Radiomics in nuclear medicine: robustness, reproducibility, standardization, and how to avoid data analysis traps and replication crisis

Radiomics for Identification and Prediction in Metastatic Prostate Cancer: A Review of Studies

Metastatic Prostate Cancer (mPCa) is associated with a poor patient prognosis. mPCa spreads throughout the body, often to bones, with spatial and temporal variations that make the clinical management of the disease difficult. The evolution of the disease leads to spatial heterogeneity that is extremely difficult to characterise with solid biopsies. Imaging provides the opportunity to quantify disease spread. Advanced image analytics methods, including radiomics, offer the opportunity to characterise heterogeneity beyond what can be achieved with simple assessment. Radiomics analysis has the potential to yield useful quantitative imaging biomarkers that can improve the early detection of mPCa, predict disease progression, assess response, and potentially inform the choice of treatment procedures. Traditional radiomics analysis involves modelling with hand-crafted features designed using significant domain knowledge. On the other hand, artificial intelligence techniques such as deep learning can facilitate end-to-end automated feature extraction and model generation with minimal human intervention. Radiomics models have the potential to become vital pieces in the oncology workflow, however, the current limitations of the field, such as limited reproducibility, are impeding their translation into clinical practice. This review provides an overview of the radiomics methodology, detailing critical aspects affecting the reproducibility of features, and providing examples of how artificial intelligence techniques can be incorporated into the workflow. The current landscape of publications utilising radiomics methods in the assessment and treatment of mPCa are surveyed and reviewed. Associated studies have incorporated information from multiple imaging modalities, including bone scintigraphy, CT, PET with varying tracers, multiparametric MRI together with clinical covariates, spanning the prediction of progression through to overall survival in varying cohorts. The methodological quality of each study is quantified using the radiomics quality score. Multiple deficits were identified, with the lack of prospective design and external validation highlighted as major impediments to clinical translation. These results inform some recommendations for future directions of the field.

Role of 18F-FDG PET/CT Radiomics Features in the Differential Diagnosis of Solitary Pulmonary Nodules: Diagnostic Accuracy and Comparison between Two Different PET/CT Scanners

The aim of this retrospective study was to investigate the ability of 18 fluorine-fluorodeoxyglucose positron emission tomography/CT (¹⁸F-FDG-PET/CT) metrics and radiomics features (RFs) in predicting the final diagnosis of solitary pulmonary nodules (SPN). We retrospectively recruited 202 patients who underwent a ¹⁸F-FDG-PET/CT before any treatment in two PET scanners. After volumetric segmentation of each lung nodule, 8 PET metrics and 42 RFs were extracted. All the features were tested for significant differences between the two PET scanners. The performances of all features in predicting the nature of SPN were analyzed by testing three classes of final logistic regression predictive models: two were built/trained through exploiting the separate data from the two scanners, and the other joined the data together. One hundred and twenty-seven patients had a final diagnosis of malignancy, while 64 were of a benign nature. Comparing the two PET scanners, we found that all metabolic features and most of RFs were significantly different, despite the cross correlation being quite similar. For scanner 1, a combination between grey level co-occurrence matrix (GLCM), histogram, and grey-level zone length matrix (GLZLM) related features presented the best performances to predict the diagnosis; for scanner 2, it was GLCM and histogram-related features and metabolic tumour volume (MTV); and for scanner 1 + 2, it was histogram features, standardized uptake value (SUV) metrics, and MTV. RFs had a significant role in predicting the diagnosis of SPN, but their accuracies were directly related to the scanner.

Current Advances and Challenges in Radiomics of Brain Tumors

Imaging diagnosis is crucial for early detection and monitoring of brain tumors. Radiomics enable the extraction of a large mass of quantitative features from complex clinical imaging arrays, and then transform them into high-dimensional data which can subsequently be mined to find their relevance with the tumor's histological features, which reflect underlying genetic mutations and malignancy, along with grade, progression, therapeutic effect, or even overall survival (OS). Compared to traditional brain imaging, radiomics provides quantitative information linked to meaningful biologic characteristics and application of deep learning which sheds light on the full automation of imaging diagnosis. Recent studies have shown that radiomics' application is broad in identifying primary tumor, differential diagnosis, grading, evaluation of mutation status and aggression, prediction of treatment response and recurrence in pituitary tumors, gliomas, and brain metastases. In this descriptive review, besides establishing a general understanding among protocols, results, and clinical significance of these studies, we further discuss the current limitations along with future development of radiomics.

On dose cube pixel spacing pre-processing for features extraction stability in dosiomic studies

PURPOSE

Dosiomics allows to parameterize regions of interest (ROIs) and to produce quantitative dose features encoding the spatial and statistical distribution of radiotherapy dose. The stability of dosiomics features extraction on dose cube pixel spacing variation has been investigated in this study.

MATERIAL AND METHODS

Based on 17 clinical delivered dose distributions (P_n), dataset has been generated considering all the possible combinations of four dose grid resolutions and two calculation algorithms. Each dose voxel cube has been post-processed considering 4 different dose cube pixel spacing values: 1x1x1, 2x2x2, 3x3x3 mm³ and the one equal to the planning CT. Dosiomics features extraction has been performed from four different ROIs. The stability of each extracted dosiomic feature has been analyzed in terms of coefficient of variation (CV) intraclass correlation coefficient (ICC).

RESULTS

The highest CV mean values were observed for PTV ROI and for the grey level size zone matrix features family. On the other hand, the lowest CV mean values have been found for RING ROI for the grey level co-occurrence matrix features family. P₃ showed the highest percentage of CV >1 (1.14%) followed by P₁₅ (0.41%), P₁ (0.29%) and P13 (0.19%). ICC analysis leads to identify features with an ICC >0.95 that could be considered stable to use in dosiomic studies when different dose cube pixel spacing are considered, especially the features in common among the seventeen plans.

CONCLUSION

Considering the observed variability, dosiomic studies should always provide a report not only on grid resolution and algorithm dose calculation, but also on dose cube pixel spacing.

Radiomics in PET Imaging:: A Practical Guide for Newcomers

Radiomics has undergone considerable development in recent years. In PET imaging, very promising results concerning the ability of handcrafted features to predict the biological characteristics of lesions and to assess patient prognosis or response to treatment have been reported in the literature. This article presents a checklist for designing a reliable radiomic study, gives an overview of the steps of the pipeline, and outlines approaches for data harmonization. Tips are provided for critical reading of the content of articles. The advantages and limitations of handcrafted radiomics compared with deep-learning approaches for the characterization of PET images are also discussed.

Repeatability of image features extracted from FET PET in application to post-surgical glioblastoma assessment

Positron emission tomography (PET) imaging using the amino acid tracer O-[2-(¹⁸F)fluoroethyl]-L-tyrosine (FET) has gained increased popularity within the past decade in the management of glioblastoma (GBM). Radiomics features extracted from FET PET images may be sensitive to variations when imaging at multiple time points. It is therefore necessary to assess feature robustness to test-retest imaging. Eight patients with histologically confirmed GBM that had undergone post-surgical test-retest FET PET imaging were recruited. In total, 1578 radiomic features were extracted from biological tumour volumes (BTVs) delineated using a semi-automatic contouring method. Feature repeatability was assessed using the intraclass correlation coefficient (ICC). The effect of both bin width and filter choice on feature repeatability was also investigated. 59/106 (55.7%) features from the original image and 843/1472 (57.3%) features from filtered images had an ICC ≥ 0.85. Shape and first order features were most stable. Choice of bin width showed minimal impact on features defined as stable. The Laplacian of Gaussian (LoG, σ = 5 mm) and Wavelet filters (HLL and LHL) significantly improved feature repeatability (p ≪ 0.0001, p = 0.003, p = 0.002, respectively). Correlation of textural features with tumour volume was reported for transparency. FET PET radiomic features extracted from post-surgical images of GBM patients that are robust to test-retest imaging were identified. An investigation with a larger dataset is warranted to validate the findings in this study.

Anatomy and Physiology of Artificial Intelligence in PET Imaging

Artificial intelligence (AI) has seen an explosion in interest within nuclear medicine. This interest is driven by the rapid progress and eye-catching achievements of machine learning algorithms. The growing foothold of AI in molecular imaging is exposing nuclear medicine personnel to new technology and terminology. Clinicians and researchers can be easily overwhelmed by numerous architectures and algorithms that have been published. This article dissects the backbone of most AI algorithms: the convolutional neural network. The algorithm training workflow and the key ingredients and operations of a convolutional neural network are described in detail. Finally, the ubiquitous U-Net is explained step-by-step.

A Multicentre Evaluation of Dosiomics Features Reproducibility, Stability and Sensitivity

Simple Summary

Dosiomics is born directly as an extension of radiomics: it entails extracting features from the patients’ three-dimensional (3D) radiotherapy dose distribution rather than from conventional medical images to obtain specific spatial and statistical information. Dosiomic studies, in a multicentre setting, require assessing the features’ stability to dose calculation settings and the features’ capability in distinguishing different dose distributions. This study provides the first multicentre evaluation of the dosiomic features in terms of reproducibility, stability and sensitivity across various dose distributions obtained from multiple technologies and techniques and considering different dose calculation algorithms of TPS and two different resolutions of the dose grid. Harmonisation strategies to account for a possible variation in the dose distribution due to these confounding factors should be adopted when investigating a correlation between dosiomic features and clinical outcomes in multicentre studies.

