Impact of rescanning and repositioning on radiomic features employing a multi-object phantom in magnetic resonance imaging

Does Dual-Energy Computed Tomography Material Decomposition Improve Radiomics Capability to Predict Survival in Head and Neck Squamous Cell Carcinoma Patients? A Preliminary Investigation

OBJECTIVE

Our study objective was to explore the additional value of dual-energy CT (DECT) material decomposition for squamous cell carcinoma of the head and neck (SCCHN) survival prognostication.

METHODS

A group of 50 SCCHN patients (male, 37; female, 13; mean age, 63.6 ± 10.82 years) with baseline head and neck DECT between September 2014 and August 2020 were retrospectively included. Primary tumors were segmented, radiomics features were extracted, and DECT material decomposition was performed. We used independent train and validation datasets with cross-validation and 100 independent iterations to identify prognostic signatures applying elastic net (EN) and random survival forest (RSF). Features were ranked and intercorrelated according to their prognostic importance. We benchmarked the models against clinical parameters. Intraclass correlation coefficients were used to analyze the interreader variation.

RESULTS

The exclusively radiomics-trained models achieved similar ( P = 0.947) prognostic performance of area under the curve (AUC) = 0.784 (95% confidence interval [CI], 0.775-0.812) (EN) and AUC = 0.785 (95% CI, 0.759-0.812) (RSF). The additional application of DECT material decomposition did not improve the model's performance (EN, P = 0.594; RSF, P = 0.198). In the clinical benchmark, the top averaged AUC value of 0.643 (95% CI, 0.611-0.675) was inferior to the quantitative imaging-biomarker models ( P < 0.001). A combined imaging and clinical model did not improve the imaging-based models ( P > 0.101). Shape features revealed high prognostic importance.

CONCLUSIONS

Radiomics AI applications may be used for SCCHN survival prognostication, but the spectral information of DECT material decomposition did not improve the model's performance in our preliminary investigation.

Role of MRI radiomics for the prediction of MYCN amplification in neuroblastomas

OBJECTIVES

We evaluate MR radiomics and develop machine learning-based classifiers to predict MYCN amplification in neuroblastomas.

METHODS

A total of 120 patients with neuroblastomas and baseline MR imaging examination available were identified of whom 74 (mean age ± standard deviation [SD] of 6 years and 2 months ± 4 years and 9 months; 43 females and 31 males, 14 MYCN amplified) underwent imaging at our institution. This was therefore used to develop radiomics models. The model was tested in a cohort of children with the same diagnosis but imaged elsewhere (n = 46, mean age ± SD: 5 years 11 months ± 3 years 9 months, 26 females and 14 MYCN amplified). Whole tumour volumes of interest were adopted to extract first-order histogram and second-order radiomics features. Interclass correlation coefficient and maximum relevance and minimum redundancy algorithm were applied for feature selection. Logistic regression, support vector machine, and random forest were employed as the classifiers. Receiver operating characteristic (ROC) analysis was performed to evaluate the diagnostic accuracy of the classifiers on the external test set.

RESULTS

The logistic regression model and the random forest both showed an AUC of 0.75. The support vector machine classifier obtained an AUC of 0.78 on the test set with a sensitivity of 64% and a specificity of 72%.

CONCLUSION

The study provides preliminary retrospective evidence demonstrating the feasibility of MRI radiomics in predicting MYCN amplification in neuroblastomas. Future studies are needed to explore the correlation between other imaging features and genetic markers and to develop multiclass predictive models.

KEY POINTS

• MYCN amplification in neuroblastomas is an important determinant of disease prognosis. • Radiomics analysis of pre-treatment MR examinations can be used to predict MYCN amplification in neuroblastomas. • Radiomics machine learning models showed good generalisability to external test set, demonstrating reproducibility of the computational models.

