for patient selection for bladder preservation along with the established clinical parameters to improve outcomes.

Robustness of radiomics to variations in segmentation methods in multimodal brain MRI

Radiomics in neuroimaging uses fully automatic segmentation to delineate the anatomical areas for which radiomic features are computed. However, differences among these segmentation methods affect radiomic features to an unknown extent. A scan-rescan dataset (n = 46) of T1-weighted and diffusion tensor images was used. Subjects were split into a sleep-deprivation and a control group. Scans were segmented using four segmentation methods from which radiomic features were computed. First, we measured segmentation agreement using the Dice-coefficient. Second, robustness and reproducibility of radiomic features were measured using the intraclass correlation coefficient (ICC). Last, difference in predictive power was assessed using the Friedman-test on performance in a radiomics-based sleep deprivation classification application. Segmentation agreement was generally high (interquartile range = 0.77–0.90) and median feature robustness to segmentation method variation was higher (ICC > 0.7) than scan-rescan reproducibility (ICC 0.3–0.8). However, classification performance differed significantly among segmentation methods (p < 0.001) ranging from 77 to 84%. Accuracy was higher for more recent deep learning-based segmentation methods. Despite high agreement among segmentation methods, subtle differences significantly affected radiomic features and their predictive power. Consequently, the effect of differences in segmentation methods should be taken into account when designing and evaluating radiomics-based research methods.

Investigation of the added value of CT-based radiomics in predicting the development of brain metastases in patients with radically treated stage III NSCLC

Introduction:

Despite radical intent therapy for patients with stage III non-small-cell lung cancer (NSCLC), cumulative incidence of brain metastases (BM) reaches 30%. Current risk stratification methods fail to accurately identify these patients. As radiomics features have been shown to have predictive value, this study aims to develop a model combining clinical risk factors with radiomics features for BM development in patients with radically treated stage III NSCLC.

Methods:

Retrospective analysis of two prospective multicentre studies. Inclusion criteria: adequately staged [¹⁸F-fluorodeoxyglucose positron emission tomography-computed tomography (18-FDG-PET-CT), contrast-enhanced chest CT, contrast-enhanced brain magnetic resonance imaging/CT] and radically treated stage III NSCLC, exclusion criteria: second primary within 2 years of NSCLC diagnosis and prior prophylactic cranial irradiation. Primary endpoint was BM development any time during follow-up (FU). CT-based radiomics features (N = 530) were extracted from the primary lung tumour on 18-FDG-PET-CT images, and a list of clinical features (N = 8) was collected. Univariate feature selection based on the area under the curve (AUC) of the receiver operating characteristic was performed to identify relevant features. Generalized linear models were trained using the selected features, and multivariate predictive performance was assessed through the AUC.

Results:

In total, 219 patients were eligible for analysis. Median FU was 59.4 months for the training cohort and 67.3 months for the validation cohort; 21 (15%) and 17 (22%) patients developed BM in the training and validation cohort, respectively. Two relevant clinical features (age and adenocarcinoma histology) and four relevant radiomics features were identified as predictive. The clinical model yielded the highest AUC value of 0.71 (95% CI: 0.58–0.84), better than radiomics or a combination of clinical parameters and radiomics (both an AUC of 0.62, 95% CIs of 0.47–076 and 0.48–0.76, respectively).

Conclusion:

CT-based radiomics features of primary NSCLC in the current setup could not improve on a model based on clinical predictors (age and adenocarcinoma histology) of BM development in radically treated stage III NSCLC patients.

The value of combined PET/MRI, CT and clinical metabolic parameters in differentiating lung adenocarcinoma from squamous cell carcinoma

Objective

This study aimed to study the diagnostic efficacy of positron emission tomography (PET)/magnetic resonance imaging (MRI), computed tomography (CT) and clinical metabolic parameters in predicting the histological classification of lung adenocarcinoma (ADC) and squamous cell carcinoma (SCC).

Methods

PET/MRI, CT and clinical metabolic data of 80 patients with lung ADC or SCC were retrospectively collected. According to the pathological results from surgery or fiberscopy, the patients were diagnosed with lung ADC (47 cases) or SCC (33 cases). All 80 patients were divided into a training group (64 cases), an internal testing group (8 cases) and an external testing group (8 cases) in the ratio of 8:1:1. Nine models were constructed by integrating features from different modalities. The Gaussian classifier was used to differentiate ADC and SCC. The prediction ability was evaluated using the receiver operating characteristic curve. The area under the curve (AUC) of the models was compared using Delong’s test. Based on the best composite model, a nomogram was established and evaluated with a calibration curve, decision curve and clinical impact curve.

Results

The composite model (PET/MRI + CT + Clinical) owned the highest AUC values in the training, internal testing and external testing sets, respectively. In the training set, significant differences in the AUC were found between the composite model and other models except for the PET/MRI + CT model. The calibration curves showed good consistency between the predicted output and actual disease. The decision curve analysis and clinical impact curves demonstrated that the composite model increased the clinical net benefit for predicting lung cancer subtypes.

Conclusion

The composite prediction model of PET/MRI + CT + Clinical better distinguished ADC from SCC pathological subtypes preoperatively and achieved clinical benefits, thus providing an accurate clinical diagnosis.

CT-based morphologic and radiomics features for the classification of MYCN gene amplification status in pediatric neuroblastoma

PURPOSE

MYCN onco-gene amplification in neuroblastoma confers patients to the high-risk disease category for which prognosis is poor and more aggressive multimodal treatment is indicated. This retrospective study leverages machine learning techniques to develop a computed tomography (CT)-based model incorporating semantic and non-semantic features for non-invasive prediction of MYCN amplification status in pediatric neuroblastoma.