Abstract

Dosiomics is a texture analysis method to produce dose features that encode the spatial 3D distribution of radiotherapy dose. Dosiomic studies, in a multicentre setting, require assessing the features’ stability to dose calculation settings and the features’ capability in distinguishing different dose distributions. Dose distributions were generated by eight Italian centres on a shared image dataset acquired on a dedicated phantom. Treatment planning protocols, in terms of planning target volume coverage and dose–volume constraints to the organs at risk, were shared among the centres to produce comparable dose distributions for measuring reproducibility/stability and sensitivity of dosiomic features. In addition, coefficient of variation (CV) was employed to evaluate the dosiomic features’ variation. We extracted 38,160 features from 30 different dose distributions from six regions of interest, grouped by four features’ families. A selected group of features (CV < 3 for the reproducibility/stability studies, CV > 1 for the sensitivity studies) were identified to support future multicentre studies, assuring both stable features when dose distributions variation is minimal and sensitive features when dose distribution variations need to be clearly identified. Dosiomic is a promising tool that could support multicentre studies, especially for predictive models, and encode the spatial and statistical characteristics of the 3D dose distribution.

Quantitative radiomics features in diffuse large B-cell lymphoma: does segmentation method matter?

INTRODUCTION

Radiomics features may predict outcome in diffuse large B-cell lymphoma (DLBCL). Currently, multiple segmentation methods are used to calculate metabolic tumor volume (MTV). We assessed the influence of segmentation method on the discriminative power of radiomics features in DLBCL for patient level and for the largest lesion. Methods: 50 baseline ¹⁸F-fluorodeoxyglucose positron emission tomography computed tomography (PET/CT) scans of DLBCL patients who progressed or relapsed within 2 years after diagnosis were matched on uptake time and reconstruction method with 50 baseline PET/CT scans of DLBCL patients without progression. Scans were analysed using 6 semi-automatic segmentation methods (standardized uptake value (SUV)4.0, SUV2.5, 41% of the maximum SUV, 50% of the SUVpeak, majority vote (MV)2 and MV3, respectively). Based on these segmentations, 490 radiomics features were extracted at patient level and 486 features for the largest lesion. To quantify the agreement between features extracted from different segmentation methods, the intra-class correlation (ICC) agreement was calculated for each method compared to SUV4.0. The feature space was reduced by deleting features that had high Pearson correlations (≥0.7) with the previously established predictors MTV and/or SUVpeak. Model performance was assessed using stratified repeated cross-validation with 5 folds and 2000 repeats yielding the mean receiver-operating characteristics curve integral (CV-AUC) for all segmentation methods using logistic regression with backward feature selection. Results: The percentage of features yielding an ICC ≥0.75 compared to the SUV4.0 segmentation was lowest for A50P both at patient level and for the largest lesion, with 77.3% and 66.7% of the features yielding an ICC ≥0.75, respectively. Features were not highly correlated with MTV, with at least 435 features at patient level and 409 features for the largest lesion for all segmentation methods with a correlation coefficient <0.7. Features were highly correlated with SUVpeak (at least 190 and 134 were uncorrelated, respectively). CV-AUCs ranged between 0.69±0.11 and 0.84±0.09 for patient level, and between 0.69±0.11 and 0.73±0.10 for lesion level. Conclusion: Even though there are differences in the actual radiomics feature values derived and selected features between segmentation methods, there is no substantial difference in the discriminative power of radiomics features between segmentation methods.

Value of PET imaging for radiation therapy

This comprehensive review written by experts in their field gives an overview on the current status of incorporating positron emission tomography (PET) into radiation treatment planning. Moreover, it highlights ongoing studies for treatment individualisation and per-treatment tumour response monitoring for various primary tumours. Novel tracers and image analysis methods are discussed. The authors believe this contribution to be of crucial value for experts in the field as well as for policy makers deciding on the reimbursement of this powerful imaging modality.

Value of PET imaging for radiation therapy

This comprehensive review written by experts in their field gives an overview on the current status of incorporating positron emission tomography (PET) into radiation treatment planning. Moreover, it highlights ongoing studies for treatment individualisation and per-treatment tumour response monitoring for various primary tumours. Novel tracers and image analysis methods are discussed. The authors believe this contribution to be of crucial value for experts in the field as well as for policy makers deciding on the reimbursement of this powerful imaging modality.

18F-FMISO-PET Hypoxia Monitoring for Head-and-Neck Cancer Patients: Radiomics Analyses Predict the Outcome of Chemo-Radiotherapy

Tumor hypoxia is associated with radiation resistance and can be longitudinally monitored by 18F-fluoromisonidazole (18F-FMISO)-PET/CT. Our study aimed at evaluating radiomics dynamics of 18F-FMISO-hypoxia imaging during chemo-radiotherapy (CRT) as predictors for treatment outcome in head-and-neck squamous cell carcinoma (HNSCC) patients. We prospectively recruited 35 HNSCC patients undergoing definitive CRT and longitudinal 18F-FMISO-PET/CT scans at weeks 0, 2 and 5 (W0/W2/W5). Patients were classified based on peritherapeutic variations of the hypoxic sub-volume (HSV) size (increasing/stable/decreasing) and location (geographically-static/geographically-dynamic) by a new objective classification parameter (CP) accounting for spatial overlap. Additionally, 130 radiomic features (RF) were extracted from HSV at W0, and their variations during CRT were quantified by relative deviations (∆RF). Prediction of treatment outcome was considered statistically relevant after being corrected for multiple testing and confirmed for the two 18F-FMISO-PET/CT time-points and for a validation cohort. HSV decreased in 64% of patients at W2 and in 80% at W5. CP distinguished earlier disease progression (geographically-dynamic) from later disease progression (geographically-static) in both time-points and cohorts. The texture feature low grey-level zone emphasis predicted local recurrence with AUCW2 = 0.82 and AUCW5 = 0.81 in initial cohort (N = 25) and AUCW2 = 0.79 and AUCW5 = 0.80 in validation cohort. Radiomics analysis of 18F-FMISO-derived hypoxia dynamics was able to predict outcome of HNSCC patients after CRT.

Radiomics in prostate cancer imaging for a personalized treatment approach - current aspects of methodology and a systematic review on validated studies

Prostate cancer (PCa) is one of the most frequently diagnosed malignancies of men in the world. Due to a variety of treatment options in different risk groups, proper diagnostic and risk stratification is pivotal in treatment of PCa. The development of precise medical imaging procedures simultaneously to improvements in big data analysis has led to the establishment of radiomics - a computer-based method of extracting and analyzing image features quantitatively. This approach bears the potential to assess and improve PCa detection, tissue characterization and clinical outcome prediction. This article gives an overview on the current aspects of methodology and systematically reviews available literature on radiomics in PCa patients, showing its potential for personalized therapy approaches. The qualitative synthesis includes all imaging modalities and focuses on validated studies, putting forward future directions.

Molecular Imaging for Thyrotoxicosis and Thyroid Nodules

After exclusion of exogenous iodine overload, radioiodine uptake (RAIU) testing with 123I or 131I enables the accurate evaluation and quantification of iodine uptake and kinetics within thyroid cells. In addition, scintigraphic evaluation with 123I or 99mTc-pertechnetate (99mTc04-) provides the topographic distribution of thyroid cell activity and allows the detection and localization of ectopic thyroid tissue. Destructive thyrotoxicosis is characterized by abolished or reduced uptake whereas productive thyrotoxicosis (i.e., hyperthyroidism "sensu strictu") is characterized by high RAIU with scintigraphically diffuse (i.e., Graves disease and diffuse thyroid autonomy) or focal (i.e., autonomously functioning thyroid nodules [AFTN]) overactivity. Accordingly, RAIU or thyroid scintigraphy are widely used to differentiate different causes of thyrotoxicosis. In addition, several radiopharmaceuticals are also available to help in differentiating benign from malignant thyroid nodules and inform clinical decision making. In fact, AFTNs can be safely excluded from fine-needle aspiration biopsy while either 99mTc-methoxyisobutylisonitrile (MIBI) and 18F-FDG may complement the work-up of cytologically indeterminate cold nodules and contribute to reducing the need for diagnostic lobectomies/thyroidectomies. Finally, RAIU studies are also useful for calculating the administered therapeutic activity of 131I to treat hyperthyroidism and euthyroid multinodular goiter. All considered, thyroid molecular imaging allows us to characterize molecular/functional aspects of different thyroid diseases, even before clinical symptoms become manifest and remains integral to properly managing such conditions. Our present paper summarizes basic concepts, clinical applications, and potential developments of thyroid molecular imaging in patients affected by thyrotoxicosis and thyroid nodules.