Comparison of radiomics models and dual-energy material decomposition to decipher abdominal lymphoma in contrast-enhanced CT

PURPOSE

The radiologists' workload is increasing, and computational imaging techniques may have the potential to identify visually unequivocal lesions, so that the radiologist can focus on equivocal and critical cases. The purpose of this study was to assess radiomics versus dual-energy CT (DECT) material decomposition to objectively distinguish visually unequivocal abdominal lymphoma and benign lymph nodes.

METHODS

Retrospectively, 72 patients [m, 47; age, 63.5 (27-87) years] with nodal lymphoma (n = 27) or benign abdominal lymph nodes (n = 45) who had contrast-enhanced abdominal DECT between 06/2015 and 07/2019 were included. Three lymph nodes per patient were manually segmented to extract radiomics features and DECT material decomposition values. We used intra-class correlation analysis, Pearson correlation and LASSO to stratify a robust and non-redundant feature subset. Independent train and test data were applied on a pool of four machine learning models. Performance and permutation-based feature importance was assessed to increase the interpretability and allow for comparison of the models. Top performing models were compared by the DeLong test.

RESULTS

About 38% (19/50) and 36% (8/22) of the train and test set patients had abdominal lymphoma. Clearer entity clusters were seen in t-SNE plots using a combination of DECT and radiomics features compared to DECT features only. Top model performances of AUC = 0.763 (CI = 0.435-0.923) were achieved for the DECT cohort and AUC = 1.000 (CI = 1.000-1.000) for the radiomics feature cohort to stratify visually unequivocal lymphomatous lymph nodes. The performance of the radiomics model was significantly (p = 0.011, DeLong) superior to the DECT model.

CONCLUSIONS

Radiomics may have the potential to objectively stratify visually unequivocal nodal lymphoma versus benign lymph nodes. Radiomics seems superior to spectral DECT material decomposition in this use case. Therefore, artificial intelligence methodologies may not be restricted to centers with DECT equipment.

One Model to Synthesize Them All: Multi-Contrast Multi-Scale Transformer for Missing Data Imputation

Multi-contrast magnetic resonance imaging (MRI) is widely used in clinical practice as each contrast provides complementary information. However, the availability of each imaging contrast may vary amongst patients, which poses challenges to radiologists and automated image analysis algorithms. A general approach for tackling this problem is missing data imputation, which aims to synthesize the missing contrasts from existing ones. While several convolutional neural networks (CNN) based algorithms have been proposed, they suffer from the fundamental limitations of CNN models, such as the requirement for fixed numbers of input and output channels, the inability to capture long-range dependencies, and the lack of interpretability. In this work, we formulate missing data imputation as a sequence-to-sequence learning problem and propose a multi-contrast multi-scale Transformer (MMT), which can take any subset of input contrasts and synthesize those that are missing. MMT consists of a multi-scale Transformer encoder that builds hierarchical representations of inputs combined with a multi-scale Transformer decoder that generates the outputs in a coarse-to-fine fashion. The proposed multi-contrast Swin Transformer blocks can efficiently capture intra- and inter-contrast dependencies for accurate image synthesis. Moreover, MMT is inherently interpretable as it allows us to understand the importance of each input contrast in different regions by analyzing the in-built attention maps of Transformer blocks in the decoder. Extensive experiments on two large-scale multi-contrast MRI datasets demonstrate that MMT outperforms the state-of-the-art methods quantitatively and qualitatively.

Radiomics for therapy-specific head and neck squamous cell carcinoma survival prognostication (part I)

BACKGROUND

Treatment plans for squamous cell carcinoma of the head and neck (SCCHN) are individually decided in tumor board meetings but some treatment decision-steps lack objective prognostic estimates. Our purpose was to explore the potential of radiomics for SCCHN therapy-specific survival prognostication and to increase the models' interpretability by ranking the features based on their predictive importance.

METHODS

We included 157 SCCHN patients (male, 119; female, 38; mean age, 64.39 ± 10.71 years) with baseline head and neck CT between 09/2014 and 08/2020 in this retrospective study. Patients were stratified according to their treatment. Using independent training and test datasets with cross-validation and 100 iterations, we identified, ranked and inter-correlated prognostic signatures using elastic net (EN) and random survival forest (RSF). We benchmarked the models against clinical parameters. Inter-reader variation was analyzed using intraclass-correlation coefficients (ICC).