METHODS

From 2009 to 2020, 54 pediatric patients treated for neuroblastoma at a specialized children's hospital with pre-treatment contrast-enhanced CT and MYCN status were identified (training cohort, n = 44; testing cohort, n = 10). Six morphologic features and 107 quantitative gray-level texture radiomics features extracted from manually drawn volume-of-interest were analyzed. Following feature selection and class balancing, the final predictive model was developed with eXtreme Gradient Boosting (XGBoost) algorithm. Accumulated local effects (ALE) plots were used to explore main effects of the predictive features. Tumor texture maps were also generated for visualization of radiomics features.

RESULTS

One morphologic and 2 radiomics features were selected for model building. The XGBoost model from the training cohort yielded an area under the receiver operating characteristics curve (AUC-ROC) of 0.930 (95% CI, 0.85-1.00), optimized F1-score of 0.878, and Matthews correlation coefficient (MCC) of 0.773. Evaluation on the testing cohort returned AUC-ROC of 0.880 (95% CI, 0.64-1.00), optimized F1-score of 0.933, and MCC of 0.764. ALE plots and texture maps showed higher "GreyLevelNonUniformity" values, lower "Strength" values, and higher number of image-defined risk factors contribute to higher predicted probability of MYCN amplification.

CONCLUSION

The machine learning model reliably classified MYCN amplification in pediatric neuroblastoma and shows potential as a surrogate imaging biomarker.

The impact of inter-observer variation in delineation on robustness of radiomics features in non-small cell lung cancer

Artificial intelligence and radiomics have the potential to revolutionise cancer prognostication and personalised treatment. Manual outlining of the tumour volume for extraction of radiomics features (RF) is a subjective process. This study investigates robustness of RF to inter-observer variation (IOV) in contouring in lung cancer. We utilised two public imaging datasets: ‘NSCLC-Radiomics’ and ‘NSCLC-Radiomics-Interobserver1’ (‘Interobserver’). For ‘NSCLC-Radiomics’, we created an additional set of manual contours for 92 patients, and for ‘Interobserver’, there were five manual and five semi-automated contours available for 20 patients. Dice coefficients (DC) were calculated for contours. 1113 RF were extracted including shape, first order and texture features. Intraclass correlation coefficient (ICC) was computed to assess robustness of RF to IOV. Cox regression analysis for overall survival (OS) was performed with a previously published radiomics signature. The median DC ranged from 0.81 (‘NSCLC-Radiomics’) to 0.85 (‘Interobserver’—semi-automated). The median ICC for the ‘NSCLC-Radiomics’, ‘Interobserver’ (manual) and ‘Interobserver’ (semi-automated) were 0.90, 0.88 and 0.93 respectively. The ICC varied by feature type and was lower for first order and gray level co-occurrence matrix (GLCM) features. Shape features had a lower median ICC in the ‘NSCLC-Radiomics’ dataset compared to the ‘Interobserver’ dataset. Survival analysis showed similar separation of curves for three of four RF apart from ‘original_shape_Compactness2’, a feature with low ICC (0.61). The majority of RF are robust to IOV, with first order, GLCM and shape features being the least robust. Semi-automated contouring improves feature stability. Decreased robustness of a feature is significant as it may impact upon the features’ prognostic capability.

Atlas-Based Lung Segmentation Combined With Automatic Densitometry Characterization In COVID-19 Patients: Training, Validation And First Application In A Longitudinal Study

Purpose

To develop and validate an automated segmentation tool for COVID-19 lung CTs. To combine it with densitometry information in identifying Aerated, Intermediate and Consolidated Volumes in admission (CT1) and follow up CT (CT3).

Materials and Methods

An Atlas was trained on manually segmented CT1 of 250 patients and validated on 10 CT1 of the training group, 10 new CT1 and 10 CT3, by comparing DICE index between automatic (AUTO), automatic-corrected (AUTOMAN) and manual (MAN) contours. A previously developed automatic method was applied on HU lung density histograms to quantify Aerated, Intermediate and Consolidated Volumes. Volumes of subregions in validation CT1 and CT3 were quantified for each method.

Results

In validation CT1/CT3, manual correction of automatic contours was not necessary in 40% of cases. Mean DICE values for both lungs were 0.94 for AUTOVsMAN and 0.96 for AUTOMANVsMAN. Differences between Aerated and Intermediate Volumes quantified with AUTOVsMAN contours were always < 6%. Consolidated Volumes showed larger differences (mean: −95 ± 72 cc). If considering AUTOMANVsMAN volumes, differences got further smaller for Aerated and Intermediate, and were drastically reduced for consolidated Volumes (mean: −36 ± 25 cc). The average time for manual correction of automatic lungs contours on CT1 was 5 ± 2 min.

Conclusions

An Atlas for automatic segmentation of lungs in COVID-19 patients was developed and validated. Combined with a previously developed method for lung densitometry characterization, it provides a fast, operator-independent way to extract relevant quantitative parameters with minimal manual intervention.

Building reliable radiomic models using image perturbation

Radiomic model reliability is a central premise for its clinical translation. Presently, it is assessed using test–retest or external data, which, unfortunately, is often scarce in reality. Therefore, we aimed to develop a novel image perturbation-based method (IPBM) for the first of its kind toward building a reliable radiomic model. We first developed a radiomic prognostic model for head-and-neck cancer patients on a training (70%) and evaluated on a testing (30%) cohort using C-index. Subsequently, we applied the IPBM to CT images of both cohorts (Perturbed-Train and Perturbed-Test cohort) to generate 60 additional samples for both cohorts. Model reliability was assessed using intra-class correlation coefficient (ICC) to quantify consistency of the C-index among the 60 samples in the Perturbed-Train and Perturbed-Test cohorts. Besides, we re-trained the radiomic model using reliable RFs exclusively (ICC > 0.75) to validate the IPBM. Results showed moderate model reliability in Perturbed-Train (ICC: 0.565, 95%CI 0.518–0.615) and Perturbed-Test (ICC: 0.596, 95%CI 0.527–0.670) cohorts. An enhanced reliability of the re-trained model was observed in Perturbed-Train (ICC: 0.782, 95%CI 0.759–0.815) and Perturbed-Test (ICC: 0.825, 95%CI 0.782–0.867) cohorts, indicating validity of the IPBM. To conclude, we demonstrated capability of the IPBM toward building reliable radiomic models, providing community with a novel model reliability assessment strategy prior to prospective evaluation.