Krizsan A, Forgács A, Berényi E, Balkay L. Effect of grey-level discretization on texture feature on different weighted MRI images of diverse disease groups

PURPOSE

Many studies of MRI radiomics do not include the discretization method used for the analyses, which might indicate that the discretization methods used are considered irrelevant. Our goals were to compare three frequently used discretization methods (lesion relative resampling (LRR), lesion absolute resampling (LAR) and absolute resampling (AR)) applied to the same data set, along with two different lesion segmentation approaches.

METHODS

We analyzed the effects of altering bin widths or bin numbers for the three different sampling methods using 40 texture indices (TIs). The impact was evaluated on brain MRI studies obtained for 71 patients divided into three different disease groups: multiple sclerosis (MS, N = 22), ischemic stroke (IS, N = 22), cancer patients (N = 27). Two different MRI acquisition protocols were considered for all patients, a T2- and a post-contrast 3D T1-weighted MRI sequence. Elliptical and manually drawn VOIs were employed for both imaging series. Three different types of gray-level discretization methods were used: LRR, LAR and AR. Hypothesis tests were done among all diseased and control areas to compare the TI values in these areas. We also did correlation analyses between TI values and lesion volumes.

RESULTS

In general, no significant differences were reported in the results when employing the AR and LAR discretization methods. It was found that employing 38 TIs introduced variation in the results when the number of bin parameters was altered, suggesting that both the degree and direction of monotonicity between each TI value and binning parameters were characteristic for each TI. Furthermore, while TIs were changing with altering binning values, no changes correlated to neither disease nor the MRI sequence. We found that most indices correlated weakly with the volume, while the correlation coefficients were independent of both diseases analyzed and MR contrast. Several cooccurrence-matrix based texture parameters show a definite higher correlation when employing the LRR discretization method However, with the best correlations obtained for the manually drawn VOI. Hypothesis tests among all disease and control areas (co-lateral hemisphere) revealed that the AR or LAR discretization techniques provide more suitable texture features than LRR. In addition, the manually drawn segmentation gave fewer significantly different TIs than the ellipsoid segmentations. In addition, the amount of TIs with significant differences was increasing with increasing the number of bins, or decreasing bin widths.

CONCLUSION

Our findings indicate that the AR discretization method may offer the best texture analysis in MR image assessments. Employing too many bins or too large bin widths might reduce the selection of TIs that can be used for differential diagnosis. In general, more statistically different TIs were observed for elliptical segmentations when compared to the manually drawn VOIs. In the texture analysis of MR studies, studies and publications should report on all important parameters and methods related to data collection, corrections, normalization, discretization, and segmentation.

Experimental phantom evaluation to identify robust positron emission tomography (PET) radiomic features

Background

Radiomics analysis usually involves, especially in multicenter and large hospital studies, different imaging protocols for acquisition, reconstruction, and processing of data. Differences in protocols can lead to differences in the quantification of the biomarker distribution, leading to radiomic feature variability. The aim of our study was to identify those radiomic features robust to the different degrading factors in positron emission tomography (PET) studies. We proposed the use of the standardized measurements of the European Association Research Ltd. (EARL) accreditation to retrospectively identify the radiomic features having low variability to the different systems and reconstruction protocols. In addition, we presented a reproducible procedure to identify PET radiomic features robust to PET/CT imaging metal artifacts. In 27 heterogeneous homemade phantoms for which ground truth was accurately defined by CT segmentation, we evaluated the segmentation accuracy and radiomic feature reliability given by the contrast-oriented algorithm (COA) and the 40% threshold PET segmentation. In the comparison of two data sets, robustness was defined by Wilcoxon rank tests, bias was quantified by Bland–Altman (BA) plot analysis, and strong correlations were identified by Spearman correlation test (r > 0.8 and p satisfied multiple test Bonferroni correction).

Results

Forty-eight radiomic features were robust to system, 22 to resolution, 102 to metal artifacts, and 42 to different PET segmentation tools. Overall, only 4 radiomic features were simultaneously robust to all degrading factors. Although both segmentation approaches significantly underestimated the volume with respect to the ground truth, with relative deviations of −62 ± 36% for COA and −50 ± 44% for 40%, radiomic features derived from the ground truth were strongly correlated and/or robust to 98 radiomic features derived from COA and to 102 from 40%.

Conclusion

In multicenter studies, we recommend the analysis of EARL accreditation measurements in order to retrospectively identify the robust PET radiomic features. Furthermore, 4 radiomic features (area under the curve of the cumulative SUV volume histogram, skewness, kurtosis, and gray-level variance derived from GLRLM after application of an equal probability quantization algorithm on the voxels within lesion) were robust to all degrading factors. In addition, the feasibility of 40% and COA segmentations for their use in radiomics analysis has been demonstrated.

Supplementary Information

The online version contains supplementary material available at 10.1186/s40658-021-00390-7.

Radiomics Analysis of 3D Dose Distributions to Predict Toxicity of Radiotherapy for Cervical Cancer

Standard treatment for locally advanced cervical cancer (LACC) is chemoradiotherapy followed by brachytherapy. Despite radiation therapy advances, the toxicity rate remains significant. In this study, we compared the prediction of toxicity events after radiotherapy for locally advanced cervical cancer (LACC), based on either dose-volume histogram (DVH) parameters or the use of a radiomics approach applied to dose maps at the voxel level. Toxicity scores using the Common Terminology Criteria for Adverse Events (CTCAE v4), spatial dose distributions, and usual clinical predictors for the toxicity of 102 patients treated with chemoradiotherapy followed by brachytherapy for LACC were used in this study. In addition to usual DVH parameters, 91 radiomic features were extracted from rectum, bladder and vaginal 3D dose distributions, after discretization into a fixed bin width of 1 Gy. They were evaluated for predictive modelling of rectal, genitourinary (GU) and vaginal toxicities (grade ≥ 2). Logistic Normal Tissue Complication Probability (NTCP) models were derived using clinical parameters only or combinations of clinical, DVH and radiomics. For rectal acute/late toxicities, the area under the curve (AUC) using clinical parameters was 0.53/0.65, which increased to 0.66/0.63, and 0.76/0.87, with the addition of DVH or radiomics parameters, respectively. For GU acute/late toxicities, the AUC increased from 0.55/0.56 (clinical only) to 0.84/0.90 (+DVH) and 0.83/0.96 (clinical + DVH + radiomics). For vaginal acute/late toxicities, the AUC increased from 0.51/0.57 (clinical only) to 0.58/0.72 (+DVH) and 0.82/0.89 (clinical + DVH + radiomics). The predictive performance of NTCP models based on radiomics features was higher than the commonly used clinical and DVH parameters. Dosimetric radiomics analysis is a promising tool for NTCP modelling in radiotherapy.

Repeatability of 18F-FDG PET Radiomic Features in Cervical Cancer

Knowledge of the intrinsic variability of radiomic features is essential to the proper interpretation of changes in these features over time. The primary aim of this study was to assess the test-retest repeatability of radiomic features extracted from ¹⁸F-FDG PET images of cervical tumors. The impact of different image preprocessing methods was also explored. Methods: Patients with cervical cancer underwent baseline and repeat ¹⁸F-FDG PET/CT imaging within 7 d. PET images were reconstructed using 2 methods: ordered-subset expectation maximization (PET_OSEM) or ordered-subset expectation maximization with point-spread function (PET_PSF). Tumors were segmented to produce whole-tumor volumes of interest (VOI_WT) and 40% isocontours (VOI₄₀). Voxels were either left at the default size or resampled to 3-mm isotropic voxels. SUV was discretized to a fixed number of bins (32, 64, or 128). Radiomic features were extracted from both VOIs, and repeatability was then assessed using the Lin concordance correlation coefficient (CCC). Results: Eleven patients were enrolled and completed the test-retest PET/CT imaging protocol. Shape, neighborhood gray-level difference matrix, and gray-level cooccurrence matrix features were repeatable, with a mean CCC value of 0.81. Radiomic features extracted from PET_OSEM images showed significantly better repeatability than features extracted from PET_PSF images (P < 0.001). Radiomic features extracted from VOI₄₀ were more repeatable than features extracted from VOI_WT (P < 0.001). For most features (78.4%), a change in bin number or voxel size resulted in less than a 10% change in feature value. All gray-level emphasis and gray-level run emphasis features showed poor repeatability (CCC values < 0.52) when extracted from VOI_WT but were highly repeatable (mean CCC values > 0.96) when extracted from VOI₄₀ Conclusion: Shape, gray-level cooccurrence matrix, and neighborhood gray-level difference matrix radiomic features were consistently repeatable, whereas gray-level run length matrix and gray-level zone length matrix features were highly variable. Radiomic features extracted from VOI₄₀ were more repeatable than features extracted from VOI_WT Changes in voxel size or SUV discretization parameters typically resulted in relatively small differences in feature value, though several features were highly sensitive to these changes.

Machine learning based on clinico-biological features integrated 18F-FDG PET/CT radiomics for distinguishing squamous cell carcinoma from adenocarcinoma of lung

PURPOSE

To develop and validate a clinico-biological features and 18F-fluorodeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) radiomic-based nomogram via machine learning for the pretherapy prediction of discriminating between adenocarcinoma (ADC) and squamous cell carcinoma (SCC) in non-small cell lung cancer (NSCLC).