RESULTS

EN and RSF achieved top prognostication performances of AUC = 0.795 (95% CI 0.767-0.822) and AUC = 0.811 (95% CI 0.782-0.839). RSF prognostication slightly outperformed the EN for the complete (ΔAUC 0.035, p = 0.002) and radiochemotherapy (ΔAUC 0.092, p < 0.001) cohort. RSF was superior to most clinical benchmarking (p ≤ 0.006). The inter-reader correlation was moderate or high for all features classes (ICC ≥ 0.77 (± 0.19)). Shape features had the highest prognostic importance, followed by texture features.

CONCLUSIONS

EN and RSF built on radiomics features may be used for survival prognostication. The prognostically leading features may vary between treatment subgroups. This warrants further validation to potentially aid clinical treatment decision making in the future.

Computed Tomography Radiomics to Differentiate Intrahepatic Cholangiocarcinoma and Hepatocellular Carcinoma

AIMS

Intrahepatic cholangiocarcinoma (iCCA) and hepatocellular carcinoma (HCC) differ in prognosis and treatment. We aimed to non-invasively differentiate iCCA and HCC by means of radiomics extracted from contrast-enhanced standard-of-care computed tomography (CT).

MATERIALS AND METHODS

In total, 94 patients (male, n = 68, mean age 63.3 ± 12.4 years) with histologically confirmed iCCA (n = 47) or HCC (n = 47) who underwent contrast-enhanced abdominal CT between August 2014 and November 2021 were retrospectively included. The enhancing tumour border was manually segmented in a clinically feasible way by defining three three-dimensional volumes of interest per tumour. Radiomics features were extracted. Intraclass correlation analysis and Pearson metrics were used to stratify robust and non-redundant features with further feature reduction by LASSO (least absolute shrinkage and selection operator). Independent training and testing datasets were used to build four different machine learning models. Performance metrics and feature importance values were computed to increase the models' interpretability.

RESULTS

The patient population was split into 65 patients for training (iCCA, n = 32) and 29 patients for testing (iCCA, n = 15). A final combined feature set of three radiomics features and the clinical features age and sex revealed a top test model performance of receiver operating characteristic (ROC) area under the curve (AUC) = 0.82 (95% confidence interval =0.66-0.98; train ROC AUC = 0.82) using a logistic regression classifier. The model was well calibrated, and the Youden J Index suggested an optimal cut-off of 0.501 to discriminate between iCCA and HCC with a sensitivity of 0.733 and a specificity of 0.857.

CONCLUSIONS

Radiomics-based imaging biomarkers can potentially help to non-invasively discriminate between iCCA and HCC.

CT-radiomics and clinical risk scores for response and overall survival prognostication in TACE HCC patients

We aimed to identify hepatocellular carcinoma (HCC) patients who will respond to repetitive transarterial chemoembolization (TACE) to improve the treatment algorithm. Retrospectively, 61 patients (mean age, 65.3 years ± 10.0 [SD]; 49 men) with 94 HCC mRECIST target-lesions who had three consecutive TACE between 01/2012 and 01/2020 were included. Robust and non-redundant radiomics features were extracted from the 24 h post-embolization CT. Five different clinical TACE-scores were assessed. Seven different feature selection methods and machine learning models were used. Radiomics, clinical and combined models were built to predict response to TACE on a lesion-wise and patient-wise level as well as its impact on overall-survival prognostication. 29 target-lesions of 19 patients were evaluated in the test set. Response rates were 37.9% (11/29) on the lesion-level and 42.1% (8/19) on the patient-level. Radiomics top lesion-wise response prognostications was AUC 0.55-0.67. Clinical scores revealed top AUCs of 0.65-0.69. The best working model combined the radiomic feature LargeDependenceHighGrayLevelEmphasis and the clinical score mHAP_II_score_group with AUC = 0.70, accuracy = 0.72. We transferred this model on a patient-level to achieve AUC = 0.62, CI = 0.41-0.83. The two radiomics-clinical features revealed overall-survival prognostication of C-index = 0.67. In conclusion, a random forest model using the radiomic feature LargeDependenceHighGrayLevelEmphasis and the clinical mHAP-II-score-group seems promising for TACE response prognostication.