Development of a robust radiomic biomarker of progression-free survival in advanced non-small cell lung cancer patients treated with first-line immunotherapy

We aim to determine the feasibility of a novel radiomic biomarker that can integrate with other established clinical prognostic factors to predict progression-free survival (PFS) in patients with non-small cell lung cancer (NSCLC) undergoing first-line immunotherapy. Our study includes 107 patients with stage 4 NSCLC treated with pembrolizumab-based therapy (monotherapy: 30%, combination chemotherapy: 70%). The ITK-SNAP software was used for 3D tumor volume segmentation from pre-therapy CT scans. Radiomic features (n = 102) were extracted using the CaPTk software. Impact of heterogeneity introduced by image physical dimensions (voxel spacing parameters) and acquisition parameters (contrast enhancement and CT reconstruction kernel) was mitigated by resampling the images to the minimum voxel spacing parameters and harmonization by a nested ComBat technique. This technique was initialized with radiomic features, clinical factors of age, sex, race, PD-L1 expression, ECOG status, body mass index (BMI), smoking status, recurrence event and months of progression-free survival, and image acquisition parameters as batch variables. Two phenotypes were identified using unsupervised hierarchical clustering of harmonized features. Prognostic factors, including PDL1 expression, ECOG status, BMI and smoking status, were combined with radiomic phenotypes in Cox regression models of PFS and Kaplan Meier (KM) curve-fitting. Cox model based on clinical factors had a c-statistic of 0.57, which increased to 0.63 upon addition of phenotypes derived from harmonized features. There were statistically significant differences in survival outcomes stratified by clinical covariates, as measured by the log-rank test (p = 0.034), which improved upon addition of phenotypes (p = 0.00022). We found that mitigation of heterogeneity by image resampling and nested ComBat harmonization improves prognostic value of phenotypes, resulting in better prediction of PFS when added to other prognostic variables.

Radiomic Analysis: Study Design, Statistical Analysis, and Other Bias Mitigation Strategies

Rapid advances in automated methods for extracting large numbers of quantitative features from medical images have led to tremendous growth of publications reporting on radiomic analyses. Translation of these research studies into clinical practice can be hindered by biases introduced during the design, analysis, or reporting of the studies. Herein, the authors review biases, sources of variability, and pitfalls that frequently arise in radiomic research, with an emphasis on study design and statistical analysis considerations. Drawing on existing work in the statistical, radiologic, and machine learning literature, approaches for avoiding these pitfalls are described.

Automated segmentation of the fractured vertebrae on CT and its applicability in a radiomics model to predict fracture malignancy

Although CT radiomics has shown promising results in the evaluation of vertebral fractures, the need for manual segmentation of fractured vertebrae limited the routine clinical implementation of radiomics. Therefore, automated segmentation of fractured vertebrae is needed for successful clinical use of radiomics. In this study, we aimed to develop and validate an automated algorithm for segmentation of fractured vertebral bodies on CT, and to evaluate the applicability of the algorithm in a radiomics prediction model to differentiate benign and malignant fractures. A convolutional neural network was trained to perform automated segmentation of fractured vertebral bodies using 341 vertebrae with benign or malignant fractures from 158 patients, and was validated on independent test sets (internal test, 86 vertebrae [59 patients]; external test, 102 vertebrae [59 patients]). Then, a radiomics model predicting fracture malignancy on CT was constructed, and the prediction performance was compared between automated and human expert segmentations. The algorithm achieved good agreement with human expert segmentation at testing (Dice similarity coefficient, 0.93-0.94; cross-sectional area error, 2.66-2.97%; average surface distance, 0.40-0.54 mm). The radiomics model demonstrated good performance in the training set (AUC, 0.93). In the test sets, automated and human expert segmentations showed comparable prediction performances (AUC, internal test, 0.80 vs 0.87, p = 0.044; external test, 0.83 vs 0.80, p = 0.37). In summary, we developed and validated an automated segmentation algorithm that showed comparable performance to human expert segmentation in a CT radiomics model to predict fracture malignancy, which may enable more practical clinical utilization of radiomics.

Reproducibility for Hepatocellular Carcinoma CT Radiomic Features: Influence of Delineation Variability Based on 3D-CT, 4D-CT and Multiple-Parameter MR Images

Purpose

Accurate lesion segmentation is a prerequisite for radiomic feature extraction. It helps to reduce the features variability so as to improve the reporting quality of radiomics study. In this research, we aimed to conduct a radiomic feature reproducibility test of inter-/intra-observer delineation variability in hepatocellular carcinoma using 3D-CT images, 4D-CT images and multiple-parameter MR images.

Materials and Methods

For this retrospective study, 19 HCC patients undergoing 3D-CT, 4D-CT and multiple-parameter MR scans were included in this study. The gross tumor volume (GTV) was independently delineated twice by two observers based on contrast-enhanced computed tomography (CECT), maximum intensity projection (MIP), LAVA-Flex, T2W FRFSE and DWI-EPI images. We also delineated the peritumoral region, which was defined as 0 to 5 mm radius surrounding the GTV. 107 radiomic features were automatically extracted from CECT images using 3D-Slicer software. Quartile coefficient of dispersion (QCD) and intraclass correlation coefficient (ICC) were applied to assess the variability of each radiomic feature. QCD<10% and ICC≥0.75 were considered small variations and excellent reliability. Finally, the principal component analysis (PCA) was used to test the feasibility of dimensionality reduction.