METHODS

A total of 315 NSCLC patients confirmed by postoperative pathology between January 2017 and June 2019 were retrospectively analyzed and randomly divided into the training (n = 220) and validation (n = 95) sets. Preoperative clinical factors, serum tumor markers, and PET, and CT radiomic features were analyzed. Prediction models were developed using the least absolute shrinkage and selection operator (LASSO) regression analysis. The performance of the models was evaluated and compared by the area under receiver-operator characteristic (ROC) curve (AUC) and DeLong test. The clinical utility of the models was determined via decision curve analysis (DCA). Then, a nomogram was developed based on the model with the best predictive efficiency and clinical utility and was validated using the calibration plots.

RESULTS

In total, 122 SCC and 193 ADC patients were enrolled in this study. Four independent prediction models were separately developed to differentiate SCC from ADC using clinical factors-tumor markers, PET radiomics, CT radiomics, and their combination. The DeLong test and DCA showed that the Combined Model, consisting of 2 clinical factors, 2 tumor markers, 7 PET radiomics, and 3 CT radiomic parameters, held the highest predictive efficiency and clinical utility in predicting the NSCLC subtypes compared with the use of these parameters alone in both the training and validation sets (AUCs (95% CIs) = 0.932 (0.900-0.964), 0.901 (0.840-0.957), respectively) (p < 0.05). A quantitative nomogram was subsequently constructed using the independently risk factors from the Combined Model. The calibration curves indicated a good consistency between the actual observations and nomogram predictions.

CONCLUSION

This study presents an integrated clinico-biologico-radiological nomogram that can be accurately and noninvasively used for the individualized differentiation SCC from ADC in NSCLC, thereby assisting in clinical decision making for precision treatment.

Combination of Quantitative MRI Fat Fraction and Texture Analysis to Evaluate Spastic Muscles of Children With Cerebral Palsy

Background: Cerebral palsy (CP) is the most common cause of physical disability in childhood. Muscle pathologies occur due to spasticity and contractures; therefore, diagnostic imaging to detect pathologies is often required. Imaging has been used to assess torsion or estimate muscle volume, but additional methods for characterizing muscle composition have not thoroughly been investigated. MRI fat fraction (FF) measurement can quantify muscle fat and is often a part of standard imaging in neuromuscular dystrophies. To date, FF has been used to quantify muscle fat and assess function in CP. In this study, we aimed to utilize a radiomics and FF analysis along with the combination of both methods to differentiate affected muscles from healthy ones. Materials and Methods: A total of 9 patients (age range 8-15 years) with CP and 12 healthy controls (age range 9-16 years) were prospectively enrolled (2018-2020) after ethics committee approval. Multi-echo Dixon acquisition of the calf muscles was used for FF calculation. The images of the second echo (TE = 2.87 ms) were used for feature extraction from the soleus, gastrocnemius medialis, and gastrocnemius lateralis muscles. The least absolute shrinkage and selection operator (LASSO) regression was employed for feature selection. RM, FF model (FFM), and combined model (CM) were built for each calf muscle. The receiver operating characteristic (ROC) curve and their respective area under the curve (AUC) values were used to evaluate model performance. Results: In total, the affected legs of 9 CP patients and the dominant legs of 12 healthy controls were analyzed. The performance of RM for soleus, gastrocnemius medialis, and gastrocnemius lateralis (AUC 0.92, 0.92, 0.82, respectively) was better than the FFM (AUC 0.88, 0.85, 0.69, respectively). The combination of both models always had a better performance than RM or FFM (AUC 0.95, 0.93, 0.83). FF was higher in the patient group (FFS 9.1%, FFGM 8.5%, and FFGL 10.2%) than control group (FFS 3.3%, FFGM 4.1%, FFGL 6.6%). Conclusion: The combination of MRI quantitative fat fraction analysis and texture analysis of muscles is a promising tool to evaluate muscle pathologies due to CP in a non-invasive manner.

FET-PET radiomics in recurrent glioblastoma: prognostic value for outcome after re-irradiation?

Purpose

The value of O-(2-[18F]fluoroethyl)-L-tyrosine (FET)-positron emission tomography (PET)-radiomics in the outcome assessment of patients with recurrent glioblastoma (rGBM) has not been evaluated until now. The aim of this study was to evaluate whether a prognostic model based on FET-PET radiomics features (RF) is feasible and can identify rGBM patients that would most benefit from re-irradiation.

Methods

We prospectively recruited rGBM patients who underwent FET-PET before re-irradiation (GLIAA-Pilot trial, DRKS00000633). Tumor volume was delineated using a semi-automatic method with a threshold of 1.8 times the standardized-uptake-value of the background. 135 FET-RF (histogram parameters, shape and texture features) were extracted. The analysis involved the characterization of tumor and non-tumor tissue with FET-RF and the evaluation of the prognostic value of FET-RF for time-to-progression (TTP), overall survival (OS) and recurrence location (RL).

Results

Thirty-two rGBM patients constituted our cohort. FET-RF discriminated significantly between tumor and non-tumor. The texture feature Small-Zone-Low-Gray-Level-Emphasis (SZLGE) showed the best performance for the prediction of TTP (p = 0.001, satisfying Bonferroni-multiple-test significance level). Additionally, two radiomics signatures could predict TTP (TTP-radiomics-signature, p = 0.001) and OS (OS-radiomics-signature, p = 0.038). SZLGE and the TTP-radiomics-signature additionally predicted RL. Specifically, high values for TTP-radiomics-signature and for SZLGE indicated not only earlier progression, but also a RL within the initial FET-PET active volume.

Conclusion

Our findings suggest that FET-PET radiomics could contribute to the prognostic assessment and selection of rGBM-patients benefiting from re-irradiation.

Trial registration DRKS00000633. Registered on 8th of December in 2010.

https://www.drks.de/drks_web/navigate.do?navigationId=trial.HTML&TRIAL_ID=DRKS00000633.

Artificial Intelligence for Response Evaluation With PET/CT

Positron emission tomography (PET)/computed tomography (CT) are nuclear diagnostic imaging modalities that are routinely deployed for cancer staging and monitoring. They hold the advantage of detecting disease related biochemical and physiologic abnormalities in advance of anatomical changes, thus widely used for staging of disease progression, identification of the treatment gross tumor volume, monitoring of disease, as well as prediction of outcomes and personalization of treatment regimens. Among the arsenal of different functional imaging modalities, nuclear imaging has benefited from early adoption of quantitative image analysis starting from simple standard uptake value normalization to more advanced extraction of complex imaging uptake patterns; thanks to application of sophisticated image processing and machine learning algorithms. In this review, we discuss the application of image processing and machine/deep learning techniques to PET/CT imaging with special focus on the oncological radiotherapy domain as a case study and draw examples from our work and others to highlight current status and future potentials.

Radiomics in immuno-oncology

With the ongoing advances in imaging techniques, increasing volumes of anatomical and functional data are being generated as part of the routine clinical workflow. This surge of available imaging data coincides with increasing research in quantitative imaging, particularly in the domain of imaging features. An important and novel approach is radiomics, where high-dimensional image properties are extracted from routine medical images. The fundamental principle of radiomics is the hypothesis that biomedical images contain predictive information, not discernible to the human eye, that can be mined through quantitative image analysis. In this review, a general outline of radiomics and artificial intelligence (AI) will be provided, along with prominent use cases in immunotherapy (e.g. response and adverse event prediction) and targeted therapy (i.e. radiogenomics). While the increased use and development of radiomics and AI in immuno-oncology is highly promising, the technology is still in its early stages, and different challenges still need to be overcome. Nevertheless, novel AI algorithms are being constructed with an ever-increasing scope of applications.

Impact of segmentation and discretization on radiomic features in 68Ga-DOTA-TOC PET/CT images of neuroendocrine tumor

OBJECTIVE

To identify the impact of segmentation methods and intensity discretization on radiomic features (RFs) extraction from 68Ga-DOTA-TOC PET images in patients with neuroendocrine tumors.

METHODS

Forty-nine patients were retrospectively analyzed. Tumor contouring was performed manually by four different operators and with a semi-automatic edge-based segmentation (SAEB) algorithm. Three SUVmax fixed thresholds (20, 30, 40%) were applied. Fifty-one RFs were extracted applying two different intensity rescale factors for gray-level discretization: one absolute (AR60 = SUV from 0 to 60) and one relative (RR = min-max of the VOI SUV). Dice similarity coefficient (DSC) was calculated to quantify segmentation agreement between different segmentation methods. The impact of segmentation and discretization on RFs was assessed by intra-class correlation coefficients (ICC) and the coefficient of variance (COVL). The RFs' correlation with volume and SUVmax was analyzed by calculating Pearson's correlation coefficients.

RESULTS

DSC mean value was 0.75 ± 0.11 (0.45-0.92) between SAEB and operators and 0.78 ± 0.09 (0.36-0.97), among the four manual segmentations. The study showed high robustness (ICC > 0.9): (a) in 64.7% of RFs for segmentation methods using AR60, improved by applying SUVmax threshold of 40% (86.5%); (b) in 50.9% of RFs for different SUVmax thresholds using AR60; and (c) in 37% of RFs for discretization settings using different segmentation methods. Several RFs were not correlated with volume and SUVmax.