Radiomics in clinical trials: perspectives on standardization

The term biomarker is used to describe a biological measure of the disease behavior. The existing imaging biomarkers are associated with the known tissue biological characteristics and follow a well-established roadmap to be implemented in routine clinical practice. Recently, a new quantitative imaging analysis approach named radiomics has emerged. It refers to the extraction of a large number of advanced imaging features with high-throughput computing. Extensive research has demonstrated its value in predicting disease behavior, progression, and response to therapeutic options. However, there are numerous challenges to establishing it as a clinically viable solution, including lack of reproducibility and transparency. The data-driven nature also does not offer insights into the underpinning biology of the observed relationships. As such, additional effort is needed to establish it as a qualified biomarker to inform clinical decisions. Here we review the technical difficulties encountered in the clinical applications of radiomics and current effort in addressing some of these challenges in clinical trial designs. By addressing these challenges, the true potential of radiomics can be unleashed.

Imaging biomarkers to stratify lymph node metastases in abdominal CT - Is radiomics superior to dual-energy material decomposition?

Purpose

To assess the potential of radiomic features in comparison to dual-energy CT (DECT) material decomposition to objectively stratify abdominal lymph node metastases.

Materials and methods

In this retrospective study, we included 81 patients (m, 57; median age, 65 (interquartile range, 58.7-73.3) years) with either lymph node metastases (n = 36) or benign lymph nodes (n = 45) who underwent contrast-enhanced abdominal DECT between 06/2015-07/2019. All malignant lymph nodes were classified as unequivocal according to RECIST criteria and confirmed by histopathology, PET-CT or follow-up imaging. Three investigators segmented lymph nodes to extract DECT and radiomics features. Intra-class correlation analysis was applied to stratify a robust feature subset with further feature reduction by Pearson correlation analysis and LASSO. Independent training and testing datasets were applied on four different machine learning models. We calculated the performance metrics and permutation-based feature importance values to increase interpretability of the models. DeLong test was used to compare the top performing models.

Results

Distance matrices and t-SNE plots revealed clearer clusters using a combination of DECT and radiomic features compared to DECT features only. Feature reduction by LASSO excluded all DECT features of the combined feature cohort. The top performing radiomic features model (AUC = 1.000; F1 = 1.000; precision = 1.000; Random Forest) was significantly superior to the top performing DECT features model (AUC = 0.942; F1 = 0.762; precision = 0.800; Stochastic Gradient Boosting) (DeLong < 0.001).

Conclusion

Imaging biomarkers have the potential to stratify unequivocal lymph node metastases. Radiomics models were superior to DECT material decomposition and may serve as a support tool to facilitate stratification of abdominal lymph node metastases.

A Review of Radiomics and Artificial Intelligence and Their Application in Veterinary Diagnostic Imaging

Simple Summary

The goal of this paper is to provide an overview of current radiomic and AI applications in veterinary diagnostic imaging. We discuss the essential elements of AI for veterinary practitioners with the aim of helping them make informed decisions in applying AI technologies to their practices and that veterinarians will play an integral role in ensuring the appropriate uses and suitable curation of data. The expertise of veterinary professionals will be vital to ensuring suitable data and, subsequently, AI that meets the needs of the profession.

Abstract

Great advances have been made in human health care in the application of radiomics and artificial intelligence (AI) in a variety of areas, ranging from hospital management and virtual assistants to remote patient monitoring and medical diagnostics and imaging. To improve accuracy and reproducibility, there has been a recent move to integrate radiomics and AI as tools to assist clinical decision making and to incorporate it into routine clinical workflows and diagnosis. Although lagging behind human medicine, the use of radiomics and AI in veterinary diagnostic imaging is becoming more frequent with an increasing number of reported applications. The goal of this paper is to provide an overview of current radiomic and AI applications in veterinary diagnostic imaging.