Results

For tumor tissues, the numbers of radiomic features with QCD<10% indicated no obvious inter-/intra-observer differences or discrepancies in 3D-CT, 4D-CT and multiple-parameter MR delineation. However, the number of radiomic features (mean 89) with ICC≥0.75 was the highest in the multiple-parameter MR group, followed by the 3DCT group (mean 77) and the MIP group (mean 73). The peritumor tissues also showed similar results. A total of 15 and 7 radiomic features presented excellent reproducibility and small variation in tumor and peritumoral tissues, respectively. Two robust features showed excellent reproducibility and small variation in tumor and peritumoral tissues. In addition, the values of the two features both represented statistically significant differences among tumor and peritumoral tissues (P<0.05). The PCA results indicated that the first seven principal components could preserve at least 90% of the variance of the original set of features.

Conclusion

Delineation on multiple-parameter MR images could help to improve the reproducibility of the HCC CT radiomic features and weaken the inter-/intra-observer influence.

HeLLePhant: A phantom mimicking non-small cell lung cancer for texture analysis in CT images

PURPOSE

Phantoms mimicking human tissue heterogeneity and intensity are required to establish radiomic features robustness in Computed Tomography (CT) images. We developed inserts with two different techniques for the radiomic study of Non-Small Cell Lung Cancer (NSCLC) lesions.

METHODS

We developed two insert prototypes: two 3D-printed made of glycol-modified polyethylene terephthalate (PET-G), and nine with sodium polyacrylate plus iodinated contrast medium. The inserts were put in a handcraft phantom (HeLLePhant). We also analysed four materials of a commercial homogeneous phantom (Catphan® 424) and collected 29 NSCLC patients for comparison. All the CT acquisitions were performed with the same clinical protocol and scanner at 120kVp. The HeLLePhant phantom was scanned ten times in fixed condition at 120kVp and 100kVp for repeatability investigation. We extracted 153 radiomic features using Pyradiomics. To compare the features between phantoms and patients, we computed how many phantom features fell in the range between 10th and 90th percentile of the corresponding patient values. We deemed repeatable the features with a coefficient of variation (CV) less than or equal to 0.10.

RESULTS

The best similarity with the patients was obtained with the polyacrylate inserts (55.6-90.2%), the worst with Catphan (15.7-19.0%). For the PET-G inserts 35.3% and 36.6% of the features match the patient range. We found high repeatability for all the inserts of the HeLLePhant phantom (74.3-100% at 120kVp, 75.7-97.9% at 100kVp), and observed a texture dependency in repeatability.

CONCLUSIONS

Our study shows a promising way to construct heterogeneous inserts mimicking a target tissue for radiomic studies.

Radiomics Analysis for Predicting Malignant Potential of Intraductal Papillary Mucinous Neoplasms of the Pancreas: Comparison of CT and MRI

RATIONALE AND OBJECTIVES

To compare the performance of CT and MRI radiomics for predicting the malignant potential of intraductal papillary mucinous neoplasms (IPMNs) of the pancreas, and to investigate their value compared to the revised 2017 international consensus Fukuoka guidelines.

MATERIALS AND METHODS

Sixty patients with surgically confirmed IPMNs (37 malignant and 23 benign) were included. Radiomics features were extracted from arterial and venous phase images of CT and T2-weighted images of MRI, respectively. Intraclass correlation coefficients for the radiomics features were calculated to assess the interobserver reproducibility. The least absolute shrinkage and selection operator algorithm was used for feature selection. Radiomics models were constructed based on selected features with logistic regression (LR) and support vector machine (SVM). A clinical and imaging model was constructed based on independent predictors of the revised 2017 Fukuoka guidelines determined in multivariate logistic regression with forward elimination.

RESULTS

The reproducibility of MRI radiomics features was higher than that of CT radiomics features, regardless of arterial or venous phase features (all p < 0.001). MRI radiomics models achieved improved AUCs (0.879 with LR and 0.940 with SVM, respectively), than that of CT radiomics models (0.811 with LR and 0.864 with SVM, respectively). All radiomics models provided better predictive performance than the clinical and imaging model (AUC = 0.764).

CONCLUSION

The MRI radiomics models with higher reproducibility radiomics features performed better than CT radiomics models for predicting the malignant potential of IPMNs. The performance of radiomics models was superior to the clinical and imaging model based on Fukuoka guidelines.

Deep Learning With Radiomics for Disease Diagnosis and Treatment: Challenges and Potential

The high-throughput extraction of quantitative imaging features from medical images for the purpose of radiomic analysis, i.e., radiomics in a broad sense, is a rapidly developing and emerging research field that has been attracting increasing interest, particularly in multimodality and multi-omics studies. In this context, the quantitative analysis of multidimensional data plays an essential role in assessing the spatio-temporal characteristics of different tissues and organs and their microenvironment. Herein, recent developments in this method, including manually defined features, data acquisition and preprocessing, lesion segmentation, feature extraction, feature selection and dimension reduction, statistical analysis, and model construction, are reviewed. In addition, deep learning-based techniques for automatic segmentation and radiomic analysis are being analyzed to address limitations such as rigorous workflow, manual/semi-automatic lesion annotation, and inadequate feature criteria, and multicenter validation. Furthermore, a summary of the current state-of-the-art applications of this technology in disease diagnosis, treatment response, and prognosis prediction from the perspective of radiology images, multimodality images, histopathology images, and three-dimensional dose distribution data, particularly in oncology, is presented. The potential and value of radiomics in diagnostic and therapeutic strategies are also further analyzed, and for the first time, the advances and challenges associated with dosiomics in radiotherapy are summarized, highlighting the latest progress in radiomics. Finally, a robust framework for radiomic analysis is presented and challenges and recommendations for future development are discussed, including but not limited to the factors that affect model stability (medical big data and multitype data and expert knowledge in medical), limitations of data-driven processes (reproducibility and interpretability of studies, different treatment alternatives for various institutions, and prospective researches and clinical trials), and thoughts on future directions (the capability to achieve clinical applications and open platform for radiomics analysis).

Cleaning Radiotherapy Contours for Radiomics Studies, is it Worth it? A Head and Neck Cancer Study

•

PET images features are more stable across different delineation of the same target.

•

Shape family features are more stable.

•

The survival model based on Dedicated contours achieved better performance for predicting PFS.