CONCLUSIONS

RFs robustness to manual segmentation resulted higher in NET 68Ga-DOTA-TOC images compared to 18F-FDG PET/CT images. Forty percent SUVmax thresholds yield superior RFs stability among operators, however leading to a possible loss of biological information. SAEB segmentation appears to be an optimal alternative to manual segmentation, but further validations are needed. Finally, discretization settings highly impacted on RFs robustness and should always be stated.

Comparison of three freeware software packages for 18F-FDG PET texture feature calculation

PURPOSE

To compare texture feature estimates obtained from ¹⁸F-FDG-PET images using three different software packages.

METHODS

PET images from 15 patients with head and neck cancer were processed with three different freeware software: CGITA, LIFEx, and Metavol. For each lesion, 38 texture features were extracted from each software package. To evaluate the statistical agreement among the features across packages a non-parametric Kruskal-Wallis test was used. Differences in the features between each couple of software were assessed using a subsequent Dunn test. Correlation between texture features was evaluated via the Spearman coefficient.

RESULTS

Twenty-three of 38 features showed a significant agreement across the three software (P < 0.05). The agreement was better between LIFEx vs. Metavol (36 of 38) and worse between CGITA and Metavol (24 of 38), and CGITA vs. LIFEx (23 of 38). All features resulted correlated (ρ >  = 0.70, P < 0.001) in comparing LIFEx vs. Metavol. Seven of 38 features were found not in agreement and slightly or not correlated (ρ < 0.70, P < 0.001) in comparing CGITA vs. LIFEx, and CGITA vs. Metavol.

CONCLUSION

Some texture discrepancies across software packages exist. Our findings reinforce the need to continue the standardization process, and to succeed in building a reference dataset to be used for comparisons.

FDG-PET Radiomics for Response Monitoring in Non-Small-Cell Lung Cancer Treated with Radiation Therapy

Simple Summary

In this study, we strive to identify clinically relevant image feature (IF) changes during chemoradiation in patients with non-small-cell lung cancer (NSCLC) to be able to predict tumor responses in an early stage of treatment. All patients underwent static (3D) and respiratory-gated 4D PET/CT scans before treatment and a 3D scan during or after treatment. Our proposed method rejects IF changes due to intrinsic variability such as noise, resolution and movement through breathing. The IF variability observed across 4D PET is employed as a patient individualized normalization factor to emphasize statistically relevant IF changes during treatment.

Abstract

The aim of this study is to identify clinically relevant image feature (IF) changes during chemoradiation and evaluate their efficacy in predicting treatment response. Patients with non-small-cell lung cancer (NSCLC) were enrolled in two prospective trials (STRIPE, PET-Plan). We evaluated 48 patients who underwent static (3D) and retrospectively-respiratory-gated 4D PET/CT scans before treatment and a 3D scan during or after treatment. Our proposed method rejects IF changes due to intrinsic variability. The IF variability observed across 4D PET is employed as a patient individualized normalization factor to emphasize statistically relevant IF changes during treatment. Predictions of overall survival (OS), local recurrence (LR) and distant metastasis (DM) were evaluated. From 135 IFs, only 17 satisfied the required criteria of being normally distributed across 4D PET and robust between 3D and 4D images. Changes during treatment in the area-under-the-curve of the cumulative standard-uptake-value histogram (δ_AUCCSH) within primary tumor discriminated (AUC = 0.87, Specificity = 0.78) patients with and without LR. The resulted prognostic model was validated with a different segmentation method (AUC = 0.83) and in a different patient cohort (AUC = 0.63). The quantification of tumor FDG heterogeneity by δ_AUCCSH during chemoradiation correlated with the incidence of local recurrence and might be recommended for monitoring treatment response in patients with NSCLC.

Evaluation of FET PET Radiomics Feature Repeatability in Glioma Patients

Simple Summary

Initial studies suggested the additional diagnostic value of amino acid positron emission tomography (PET) radiomics using the tracer O-(2-[¹⁸F]fluoroethyl)-L-tyrosine (FET) in brain tumor patient management. However, to ensure the reliable performance of the generated FET PET radiomics models for clinical diagnostics, repeatability of radiomics features is essential. Hence, we assessed the impact of brain tumor volumes and key molecular alterations such as an isocitrate dehydrogenase (IDH) mutation on the repeatability of FET PET radiomics features in 50 newly diagnosed glioma patients. In a test–retest approach based on routinely acquired FET PET scans, we identified 297 repeatable features. The IDH genotype did not affect feature repeatability. Moreover, these robust features were able to differentiate patients with IDH-wildtype glioma from those with an IDH mutation. Our results suggest that robust radiomics features can be obtained from routinely acquired FET PET scans, which are valuable for further standardization of radiomics analyses in neurooncology.

Abstract

Amino acid PET using the tracer O-(2-[¹⁸F]fluoroethyl)-L-tyrosine (FET) has attracted considerable interest in neurooncology. Furthermore, initial studies suggested the additional diagnostic value of FET PET radiomics in brain tumor patient management. However, the conclusiveness of radiomics models strongly depends on feature generalizability. We here evaluated the repeatability of feature-based FET PET radiomics. A test–retest analysis based on equivalent but statistically independent subsamples of FET PET images was performed in 50 newly diagnosed and histomolecularly characterized glioma patients. A total of 1,302 radiomics features were calculated from semi-automatically segmented tumor volumes-of-interest (VOIs). Furthermore, to investigate the influence of the spatial resolution of PET on repeatability, spherical VOIs of different sizes were positioned in the tumor and healthy brain tissue. Feature repeatability was assessed by calculating the intraclass correlation coefficient (ICC). To further investigate the influence of the isocitrate dehydrogenase (IDH) genotype on feature repeatability, a hierarchical cluster analysis was performed. For tumor VOIs, 73% of first-order features and 71% of features extracted from the gray level co-occurrence matrix showed high repeatability (ICC 95% confidence interval, 0.91–1.00). In the largest spherical tumor VOIs, 67% of features showed high repeatability, significantly decreasing towards smaller VOIs. The IDH genotype did not affect feature repeatability. Based on 297 repeatable features, two clusters were identified separating patients with IDH-wildtype glioma from those with an IDH mutation. Our results suggest that robust features can be obtained from routinely acquired FET PET scans, which are valuable for further standardization of radiomics analyses in neurooncology.

Influence of Magnetic Field Strength on Magnetic Resonance Imaging Radiomics Features in Brain Imaging, an In Vitro and In Vivo Study

BACKGROUND

The development and clinical adoption of quantitative imaging biomarkers (radiomics) has established the need for the identification of parameters altering radiomics reproducibility. The aim of this study was to assess the impact of magnetic field strength on magnetic resonance imaging (MRI) radiomics features in neuroradiology clinical practice.

METHODS

T1 3D SPGR sequence was acquired on two phantoms and 10 healthy volunteers with two clinical MR devices from the same manufacturer using two different magnetic fields (1.5 and 3T). Phantoms varied in terms of gadolinium concentrations and textural heterogeneity. 27 regions of interest were segmented (phantom: 21, volunteers: 6) using the LIFEX software. 34 features were analyzed.

RESULTS

In the phantom dataset, 10 (67%) out of 15 radiomics features were significantly different when measured at 1.5T or 3T (student's t-test, p < 0.05). Gray levels resampling, and pixel size also influence part of texture features. These findings were validated in healthy volunteers.

CONCLUSIONS

According to daily used protocols for clinical examinations, radiomic features extracted on 1.5T should not be used interchangeably with 3T when evaluating texture features. Such confounding factor should be adjusted when adapting the results of a study to a different platform, or when designing a multicentric trial.

A comparative study to evaluate CT-based semantic and radiomic features in preoperative diagnosis of invasive pulmonary adenocarcinomas manifesting as subsolid nodules

This study aims to predict the histological invasiveness of pulmonary adenocarcinoma spectrum manifesting with subsolid nodules ≦ 3 cm using the preoperative CT-based radiomic approach. A total of 186 patients with 203 SSNs confirmed with surgically pathologic proof were retrospectively reviewed from February 2016 to March 2020 for training cohort modeling. The validation cohort included 50 subjects with 57 SSNs confirmed with surgically pathologic proof from April 2020 to August 2020. CT-based radiomic features were extracted using an open-source software with 3D nodular volume segmentation manually. The association between CT-based conventional features/selected radiomic features and histological invasiveness of pulmonary adenocarcinoma status were analyzed. Diagnostic models were built using conventional CT features, selected radiomic CT features and experienced radiologists. In addition, we compared diagnostic performance between radiomic CT feature, conventional CT features and experienced radiologists. In the training cohort of 203 SSNs, there were 106 invasive lesions and 97 pre-invasive lesions. Logistic analysis identified that a selected radiomic feature named GLCM_Entropy_log10 was the predictor for histological invasiveness of pulmonary adenocarcinoma spectrum (OR: 38.081, 95% CI 2.735–530.309, p = 0.007). The sensitivity and specificity for predicting histological invasiveness of pulmonary adenocarcinoma spectrum using the cutoff value of CT-based radiomic parameter (GLCM_Entropy_log10) were 84.8% and 79.2% respectively (area under curve, 0.878). The diagnostic model of CT-based radiomic feature was compared to those of conventional CT feature (morphologic and quantitative) and three experienced radiologists. The diagnostic performance of radiomic feature was similar to those of the quantitative CT feature (nodular size and solid component, both lung and mediastinal window) in prediction invasive pulmonary adenocarcinoma (IPA). The AUC value of CT radiomic feature was higher than those of conventional CT morphologic feature and three experienced radiologists. The c-statistic of the training cohort model was 0.878 (95% CI 0.831–0.925) and 0.923 (0.854–0.991) in the validation cohort. Calibration was good in both cohorts. The diagnostic performance of CT-based radiomic feature is not inferior to solid component (lung and mediastinal window) and nodular size for predicting invasiveness. CT-based radiomic feature and nomogram could help to differentiate IPA lesions from preinvasive lesions in the both independent training and validation cohorts. The nomogram may help clinicians with decision making in the management of subsolid nodules.