Harmonization Strategies in Multicenter MRI-Based Radiomics

Radiomics analysis is a powerful tool aiming to provide diagnostic and prognostic patient information directly from images that are decoded into handcrafted features, comprising descriptors of shape, size and textural patterns. Although radiomics is gaining momentum since it holds great promise for accelerating digital diagnostics, it is susceptible to bias and variation due to numerous inter-patient factors (e.g., patient age and gender) as well as inter-scanner ones (different protocol acquisition depending on the scanner center). A variety of image and feature based harmonization methods has been developed to compensate for these effects; however, to the best of our knowledge, none of these techniques has been established as the most effective in the analysis pipeline so far. To this end, this review provides an overview of the challenges in optimizing radiomics analysis, and a concise summary of the most relevant harmonization techniques, aiming to provide a thorough guide to the radiomics harmonization process.

Phantom Study on the Robustness of MR Radiomics Features: Comparing the Applicability of 3D Printed and Biological Phantoms

The objectives of our study were to (a) evaluate the feasibility of using 3D printed phantoms in magnetic resonance imaging (MR) in assessing the robustness and repeatability of radiomic parameters and (b) to compare the results obtained from the 3D printed phantoms to metrics obtained in biological phantoms. To this end, three different 3D phantoms were printed: a Hilbert cube (5 × 5 × 5 cm³) and two cubic quick response (QR) code phantoms (a large phantom (large QR) (5 × 5 × 4 cm³) and a small phantom (small QR) (4 × 4 × 3 cm³)). All 3D printed and biological phantoms (kiwis, tomatoes, and onions) were scanned thrice on clinical 1.5 T and 3 T MR with 1 mm and 2 mm isotropic resolution. Subsequent analyses included analyses of several radiomics indices (RI), their repeatability and reliability were calculated using the coefficient of variation (CV), the relative percentage difference (RPD), and the interclass coefficient (ICC) parameters. Additionally, the readability of QR codes obtained from the MR images was examined with several mobile phones and algorithms. The best repeatability (CV ≤ 10%) is reported for the acquisition protocols with the highest spatial resolution. In general, the repeatability and reliability of RI were better in data obtained at 1.5 T (CV = 1.9) than at 3 T (CV = 2.11). Furthermore, we report good agreements between results obtained for the 3D phantoms and biological phantoms. Finally, analyses of the read-out rate of the QR code revealed better texture analyses for images with a spatial resolution of 1 mm than 2 mm. In conclusion, 3D printing techniques offer a unique solution to create textures for analyzing the reliability of radiomic data from MR scans.

A Pipeline for the Implementation and Visualization of Explainable Machine Learning for Medical Imaging Using Radiomics Features

Machine learning (ML) models have been shown to predict the presence of clinical factors from medical imaging with remarkable accuracy. However, these complex models can be difficult to interpret and are often criticized as “black boxes”. Prediction models that provide no insight into how their predictions are obtained are difficult to trust for making important clinical decisions, such as medical diagnoses or treatment. Explainable machine learning (XML) methods, such as Shapley values, have made it possible to explain the behavior of ML algorithms and to identify which predictors contribute most to a prediction. Incorporating XML methods into medical software tools has the potential to increase trust in ML-powered predictions and aid physicians in making medical decisions. Specifically, in the field of medical imaging analysis the most used methods for explaining deep learning-based model predictions are saliency maps that highlight important areas of an image. However, they do not provide a straightforward interpretation of which qualities of an image area are important. Here, we describe a novel pipeline for XML imaging that uses radiomics data and Shapley values as tools to explain outcome predictions from complex prediction models built with medical imaging with well-defined predictors. We present a visualization of XML imaging results in a clinician-focused dashboard that can be generalized to various settings. We demonstrate the use of this workflow for developing and explaining a prediction model using MRI data from glioma patients to predict a genetic mutation.