A vast majority of studies in the radiomics field are based on contours originating from radiotherapy planning. This kind of delineation (e.g. Gross Tumor Volume, GTV) is often larger than the true tumoral volume, sometimes including parts of other organs (e.g. trachea in Head and Neck, H&N studies) and the impact of such over-segmentation was little investigated so far. In this paper, we propose to evaluate and compare the performance between models using two contour types: those from radiotherapy planning, and those specifically delineated for radiomics studies. For the latter, we modified the radiotherapy contours to fit the true tumoral volume. The two contour types were compared when predicting Progression-Free Survival (PFS) using Cox models based on radiomics features extracted from FluoroDeoxyGlucose-Positron Emission Tomography (FDG-PET) and CT images of 239 patients with oropharyngeal H&N cancer collected from five centers, the data from the 2020 HECKTOR challenge. Using Dedicated contours demonstrated better performance for predicting PFS, where Harell’s concordance indices of 0.61 and 0.69 were achieved for Radiotherapy and Dedicated contours, respectively. Using automatically Resegmented contours based on a fixed intensity range was associated with a C-index of 0.63. These results illustrate the importance of using clean dedicated contours that are close to the true tumoral volume in radiomics studies, even when tumor contours are already available from radiotherapy treatment planning

Magnetic Resonance Imaging-Based Radiomics for the Prediction of Progression-Free Survival in Patients with Nasopharyngeal Carcinoma: A Systematic Review and Meta-Analysis

Simple Summary

More than 70% of patients with nasopharyngeal carcinoma (NPC) present with a locoregionally advanced state. Although the initial staging of NPC is primarily based on TNM staging, there is currently no well-established prognostic marker for NPC. Recently, radiomics has received considerable research attention as a potential prognostic biomarker for NPC. The aim of this systematic review and meta-analysis was to comprehensively evaluate the prognostic value of pretreatment magnetic resonance imaging (MRI)-based radiomics for NPC. The analyzed radiomic models demonstrated modest prognostic values, with a pooled mean estimated Harrell’s concordance index (C index) of 0.762. The prognostic models developed using more than eight radiomic features had significantly higher C-indices than those developed using fewer features. Our findings provide evidence that MRI-based radiomics may have a modest prognostic role in the treatment of NPC. However, more consistent study protocols are needed to verify the generalizability of radiomics.

Abstract

Advanced non-metastatic nasopharyngeal carcinoma (NPC) has variable treatment outcomes. However, there are no prognostic biomarkers for identifying high-risk patients with NPC. The aim of this systematic review and meta-analysis was to comprehensively assess the prognostic value of magnetic resonance imaging (MRI)-based radiomics for untreated NPC. The PubMed-Medline and EMBASE databases were searched for relevant articles published up to 12 August 2021. The Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) checklist was used to determine the qualities of the selected studies. Random-effects modeling was used to calculate the pooled estimates of Harrell’s concordance index (C-index) for progression-free survival (PFS). Between-study heterogeneity was evaluated using Higgins’ inconsistency index (I²). Among the studies reported in the 57 articles screened, 10 with 3458 patients were eligible for qualitative and quantitative data syntheses. The mean adherence rate to the TRIPOD checklist was 68.6 ± 7.1%. The pooled estimate of the C-index was 0.762 (95% confidence interval, 0.687–0.837). Substantial between-study heterogeneity was observed (I² = 89.2%). Overall, MRI-based radiomics shows good prognostic performance in predicting the PFS of patients with untreated NPC. However, more consistent and robust study protocols are necessary to validate the prognostic role of radiomics for NPC.

Reproducibility of radiomic features in CT images of NSCLC patients: an integrative analysis on the impact of acquisition and reconstruction parameters

BACKGROUND

We investigated to what extent tube voltage, scanner model, and reconstruction algorithm affect radiomic feature reproducibility in a single-institution retrospective database of computed tomography images of non-small-cell lung cancer patients.

METHODS

This study was approved by the Institutional Review Board (UID 2412). Images of 103 patients were considered, being acquired on either among two scanners, at 100 or 120 kVp. For each patient, images were reconstructed with six iterative blending levels, and 1414 features were extracted from each reconstruction. At univariate analysis, Wilcoxon-Mann-Whitney test was applied to evaluate feature differences within scanners and voltages, whereas the impact of the reconstruction was established with the overall concordance correlation coefficient (OCCC). A multivariable mixed model was also applied to investigate the independent contribution of each acquisition/reconstruction parameter. Univariate and multivariable analyses were combined to analyse feature behaviour.

RESULTS

Scanner model and voltage did not affect features significantly. The reconstruction blending level showed a significant impact at both univariate analysis (154/1414 features yielding an OCCC < 0.85) and multivariable analysis, with most features (1042/1414) revealing a systematic trend with the blending level (multiple comparisons adjusted p < 0.05). Reproducibility increased in association to image processing with smooth filters, nonetheless specific investigation in relation to clinical endpoints should be performed to ensure that textural information is not removed.

CONCLUSIONS

Combining univariate and multivariable models is allowed to identify features for which corrections may be applied to reduce the trend with the algorithm and increase reproducibility. Subsequent clustering may be applied to eliminate residual redundancy.

Preoperative CT-based peritumoral and tumoral radiomic features prediction for tumor spread through air spaces in clinical stage I lung adenocarcinoma

OBJECTIVES

This study aims to develop and evaluate preoperative CT-based peritumoral and tumoral radiomic features to predict tumor spread through air space (STAS) status in clinical stage I lung adenocarcinoma (LUAD).

MATERIALS AND METHODS

From June 2018 to December 2019, a retrospective diagnostic investigation was done. Patients with pathologically confirmed STAS status (N = 256) were eventually enrolled. The development cohort consisted of 191 patients (74.6%) chosen randomly in a 7:3 ratio, whereas the validation group consisted of 65 patients (25.4%). The performance of models was assessed using receiver operating characteristic analysis, accuracy, sensitivity, specificity, negative predictive values, and positive predictive values.