Hormonal Receptor Immunochemistry Heterogeneity and 18F-FDG Metabolic Heterogeneity: Preliminary Results of Their Relationship and Prognostic Value in Luminal Non-Metastatic Breast Cancers

Introduction

We aimed to investigate whether ¹⁸F-FDG PET metabolic heterogeneity reflects the heterogeneity of estrogen receptor (ER) and progesterone receptor (PR) expressions within luminal non-metastatic breast tumors and if it could help in identifying patients with worst event-free survival (EFS).

Materials and methods

On 38 PET high-resolution breast bed positions, a single physician drew volumes of interest encompassing the breast tumors to extract SUV_max, histogram parameters and textural features. High-resolution immunochemistry (IHC) scans were analyzed to extract Haralick parameters and descriptors of the distribution shape. Correlation between IHC and PET parameters were explored using Spearman tests. Variables of interest to predict the EFS status at 8 years (EFS-8y) were sought by means of a random forest classification. EFS-8y analyses were then performed using univariable Kaplan-Meier analyses and Cox regression analysis. When appropriate, Mann-Whitney tests and Spearman correlations were used to explore the relationship between clinical data and tumoral PET heterogeneity variables.

Results

For ER expression, correlations were mainly observed with ¹⁸F-FDG histogram parameters, whereas for PR expression correlations were mainly observed with gray-level co-occurrence matrix (GLCM) parameters. The strongest correlations were observed between skewness__ER and uniformity__HISTO (ρ = −0.386, p = 0.017) and correlation__PR and entropy__GLCM (ρ = 0.540, p = 0.001), respectively. The median follow-up was 6.5 years and the 8y-EFS was 71.0%. Random forest classification found age, clinical stage, SUV_max, skewness__ER, kurtosis__ER, entropy__HISTO, and uniformity__HISTO to be variables of importance to predict the 8y-EFS. Univariable Kaplan-Meier survival analyses showed that skewness__ER was a predictor of 8y-EFS (66.7 ± 27.2 versus 19.1 ± 15.2, p = 0.018 with a cut-off value set to 0.163) whereas other IHC and PET parameters were not. On multivariable analysis including age, clinical stage and skewness__ER, none of the parameters were independent predictors. Indeed, skewness__ER was significantly higher in youngest patients (ρ = −0.351, p = 0.031) and in clinical stage III tumors (p = 0.023).

Conclusion

A heterogeneous distribution of ER within the tumor in IHC appeared as an EFS-8y prognosticator in luminal non-metastatic breast cancers. Interestingly, it appeared to be correlated with PET histogram parameters which could therefore become potential non-invasive prognosticator tools, provided these results are confirmed by further larger and prospective studies.

A decade of radiomics research: are images really data or just patterns in the noise?

KEY POINTS

• Although radiomics is potentially a promising approach to analyze medical image data, many pitfalls need to be considered to avoid a reproducibility crisis.• There is a translation gap in radiomics research, with many studies being published but so far little to no translation into clinical practice.• Going forward, more studies with higher levels of evidence are needed, ideally also focusing on prospective studies with relevant clinical impact.

Prognostic Value of Baseline Radiomic Features of 18F-FDG PET in Patients with Diffuse Large B-Cell Lymphoma

This study investigates whether baseline ¹⁸F-FDG PET radiomic features can predict survival outcomes in patients with diffuse large B-cell lymphoma (DLBCL). We retrospectively enrolled 83 patients diagnosed with DLBCL who underwent ¹⁸F-FDG PET scans before treatment. The patients were divided into the training cohort (n = 58) and the validation cohort (n = 25). Eighty radiomic features were extracted from the PET images for each patient. Least absolute shrinkage and selection operator regression were used to reduce the dimensionality within radiomic features. Cox proportional hazards model was used to determine the prognostic factors for progression-free survival (PFS) and overall survival (OS). A prognostic stratification model was built in the training cohort and validated in the validation cohort using Kaplan-Meier survival analysis. In the training cohort, run length non-uniformity (RLN), extracted from a gray level run length matrix (GLRLM), was independently associated with PFS (hazard ratio (HR) = 15.7, p = 0.007) and OS (HR = 8.64, p = 0.040). The International Prognostic Index was an independent prognostic factor for OS (HR = 2.63, p = 0.049). A prognostic stratification model was devised based on both risk factors, which allowed identification of three risk groups for PFS and OS in the training (p < 0.001 and p < 0.001) and validation (p < 0.001 and p = 0.020) cohorts. Our results indicate that the baseline ¹⁸F-FDG PET radiomic feature, RLN_GLRLM, is an independent prognostic factor for survival outcomes. Furthermore, we propose a prognostic stratification model that may enable tailored therapeutic strategies for patients with DLBCL.

Incorporating radiomic feature of pretreatment 18F-FDG PET improves survival stratification in patients with EGFR-mutated lung adenocarcinoma

BACKGROUND

To investigate the survival prognostic value of the radiomic features of 18F-FDG PET in patients who had EGFR (epidermal growth factor receptor) mutated lung adenocarcinoma and received targeted TKI (tyrosine kinase inhibitor) treatment.

METHODS

Fifty-one patients with stage III-IV lung adenocarcinoma and actionable EGFR mutation who received first-line TKI were retrospectively analyzed. All patients underwent pretreatment 18F-FDG PET/CT, and we calculated the PET-derived radiomic features. Cox proportional hazard model was used to examine the association between the radiomic features and the survival outcomes, including progression-free survival (PFS) and overall survival (OS). A score model was established according to the independent prognostic predictors and we compared this model to the TNM staging system using Harrell's concordance index (c-index).

RESULTS

Forty-eight patients (94.1%) experienced disease progression and 41 patients (80.4%) died. Primary tumor SUV entropy > 5.36, and presence of pleural effusion were independently associated with worse OS (both p < 0.001) and PFS (p = 0.001, and 0.003, respectively). We used these two survival predictors to devise a scoring system (score 0-2). Patients with a score of 1 or 2 had a worse survival than those with a score of 0 (HR for OS: 3.6, p = 0.006 for score 1, and HR: 21.8, p < 0.001 for score 2; HR for PFS: 2.2, p = 0.027 for score 1 and HR: 8.8, p < 0.001 for score 2). Our scoring system surpassed the TNM staging system (c-index = 0.691 versus 0.574, p = 0.013 for OS, and c-index = 0.649 versus 0.517, p = 0.004 for PFS).

CONCLUSIONS

In this preliminary study, combining PET radiomics with clinical risk factors may improve survival stratification in stage III-IV lung adenocarcinoma with actionable EFGR mutation. Our proposed scoring system may assist with optimization of individualized treatment strategies in these patients.

Harmonization strategies for multicenter radiomics investigations

Carrying out large multicenter studies is one of the key goals to be achieved towards a faster transfer of the radiomics approach in the clinical setting. This requires large-scale radiomics data analysis, hence the need for integrating radiomic features extracted from images acquired in different centers. This is challenging as radiomic features exhibit variable sensitivity to differences in scanner model, acquisition protocols and reconstruction settings, which is similar to the so-called 'batch-effects' in genomics studies. In this review we discuss existing methods to perform data integration with the aid of reducing the unwanted variation associated with batch effects. We also discuss the future potential role of deep learning methods in providing solutions for addressing radiomic multicentre studies.

The Challenge of Evaluating Response to Peptide Receptor Radionuclide Therapy in Gastroenteropancreatic Neuroendocrine Tumors: The Present and the Future

The NETTER-1 study has proven peptide receptor radionuclide therapy (PRRT) to be one of the most effective therapeutic options for metastatic neuroendocrine tumors (NETs), improving progression-free survival and overall survival. However, PRRT response assessment is challenging and no consensus on methods and timing has yet been reached among experts in the field. This issue is owed to the suboptimal sensitivity and specificity of clinical biomarkers, limitations of morphological response criteria in slowly growing tumors and necrotic changes after therapy, a lack of standardized parameters and timing of functional imaging and the heterogeneity of PRRT protocols in the literature. The aim of this article is to review the most relevant current approaches for PRRT efficacy prediction and response assessment criteria in order to provide an overview of suitable tools for safe and efficacious PRRT.