CT-Based Radiomics Analysis for Noninvasive Prediction of Perineural Invasion of Perihilar Cholangiocarcinoma

Purpose

The study aimed to construct and evaluate a CT-Based radiomics model for noninvasive detecting perineural invasion (PNI) of perihilar cholangiocarcinoma (pCCA) preoperatively.

Materials and Methods

From February 2012 to October 2021, a total of 161 patients with pCCA who underwent resection were retrospectively enrolled in this study. Patients were allocated into the training cohort and the validation cohort according to the diagnostic time. Venous phase images of contrast-enhanced CT were used for radiomics analysis. The intraclass correlation efficient (ICC), the correlation analysis, and the least absolute shrinkage and selection operator (LASSO) regression were applied to select radiomics features and built radiomics signature. Logistic regression analyses were performed to establish a clinical model, a radiomics model, and a combined model. The performance of the predictive models was measured by area under the receiver operating characteristic curve (AUC), and pairwise ROC comparisons between models were tested using the Delong method. Finally, the model with the best performance was presented as a nomogram, and its calibration and clinical usefulness were assessed.

Results

Finally, 15 radiomics features were selected to build a radiomics signature, and three models were developed through logistic regression. In the training cohort, the combined model showed a higher predictive capability (AUC = 0.950) than the radiomics model and the clinical model (AUC: radiomics = 0.914, clinical = 0.756). However, in the validation cohort, the AUC of the radiomics model (AUC = 0.885) was significantly higher than the other two models (AUC: combined = 0.791, clinical = 0.567). After comprehensive consideration, the radiomics model was chosen to develop the nomogram. The calibration curve and decision curve analysis (DCA) suggested that the nomogram had a good consistency and clinical utility.

Conclusion

We developed a CT-based radiomics model with good performance to noninvasively predict PNI of pCCA preoperatively.

The stability of oncologic MRI radiomic features and the potential role of deep learning: a review

The use of MRI radiomic models for the diagnosis, prognosis and treatment response prediction of tumors has been increasingly reported in literature. However, its widespread adoption in clinics is hampered by issues related to features stability. In the MRI radiomic workflow, the main factors that affect radiomic features computation can be found in the image acquisition and reconstruction phase, in the image pre-processing steps, and in the segmentation of the region of interest on which radiomic indices are extracted. Deep Neural Networks (DNNs), having shown their potentiality in the medical image processing and analysis field, can be seen as an attractive strategy to partially overcome the issues related to radiomic stability and mitigate their impact. In fact, DNN approaches can be prospectively integrated in the MRI radiomic workflow to improve image quality, obtain accurate and reproducible segmentations and generate standardized images. In this review, DNN methods that can be included in the image processing steps of the radiomic workflow are described and discussed, in the light of a detailed analysis of the literature in the context of MRI radiomic reliability.

Robustness and reproducibility of radiomics in T2 weighted images from magnetic resonance image guided linear accelerator in a phantom study

PURPOSE

Quantitative radiomics features extracted from medical images have been shown to provide value in predicting clinical outcomes. The study for robustness and reproducibility of radiomics features obtained with magnetic resonance image guided linear accelerator (MR-Linac) is insufficient. The objective of this work was to investigate the stability of radiomics features extracted from T2-weighted images of MR-Linac for five common effect factors.

MATERIALS AND METHOD

In this work, ten jellies, five fruits/vegetables, and a dynamic phantom were used to evaluate the impact of test-retest, intraobserver, varied thicknesses, radiation, and motion. These phantoms were scanned on a 1.5 T MRI system of MR-Linac. For test-retest data, the phantoms were scanned twice with repositioning within 15 min. To assess for intraobserver comparison, the segmentation of MR images was repeated by one observer in a double-blind manner. Three slice thicknesses (1.2 mm, 2.4 mm, and 4.8 mm) were used to select robust features that were insensitive to different thicknesses. The effect of radiation on features was studied by acquiring images when the beam was on. Common movement images of patients during radiotherapy were simulated by a dynamic phantom with five motion states to study the motion effect. A total of 1409 radiomics features, including shape features, first-order features, and texture features, were extracted from the original, wavelet, square, logarithmic, exponential and gradient images. The robustness and reproducibility features were evaluated using the concordance correlation coefficient (CCC).