RESULTS

The STAS positive status was found in 85 (33.2%) of the 256 patients (female: 53.2%; median [IQR] age: 62.0, [53.0-79.0] years), while the STAS negative status was found in 171 patients (66.8%) (female:50.6%; median [IQR] age: 62.0, [53.0-87.0] years). The combined TRS and PRS-15 mm model had an AUC of 0.854 (95% CI, 0.799-0.909) in the development cohort and 0.870 (95% CI, 0.781-0.958) in the validation cohort, indicating that the tumor radiomic signature (TRS) model and different peritumoral radiomic signature (PRS) models were used to build the optimal gross radiomic signature (GRS) model. The radiomic nomogram achieves superior discriminatory performance than GRS and clinical and radiological signatures (CRS), with an AUC of 0.871 (95% CI, 0.820-0.922) in the development cohort and AUC of 0.869 (95% CI, 0.776-0.961) in the validation cohort. Based on the Akaike information criterion (AIC) and decision curve analysis (DCA), the radiomic nomogram provided greater clinical predictive capacity than clinical or any radiomic signatures alone.

CONCLUSION

In conclusion, we discovered that peritumoral characteristics were substantially related to STAS status. This study revealed the unit of radiomic signature and clinical signatures may have a better performance in STAS status.

Assessment of CT to CBCT contour mapping for radiomic feature analysis in prostate cancer

This study provides a quantitative assessment of the accuracy of a commercially available deformable image registration (DIR) algorithm to automatically generate prostate contours and additionally investigates the robustness of radiomic features to differing contours. Twenty-eight prostate cancer patients enrolled on an institutional review board (IRB) approved protocol were selected. Planning CTs (pCTs) were deformably registered to daily cone-beam CTs (CBCTs) to generate prostate contours (auto contours). The prostate contours were also manually drawn by a physician. Quantitative assessment of deformed versus manually drawn prostate contours on daily CBCT images was performed using Dice similarity coefficient (DSC), mean distance-to-agreement (MDA), difference in center-of-mass position (ΔCM) and difference in volume (ΔVol). Radiomic features from 6 classes were extracted from each contour. Lin’s concordance correlation coefficient (CCC) and mean absolute percent difference in radiomic feature-derived data (mean |%Δ|RF) between auto and manual contours were calculated. The mean (± SD) DSC, MDA, ΔCM and ΔVol between the auto and manual prostate contours were 0.90 ± 0.04, 1.81 ± 0.47 mm, 2.17 ± 1.26 mm and 5.1 ± 4.1% respectively. Of the 1,010 fractions under consideration, 94.8% of DIRs were within TG-132 recommended tolerance. 30 radiomic features had a CCC > 0.90 and 21 had a mean |%∆|RF < 5%. Auto-propagation of prostate contours resulted in nearly 95% of DIRs within tolerance recommendations of TG-132, leading to the majority of features being regarded as acceptably robust. The use of auto contours for radiomic feature analysis is promising but must be done with caution.

Robustness and performance of radiomic features in diagnosing cystic renal masses

PURPOSE

We study the inter-reader variability in manual delineation of cystic renal masses (CRMs) presented in computerized tomography (CT) images and its effect on the classification performance of a machine learning algorithm in distinguishing benign from potentially malignant CRMs. In addition, we assessed whether the inclusion of higher-order robust radiomic features improves the classification performance over the use of first-order features.

METHODS

230 CRMs were independently delineated by two radiologists. Through a combination of random fluctuations, dilation, and erosion operations over the original region of interests (ROIs), we generated four additional sets of synthetic ROIs to capture the inter-reader variability realistically, as confirmed by dice coefficient measurements and visual assessment. We then identified the robust features based on the intra-class coefficient (ICC > 0.85) across these datasets. We applied a tenfold stratified cross-validation (CV) to train and test the performance of the random forest model for the classification of CRMs into benign and potentially malignant.

RESULTS

The mean area under the curve (AUC), sensitivity, specificity, positive predictive value, and negative predictive value were 0.87, 0.82, 0.90, 0.85, and 0.93, respectively. With the usage of first-order features alone, the corresponding values were nearly identical.

CONCLUSION

AUC ranged for the robust and uncorrelated features from 0.83 ± 0.09 to 0.93 ± 0.04 and for the first-order features from 0.84 ± 0.09 to 0.91 ± 0.04. Our study indicates that the first-order features alone are sufficient for the classification of CRMs, and that inclusion of higher-order features does not necessarily improve performance.

Radiomics in breast MRI: current progress toward clinical application in the era of artificial intelligence

Breast magnetic resonance imaging (MRI) is the most sensitive imaging modality for breast cancer diagnosis and is widely used clinically. Dynamic contrast-enhanced MRI is the basis for breast MRI, but ultrafast images, T2-weighted images, and diffusion-weighted images are also taken to improve the characteristics of the lesion. Such multiparametric MRI with numerous morphological and functional data poses new challenges to radiologists, and thus, new tools for reliable, reproducible, and high-volume quantitative assessments are warranted. In this context, radiomics, which is an emerging field of research involving the conversion of digital medical images into mineable data for clinical decision-making and outcome prediction, has been gaining ground in oncology. Recent development in artificial intelligence has promoted radiomics studies in various fields including breast cancer treatment and numerous studies have been conducted. However, radiomics has shown a translational gap in clinical practice, and many issues remain to be solved. In this review, we will outline the steps of radiomics workflow and investigate clinical application of radiomics focusing on breast MRI based on published literature, as well as current discussion about limitations and challenges in radiomics.

Radiomic machine learning for pretreatment assessment of prognostic risk factors for endometrial cancer and its effects on radiologists' decisions of deep myometrial invasion

PURPOSE

To evaluate radiomic machine learning (ML) classifiers based on multiparametric magnetic resonance images (MRI) in pretreatment assessment of endometrial cancer (EC) risk factors and to examine effects on radiologists' interpretation of deep myometrial invasion (dMI).