Tumor response prediction in 90Y radioembolization with PET-based radiomics features and absorbed dose metrics

Purpose

To evaluate whether lesion radiomics features and absorbed dose metrics extracted from post-therapy ⁹⁰Y PET can be integrated to better predict outcomes in microsphere radioembolization of liver malignancies

Methods

Given the noisy nature of ⁹⁰Y PET, first, a liver phantom study with repeated acquisitions and varying reconstruction parameters was used to identify a subset of robust radiomics features for the patient analysis. In 36 radioembolization procedures, ⁹⁰Y PET/CT was performed within a couple of hours to extract 46 radiomics features and estimate absorbed dose in 105 primary and metastatic liver lesions. Robust radiomics modeling was based on bootstrapped multivariate logistic regression with shrinkage regularization (LASSO) and Cox regression with LASSO. Nested cross-validation and bootstrap resampling were used for optimal parameter/feature selection and for guarding against overfitting risks. Spearman rank correlation was used to analyze feature associations. Area under the receiver-operating characteristics curve (AUC) was used for lesion response (at first follow-up) analysis while Kaplan-Meier plots and c-index were used to assess progression model performance. Models with absorbed dose only, radiomics only, and combined models were developed to predict lesion outcome.

Results

The phantom study identified 15/46 reproducible and robust radiomics features that were subsequently used in the patient models. A lesion response model with zone percentage (ZP) and mean absorbed dose achieved an AUC of 0.729 (95% CI 0.702–0.758), and a progression model with zone size nonuniformity (ZSN) and absorbed dose achieved a c-index of 0.803 (95% CI 0.790–0.815) on nested cross-validation (CV). Although the combined models outperformed the radiomics only and absorbed dose only models, statistical significance was not achieved with the current limited data set to establish expected superiority.

Conclusion

We have developed new lesion-level response and progression models using textural radiomics features, derived from ⁹⁰Y PET combined with mean absorbed dose for predicting outcome in radioembolization. These encouraging, but limited results, will need further validation in independent and larger datasets prior to any clinical adoption.

Supplementary Information

Supplementary information accompanies this paper at 10.1186/s40658-020-00340-9.

Development and Validation of a Machine Learning Model to Explore Tyrosine Kinase Inhibitor Response in Patients With Stage IV EGFR Variant-Positive Non-Small Cell Lung Cancer

IMPORTANCE

An end-to-end efficacy evaluation approach for identifying progression risk after epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitor (TKI) therapy in patients with stage IV EGFR variant-positive non-small cell lung cancer (NSCLC) is lacking.

OBJECTIVE

To propose a clinically applicable large-scale bidirectional generative adversarial network for predicting the efficacy of EGFR-TKI therapy in patients with NSCLC.

DESIGN, SETTING, AND PARTICIPANTS

This diagnostic/prognostic study enrolled 465 patients from January 1, 2010, to August 1, 2017, with follow-up from February 1, 2010, to June 1, 2020. A deep learning (DL) semantic signature to predict progression-free survival (PFS) was constructed in the training cohort, validated in 2 external validation and 2 control cohorts, and compared with the radiomics signature.

EXPOSURES

An end-to-end bidirectional generative adversarial network framework was designed to predict the progression risk in patients with NSCLC.

MAIN OUTCOMES AND MEASURES

The primary end point was PFS, considering the time from the initiation of therapy to the date of recurrence, confirmed disease progression, or death.

RESULTS

A total of 342 patients with stage IV EGFR variant-positive NSCLC receiving EGFR-TKI therapy met the inclusion criteria. Of these, 145 patients from 2 of the hospitals (n = 117 and 28) formed a training cohort (mean [SD] age, 61 [11] years; 87 [60.0%] female), and the patients from 2 other hospitals comprised 2 external validation cohorts (validation cohort 1: n = 101; mean [SD] age, 57 [12] years; 60 [59.4%] female; and validation cohort 2: n = 96, mean [SD] age, 58 [9] years; 55 [57.3%] female). Fifty-six patients with advanced-stage EGFR variant-positive NSCLC (mean [SD] age, 52 [11] years; 26 [46.4%] female) and 67 patients with advanced-stage EGFR wild-type NSCLC (mean [SD] age, 54 [10] years; 10 [15.0%] female) who received first-line chemotherapy were included. A total of 90 (26%) receiving EGFR-TKI therapy with a high risk of rapid disease progression were identified (median [range] PFS, 7.3 [1.4-32.0] months in the training cohort, 5.0 [0.6-34.6] months in validation cohort 1, and 6.4 [1.8-20.1] months, in validation cohort 2) using the DL semantic signature.The PFS decreased by 36% (hazard ratio, 2.13; 95% CI, 1.30-3.49; P < .001) compared with that in other patients (median [range] PFS, 11.5 [1.5-64.2] months in the training cohort, 10.9 [1.1-50.5] in validation cohort 1, and 8.9 [0.8-40.6] months in validation cohort 2. No significant differences were observed when comparing the PFS of high-risk patients receiving EGFR-TKI therapy with the chemotherapy cohorts (median PFS, 6.9 vs 4.4 months; P = .08). In terms of predicting the tumor progression risk after EGFR-TKI therapy, clinical decisions based on the DL semantic signature led to better survival outcomes than those based on radiomics signature across all risk probabilities by the decision curve analysis.

CONCLUSIONS AND RELEVANCE

This diagnostic/prognostic study provides a clinically applicable approach for identifying patients with stage IV EGFR variant-positive NSCLC who are not likely to benefit from EGFR-TKI therapy. The end-to-end DL-derived semantic features eliminated all manual interventions required while using previous radiomics methods and have a better prognostic performance.

Artificial intelligence applications for oncological positron emission tomography imaging

Positron emission tomography (PET), a functional and dynamic molecular imaging technique, is generally used to reveal tumors' biological behavior. Radiomics allows a high-throughput extraction of multiple features from images with artificial intelligence (AI) approaches and develops rapidly worldwide. Quantitative and objective features of medical images have been explored to recognize reliable biomarkers, with the development of PET radiomics. This paper will review the current clinical exploration of PET-based classical machine learning and deep learning methods, including disease diagnosis, the prediction of histological subtype, gene mutation status, tumor metastasis, tumor relapse, therapeutic side effects, therapeutic intervention and evaluation of prognosis. The applications of AI in oncology will be mainly discussed. The image-guided biopsy or surgery assisted by PET-based AI will be introduced as well. This paper aims to present the applications and methods of AI for PET imaging, which may offer important details for further clinical studies. Relevant precautions are put forward and future research directions are suggested.

Radiographic Texture Reproducibility: The Impact of Different Materials, their Arrangement, and Focal Spot Size

Background:

Feature reproducibility is a critical issue in quantitative radiomic studies. The aim of this study is to assess how radiographic radiomic textures behave against changes in phantom materials, their arrangements, and focal spot size.

Method:

A phantom with detachable parts was made using wood, sponge, Plexiglas, and rubber. Each material had 1 cm thickness and was imaged for consecutive time. The phantom also was imaged by change in the arrangement of its materials. Imaging was done with two focal spot sizes including 0.6 and 1.2 mm. All images were acquired with a digital radiography machine. Several texture features were extracted from the same size region of interest in all images. To assess reproducibility, coefficient of variation (COV), intraclass correlation coefficient (ICC), and Bland–Altman tests were used.

Results:

Results show that 59%, 50%, and 4.5% of all features are most reproducible (COV ≤5%) against change in focal spot size, material arrangements, and phantom's materials, respectively. Results on Bland–Altman analysis showed that there is just a nonreproducible feature against change in the focal spot size. On the ICC results, we observed that the ICCs for more features are >0.90 and there were few features with ICC lower than 0.90.

Conclusion:

We showed that radiomic textures are vulnerable against changes in materials, arrangement, and different focal spot sizes. These results suggest that a careful analysis of the effects of these parameters is essential before any radiomic clinical application.

Radiomics in PET/CT: Current Status and Future AI-Based Evolutions

This short review aims at providing the readers with an update on the current status, as well as future perspectives in the quickly evolving field of radiomics applied to the field of PET/CT imaging. Numerous pitfalls have been identified in study design, data acquisition, segmentation, features calculation and modeling by the radiomics community, and these are often the same issues across all image modalities and clinical applications, however some of these are specific to PET/CT (and SPECT/CT) imaging and therefore the present paper focuses on those. In most cases, recommendations and potential methodological solutions do exist and should therefore be followed to improve the overall quality and reproducibility of published studies. In terms of future evolutions, the techniques from the larger field of artificial intelligence (AI), including those relying on deep neural networks (also known as deep learning) have already shown impressive potential to provide solutions, especially in terms of automation, but also to maybe fully replace the tools the radiomics community has been using until now in order to build the usual radiomics workflow. Some important challenges remain to be addressed before the full impact of AI may be realized but overall the field has made striking advances over the last few years and it is expected advances will continue at a rapid pace.

Are lesion features reproducible between 18F-FDG PET/CT images when acquired on analog or digital PET/CT scanners?

OBJECTIVES

To compare lesion features extracted from ¹⁸F-FDG PET/CT images acquired on analog and digital scanners, on consecutive imaging data from the same subjects.