RESULT

The intraobserver group had the most robust features (936/1079, 86.7%), while the group of motion effects had the lowest robustness (56/936, 6.0%), followed by the group of different thickness cohorts (374/936, 40.0%). The stability of features in the test-retest and radiation groups was 1072 of 1312 (81.7%) and 810 of 936 (86.5%), respectively. Overall, 25 of 1409 (2.4%) radiomics features remained robust in all five tests, mostly focusing on the image type of the wavelet. The number of stable features extracted from when the beam was on was less than that extracted when the beam was off. Shape features were the most robust of all of the features in all of the groups, excluding the motion group.

CONCLUSION

Compared with other factors fewer features remained robust to the effect of motion. This result emphasizes the need to consider the effect of respiration motion. The study for T2-weighted images from MR-Linac under different conditions will help us to build a robust predictive model applicable for radiotherapy.

Evaluation of the dependence of radiomic features on the machine learning model

BACKGROUND

In radiomic studies, several models are often trained with different combinations of feature selection methods and classifiers. The features of the best model are usually considered relevant to the problem, and they represent potential biomarkers. Features selected from statistically similarly performing models are generally not studied. To understand the degree to which the selected features of these statistically similar models differ, 14 publicly available datasets, 8 feature selection methods, and 8 classifiers were used in this retrospective study. For each combination of feature selection and classifier, a model was trained, and its performance was measured with AUC-ROC. The best-performing model was compared to other models using a DeLong test. Models that were statistically similar were compared in terms of their selected features.

RESULTS

Approximately 57% of all models analyzed were statistically similar to the best-performing model. Feature selection methods were, in general, relatively unstable (0.58; range 0.35-0.84). The features selected by different models varied largely (0.19; range 0.02-0.42), although the selected features themselves were highly correlated (0.71; range 0.4-0.92).

CONCLUSIONS

Feature relevance in radiomics strongly depends on the model used, and statistically similar models will generally identify different features as relevant. Considering features selected by a single model is misleading, and it is often not possible to directly determine whether such features are candidate biomarkers.

Radiomics in prostate cancer: an up-to-date review

Prostate cancer (PCa) is the most common worldwide diagnosed malignancy in male population. The diagnosis, the identification of aggressive disease, and the post-treatment follow-up needs a more comprehensive and holistic approach. Radiomics is the extraction and interpretation of images phenotypes in a quantitative manner. Radiomics may give an advantage through advancements in imaging modalities and through the potential power of artificial intelligence techniques by translating those features into clinical outcome prediction. This article gives an overview on the current evidence of methodology and reviews the available literature on radiomics in PCa patients, highlighting its potential for personalized treatment and future applications.

Measuring the bias of incorrect application of feature selection when using cross-validation in radiomics

BACKGROUND

Many studies in radiomics are using feature selection methods to identify the most predictive features. At the same time, they employ cross-validation to estimate the performance of the developed models. However, if the feature selection is performed before the cross-validation, data leakage can occur, and the results can be biased. To measure the extent of this bias, we collected ten publicly available radiomics datasets and conducted two experiments. First, the models were developed by incorrectly applying the feature selection prior to cross-validation. Then, the same experiment was conducted by applying feature selection correctly within cross-validation to each fold. The resulting models were then evaluated against each other in terms of AUC-ROC, AUC-F1, and Accuracy.

RESULTS

Applying the feature selection incorrectly prior to the cross-validation showed a bias of up to 0.15 in AUC-ROC, 0.29 in AUC-F1, and 0.17 in Accuracy.

CONCLUSIONS

Incorrect application of feature selection and cross-validation can lead to highly biased results for radiomic datasets.