METHODS

This retrospective study examined 200 consecutive patients with EC during January 2004 -March 2017, divided randomly to Discovery (n = 150) and Test (n = 50) datasets. Radiomic features of tumors were extracted from T2-weighted images, apparent diffusion coefficient map, and contrast enhanced T1-weighed images. Using the Discovery dataset, feature selection and hyperparameter tuning for XGBoost were performed. Ten classifiers were built to predict dMI, histological grade, lymphovascular invasion (LVI), and pelvic/paraaortic lymph node metastasis (PLNM/PALNM), respectively. Using the Test dataset, the diagnostic performances of ten classifiers were assessed by the area under the receiver operator characteristic curve (AUC). Next, four radiologists assessed dMI independently using MRI with a Likert scale before and after referring to inference of the ML classifier for the Test dataset. Then, AUCs obtained before and after reference were compared.

RESULTS

In the Test dataset, mean AUC of ML classifiers for dMI, histological grade, LVI, PLNM, and PALNM were 0.83, 0.77, 0.81, 0.72, and 0.82. AUCs of all radiologists for dMI (0.83-0.88) were better than or equal to mean AUC of the ML classifier, which showed no statistically significant difference before and after the reference.

CONCLUSION

Radiomic classifiers showed promise for pretreatment assessment of EC risk factors. Radiologists' inferences outperformed the ML classifier for dMI and showed no improvement by review.

A deep learning- and CT image-based prognostic model for the prediction of survival in non-small cell lung cancer

OBJECTIVE

To assist clinicians in arranging personalized treatment, planning follow-up programs and extending survival times for non-small cell lung cancer (NSCLC) patients, a method of deep learning combined with computed tomography (CT) imaging for survival prediction was designed.

METHODS

Data were collected from 484 patients from four research centers. The data from 344 patients were utilized to build the A_CNN survival prognosis model to classify 2-year overall survival time ranges (730 days cut-off). Data from 140 patients, including independent internal and external test sets, were utilized for model testing. First, a series of preprocessing techniques were used to process the original CT images and generate training and test data sets from the axial, coronal, and sagittal planes. Second, the structure of the A_CNN model was designed based on asymmetric convolution, bottleneck blocks, the uniform cross-entropy (UC) loss function, and other advanced techniques. After that, the A_CNN model was trained, and numerous comparative experiments were designed to obtain the best prognostic survival model. Last, the model performance was evaluated, and the predicted survival curves were analyzed.

RESULTS

The A_CNN survival prognosis model yielded a high patient-level accuracy of 88.8%, a patch-level accuracy of 82.9%, and an area under the receiver operating characteristic (ROC) curve (AUC) of 0.932. When tested on an external data set, the maximum patient-level accuracy was 80.0%.

CONCLUSIONS

The results suggest that using a deep learning method can improve prognosis in patients with NSCLC and has important application value in establishing individualized prognostic models.

Radiomics feature reliability assessed by intraclass correlation coefficient: a systematic review

Radiomics research is rapidly growing in recent years, but more concerns on radiomics reliability are also raised. This review attempts to update and overview the current status of radiomics reliability research in the ever expanding medical literature from the perspective of a single reliability metric of intraclass correlation coefficient (ICC). To conduct this systematic review, Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were followed. After literature search and selection, a total of 481 radiomics studies using CT, PET, or MRI, covering a wide range of subject and disease types, were included for review. In these highly heterogeneous studies, feature reliability to image segmentation was much more investigated than reliability to other factors, such as image acquisition, reconstruction, post-processing, and feature quantification. The reported ICCs also suggested high radiomics feature reliability to image segmentation. Image acquisition was found to introduce much more feature variability than image segmentation, in particular for MRI, based on the reported ICC values. Image post-processing and feature quantification yielded different levels of radiomics reliability and might be used to mitigate image acquisition-induced variability. Some common flaws and pitfalls in ICC use were identified, and suggestions on better ICC use were given. Due to the extremely high study heterogeneities and possible risks of bias, the degree of radiomics feature reliability that has been achieved could not yet be safely synthesized or derived in this review. More future researches on radiomics reliability are warranted.

Prediction of Clinical Outcome for High-Intensity Focused Ultrasound Ablation of Uterine Leiomyomas Using Multiparametric MRI Radiomics-Based Machine Leaning Model

Objectives

This study sought to develop a multiparametric MRI radiomics-based machine learning model for the preoperative prediction of clinical success for high-intensity-focused ultrasound (HIFU) ablation of uterine leiomyomas.

Methods

One hundred and thirty patients who received HIFU ablation therapy for uterine leiomyomas were enrolled in this retrospective study. Radiomics features were extracted from T2-weighted (T2WI) image and ADC map derived from diffusion-weighted imaging (DWI). Three feature selection algorithms including least absolute shrinkage and selection operator (LASSO), recursive feature elimination (RFE), and ReliefF algorithm were used to select radiomics features, respectively, which were fed into four machine learning classifiers including k-nearest neighbors (KNN), logistic regression (LR), random forest (RF), and support vector machine (SVM) for the construction of outcome prediction models before HIFU treatment. The performance, predication ability, and clinical usefulness of these models were verified and evaluated using receiver operating characteristics (ROC), calibration, and decision curve analyses.

Results

The radiomics analysis provided an effective preoperative prediction for HIFU ablation of uterine leiomyomas. Using SVM with ReliefF algorithm, the multiparametric MRI radiomics model showed the favorable performance with average accuracy of 0.849, sensitivity of 0.814, specificity of 0.896, positive predictive value (PPV) of 0.903, negative predictive value (NPV) of 0.823, and the area under the ROC curve (AUC) of 0.887 (95% CI = 0.848-0.939) in fivefold cross-validation, followed by RF with ReliefF. Calibration and decision curve analyses confirmed the potential of model in predication ability and clinical usefulness.

Conclusions

The radiomics-based machine learning model can predict preoperatively HIFU ablation response for the patients with uterine leiomyomas and contribute to determining individual treatment strategies.