METHODS

Whole-body ¹⁸F-FDG PET/CT images from 55 oncological patients were acquired twice after a single ¹⁸F-FDG injection, with a digital and an analog PET/CT scanner, alternately. Twenty-nine subjects were examined first on the digital, and 26 first on the analog equipment. Image reconstruction was performed using manufacturer standard clinical protocols and protocols that fulfilled EARL1 specifications. Twenty-five features based on lesion standardized uptake value (SUV) and geometry were assessed. To compare these features, intraclass correlation coefficient (ICC), relative difference (RD), absolute value of RD (|RD|), and repeatability coefficient (RC) were used.

RESULTS

In total, 323 ¹⁸F-FDG avid lesions were identified. High agreement (ICC > 0.75) was obtained for most of the lesion features pulled out from both scanners' imaging data, especially when reconstruction protocols fulfilled EARL1 specifications. For EARL1 reconstruction images, the features frequently used in clinics, SUV_max, SUV_peak, SUV_mean, metabolic tumor volume, and total lesion glycolysis, reached an ICC of 0.92, 0.95, 0.87, 0.98, and 0.98, and a median RD (digital-analog) of 3%, 5%, 4%, - 3% and 1%, respectively. Using standard reconstruction protocols, the ICC were 0.84, 0.93, 0.80, 0.98, and 0.98, and the RD were 20%, 11%, 13%, - 7%, and 7%, respectively.

CONCLUSION

Under controlled acquisition and reconstruction parameters, most of the features studied can be used for research and clinical work. This is especially important for multicenter studies and patient follow-ups.

KEY POINTS

• Using manufacturer standard clinical reconstruction protocols, lesions SUV was significantly higher when using the digital scanner, especially the SUV_max that was approximately 20% higher. • High agreement was obtained for the majority of the lesion features when using reconstruction protocols that fulfilled EARL1 specifications. • Longitudinal patient studies can be performed interchangeably between digital and analog scanners when both fulfill EARL1 specifications.

A systematic review and meta-analysis of the prognostic value of radiomics based models in non-small cell lung cancer treated with curative radiotherapy

BACKGROUND AND PURPOSE

Radiomics allows extraction of quantifiable features from imaging. This study performs a systematic review and meta-analysis of the performance of radiomics based prognostic models in non-small cell lung cancer (NSCLC).

MATERIALS AND METHODS

A literature review was performed following PRISMA guidelines. Medline, EMBASE and Cochrane databases were searched for articles investigating radiomics features predictive of overall survival (OS) in NSCLC treated with curative intent radiotherapy. A random-effects meta-analysis of Harrell's Concordance Index (C-index) was performed on the performance of radiomics models.

RESULTS

Of the 2746 articles retrieved, 40 studies of 55 datasets and 6223 patients were eligible for inclusion in the systematic review. There was significant heterogeneity in the methodology for feature selection and model development. Twelve datasets reported the C-index of radiomics based models in predicting OS and were included in the meta-analysis. The C-index random effects estimate was 0.57 (95% CI 0.53-0.62). There was significant heterogeneity (I² = 70.3%).

CONCLUSIONS

Based on this review, radiomics based models for lung cancer have to date demonstrated modest prognostic capabilities. Future research should consider using standardised radiomics features, robust feature selection and model development, and deep learning techniques, absolving the need for pre-defined features, to improve imaging-based models.

Potentials and caveats of AI in hybrid imaging

State-of-the-art patient management frequently mandates the investigation of both anatomy and physiology of the patients. Hybrid imaging modalities such as the PET/MRI, PET/CT and SPECT/CT have the ability to provide both structural and functional information of the investigated tissues in a single examination. With the introduction of such advanced hardware fusion, new problems arise such as the exceedingly large amount of multi-modality data that requires novel approaches of how to extract a maximum of clinical information from large sets of multi-dimensional imaging data. Artificial intelligence (AI) has emerged as one of the leading technologies that has shown promise in facilitating highly integrative analysis of multi-parametric data. Specifically, the usefulness of AI algorithms in the medical imaging field has been heavily investigated in the realms of (1) image acquisition and reconstruction, (2) post-processing and (3) data mining and modelling. Here, we aim to provide an overview of the challenges encountered in hybrid imaging and discuss how AI algorithms can facilitate potential solutions. In addition, we highlight the pitfalls and challenges in using advanced AI algorithms in the context of hybrid imaging and provide suggestions for building robust AI solutions that enable reproducible and transparent research.

From multisource data to clinical decision aids in radiation oncology: The need for a clinical data science community

Big data are no longer an obstacle; now, by using artificial intelligence (AI), previously undiscovered knowledge can be found in massive data collections. The radiation oncology clinic daily produces a large amount of multisource data and metadata during its routine clinical and research activities. These data involve multiple stakeholders and users. Because of a lack of interoperability, most of these data remain unused, and powerful insights that could improve patient care are lost. Changing the paradigm by introducing powerful AI analytics and a common vision for empowering big data in radiation oncology is imperative. However, this can only be achieved by creating a clinical data science community in radiation oncology. In this work, we present why such a community is needed to translate multisource data into clinical decision aids.

Sensitivity analysis of FDG PET tumor voxel cluster radiomics and dosimetry for predicting mid-chemoradiation regional response of locally advanced lung cancer

We investigated the sensitivity of regional tumor response prediction to variability in voxel clustering techniques, imaging features, and machine learning algorithms in 25 patients with locally advanced non-small cell lung cancer (LA-NSCLC) enrolled on the FLARE-RT clinical trial. Metabolic tumor volumes (MTV) from pre-chemoradiation (PETpre) and mid-chemoradiation fluorodeoxyglucose-positron emission tomography (FDG PET) images (PETmid) were subdivided into K-means or hierarchical voxel clusters by standardized uptake values (SUV) and 3D-positions. MTV cluster separability was evaluated by CH index, and morphologic changes were captured by Dice similarity and centroid Euclidean distance. PETpre conventional features included SUVmean, MTV/MTV cluster size, and mean radiation dose. PETpre radiomics consisted of 41 intensity histogram and 3D texture features (PET Oncology Radiomics Test Suite) extracted from MTV or MTV clusters. Machine learning models (multiple linear regression, support vector regression, logistic regression, support vector machines) of conventional features or radiomic features were constructed to predict PETmid response. Leave-one-out-cross-validated root-mean-squared-error (RMSE) for continuous response regression (ΔSUVmean) and area-under-receiver-operating-characteristic-curve (AUC) for binary response classification were calculated. K-means MTV 2-clusters (MTVhi, MTVlo) achieved maximum CH index separability (Friedman p < 0.001). Between PETpre and PETmid, MTV cluster pairs overlapped (Dice 0.70-0.87) and migrated 0.6-1.1 cm. PETmid ΔSUVmean response prediction was superior in MTV and MTVlo (RMSE = 0.17-0.21) compared to MTVhi (RMSE = 0.42-0.52, Friedman p < 0.001). PETmid ΔSUVmean response class prediction performance trended higher in MTVlo (AUC = 0.83-0.88) compared to MTVhi (AUC = 0.44-0.58, Friedman p = 0.052). Models were more sensitive to MTV/MTV cluster regions (Friedman p = 0.026) than feature sets/algorithms (Wilcoxon signed-rank p = 0.36). Top-ranked radiomic features included GLZSM-LZHGE (large-zone-high-SUV), GTSDM-CP (cluster-prominence), GTSDM-CS (cluster-shade) and NGTDM-CNT (contrast). Top-ranked features were consistent between MTVhi and MTVlo cluster pairs but varied between MTVhi-MTVlo clusters, reflecting distinct regional radiomic phenotypes. Variability in tumor voxel cluster response prediction can inform robust radiomic target definition for risk-adaptive chemoradiation in patients with LA-NSCLC. FLARE-RT trial: NCT02773238.

How to read and review papers on machine learning and artificial intelligence in radiology: a survival guide to key methodological concepts

In recent years, there has been a dramatic increase in research papers about machine learning (ML) and artificial intelligence in radiology. With so many papers around, it is of paramount importance to make a proper scientific quality assessment as to their validity, reliability, effectiveness, and clinical applicability. Due to methodological complexity, the papers on ML in radiology are often hard to evaluate, requiring a good understanding of key methodological issues. In this review, we aimed to guide the radiology community about key methodological aspects of ML to improve their academic reading and peer-review experience. Key aspects of ML pipeline were presented within four broad categories: study design, data handling, modelling, and reporting. Sixteen key methodological items and related common pitfalls were reviewed with a fresh perspective: database size, robustness of reference standard, information leakage, feature scaling, reliability of features, high dimensionality, perturbations in feature selection, class balance, bias-variance trade-off, hyperparameter tuning, performance metrics, generalisability, clinical utility, comparison with traditional tools, data sharing, and transparent reporting.Key Points• Machine learning is new and rather complex for the radiology community.• Validity, reliability, effectiveness, and clinical applicability of studies on machine learning can be evaluated with a proper understanding of key methodological concepts about study design, data handling, modelling, and reporting.• Understanding key methodological concepts will provide a better academic reading and peer-review experience for the radiology community.