Nonenhanced MRI-based radiomics model for preoperative prediction of nonperfused volume ratio for high-intensity focused ultrasound ablation of uterine leiomyomas

OBJECTIVES

To develop and assess nonenhanced MRI-based radiomics model for the preoperative prediction of nonperfused volume (NPV) ratio of uterine leiomyomas after high-intensity focused ultrasound (HIFU) treatment.

METHODS

Two hundred and five patients with uterine leiomyomas treated by HIFU were enrolled and allocated to training (N =164) and testing cohorts (N = 41). Pyradiomics was used to extract radiomics features from T2-weighted images and apparent diffusion coefficient (ADC) map generated from diffusion-weighted imaging (DWI). The clinico-radiological model, radiomics model, and radiomics-clinical model which combined the selected radiomics features and clinical parameters were used to predict technical outcomes determined by NPV ratios where three classification groups were created (NPV ratio ≤ 50%, 50-80% or ≥ 80%). The receiver operating characteristic (ROC) curve, area under the curve (AUC), and calibration and decision curve analyses were performed to illustrate the prediction performance and clinical usefulness of model in the training and testing cohorts.

RESULTS

The multi-parametric MRI-based radiomics model outperformed T2-weighted imaging (T2WI)-based radiomics model, which achieved an average AUC of 0.769 (95% confidence interval [CI], 0.701-0.842), and showed satisfactory prediction performance for NPV ratio classification. The radiomics-clinical model demonstrated best prediction performance for HIFU treatment outcome, with an average AUC of 0.802 (95% CI, 0.796-0.850) and an accuracy of 0.762 (95% CI, 0.698-0.815) in the testing cohort, compared to the clinico-radiological and radiomics models. The decision curve also indicated favorable clinical usefulness of the radiomics-clinical model.

CONCLUSIONS

Nonenhanced MRI-based radiomics has potential in the preoperative prediction of NPV ratio for HIFU ablation of uterine leiomyomas.

An Automatic Deep Learning-Based Workflow for Glioblastoma Survival Prediction Using Preoperative Multimodal MR Images: A Feasibility Study

PURPOSE

Most radiomic studies use the features extracted from the manually drawn tumor contours for classification or survival prediction. However, large interobserver segmentation variations lead to inconsistent features and hence introduce more challenges in constructing robust prediction models. Here, we proposed an automatic workflow for glioblastoma (GBM) survival prediction based on multimodal magnetic resonance (MR) images.

METHODS AND MATERIALS

Two hundred eighty-five patients with glioma (210 GBM, 75 low-grade glioma) were included. One hundred sixty-three of the patients with GBM had overall survival data. Every patient had 4 preoperative MR images and manually drawn tumor contours. A 3-dimensional convolutional neural network, VGG-Seg, was trained and validated using 122 patients with glioma for automatic GBM segmentation. The trained VGG-Seg was applied to the remaining 163 patients with GBM to generate their autosegmented tumor contours. The handcrafted and deep learning (DL)-based radiomic features were extracted from the autosegmented contours using explicitly designed algorithms and a pretrained convolutional neural network, respectively. One hundred sixty-three patients with GBM were randomly split into training (n = 122) and testing (n = 41) sets for survival analysis. Cox regression models were trained to construct the handcrafted and DL-based signatures. The prognostic powers of the 2 signatures were evaluated and compared.

RESULTS

The VGG-Seg achieved a mean Dice coefficient of 0.86 across 163 patients with GBM for GBM segmentation. The handcrafted signature achieved a C-index of 0.64 (95% confidence interval, 0.55-0.73), whereas the DL-based signature achieved a C-index of 0.67 (95% confidence interval, 0.57-0.77). Unlike the handcrafted signature, the DL-based signature successfully stratified testing patients into 2 prognostically distinct groups.

CONCLUSIONS

The VGG-Seg generated accurate GBM contours from 4 MR images. The DL-based signature achieved a numerically higher C-index than the handcrafted signature and significant patient stratification. The proposed automatic workflow demonstrated the potential of improving patient stratification and survival prediction in patients with GBM.

Prognostic role of pre-treatment magnetic resonance imaging (MRI)-based radiomic analysis in effectively cured head and neck squamous cell carcinoma (HNSCC) patients

OBJECTIVES

To identify and validate baseline magnetic resonance imaging (b-MRI) radiomic features (RFs) as predictors of disease outcomes in effectively cured head and neck squamous cell carcinoma (HNSCC) patients.

MATERIALS AND METHODS

Training set (TS) and validation set (VS) were retrieved from preexisting datasets (HETeCo and BD2Decide trials, respectively). Only patients with both pre- and post-contrast enhancement T1 and T2-weighted b-MRI and at least 2 years of follow-up (FUP) were selected. The combination of the best extracted RFs was used to classify low risk (LR) vs. high risk (HR) of disease recurrence. Sensitivity, specificity, and area under the curve (AUC) of the radiomic model were computed on both TS and VS. Overall survival (OS) and 5-year disease-free survival (DFS) Kaplan-Meier (KM) curves were compared for LR vs. HR. The radiomic-based risk class was used in a multivariate Cox model, including well-established clinical prognostic factors (TNM, sub-site, human papillomavirus [HPV]).

RESULTS

In total, 57 patients of TS and 137 of VS were included. Three RFs were selected for the signature. Sensitivity of recurrence risk classifier was 0.82 and 0.77, specificity 0.78 and 0.81, AUC 0.83 and 0.78 for TS and VS, respectively. VS KM curves for LR vs. HR groups significantly differed both for 5-year DFS (p<.0001) and OS (p=.0004). A combined model of RFs plus TNM improved prognostic performance as compared to TNM alone, both for VS 5-year DFS (C-index: 0.76 vs. 0.60) and OS (C-index: 0.74 vs. 0.64).

CONCLUSIONS

Radiomics of b-MRI can help to predict recurrence and survival outcomes in HNSCC.