4DCT imaging to assess radiomics feature stability: An investigation for thoracic cancers

Technical and clinical considerations of a physical liver phantom for CT radiomics analysis

OBJECTIVE

This study identifies key characteristics to help build a physical liver computed tomography (CT) phantom for radiomics harmonization; particularly, the higher-order texture metrics.

MATERIALS AND METHODS

CT scans of a radiomics phantom comprising of 18 novel 3D printed inserts with varying size, shape, and material combinations were acquired on a 64-slice CT scanner (Brilliance 64, Philips Healthcare). The images were acquired at 120 kV, 250 mAs, CTDIvol of 16.36 mGy, 2 mm slice thickness, and iterative noise-reduction reconstruction (iDose, Philips Healthcare, Andover, MA). Radiomics analysis was performed using the Cancer Imaging Phenomics Toolkit (CaPTk), following automated segmentation of 3D regions of interest (ROI) of the 18 inserts. The findings were compared to three additional ROI obtained of an anthropomorphic liver phantom, a patient liver CT scan, and a water phantom, at comparable imaging settings. Percentage difference in radiomic metrics values between phantom and tissue was used to assess the biological equivalency and <10% was used to claim equivalent.

RESULTS

The HU for all 18 ROI from the phantom ranged from -30 to 120 which is within clinically observed HU range of the liver, showing that our phantom material (T3-6B) is representative of biological CT tissue densities (liver) with >50% radiomic features having <10% difference from liver tissue. Based on the assessment of the Neighborhood Gray Tone Difference Matrix (NGTDM) metrics it is evident that the water phantom ROI show extreme values compared to the ROIs from the phantom. This result may further reinforce the difference between a structureless quantity such as water HU values and tissue HU values found in liver.

CONCLUSION

The 3-D printed patterns of the constructed radiomics phantom cover a wide span of liver tissue textures seen in CT images. Using our results, texture metrics can be selectively harmonized to establish clinically relevant and reliable radiomics panels.

Multi-omics and Multi-VOIs to predict esophageal fistula in esophageal cancer patients treated with radiotherapy

OBJECTIVE

This study aimed to develop a prediction model for esophageal fistula (EF) in esophageal cancer (EC) patients treated with intensity-modulated radiation therapy (IMRT), by integrating multi-omics features from multiple volumes of interest (VOIs).

METHODS

We retrospectively analyzed pretreatment planning computed tomographic (CT) images, three-dimensional dose distributions, and clinical factors of 287 EC patients. Nine groups of features from different combination of omics [Radiomics (R), Dosiomics (D), and RD (the combination of R and D)], and VOIs [esophagus (ESO), gross tumor volume (GTV), and EG (the combination of ESO and GTV)] were extracted and separately selected by unsupervised (analysis of variance (ANOVA) and Pearson correlation test) and supervised (Student T test) approaches. The final model performance was evaluated using five metrics: average area under the receiver-operator-characteristics curve (AUC), accuracy, precision, recall, and F1 score.

RESULTS

For multi-omics using RD features, the model performance in EG model shows: AUC, 0.817 ± 0.031; 95% CI 0.805, 0.825; p < 0.001, which is better than single VOI (ESO or GTV).

CONCLUSION

Integrating multi-omics features from multi-VOIs enables better prediction of EF in EC patients treated with IMRT. The incorporation of dosiomics features can enhance the model performance of the prediction.

Limited impact of discretization/interpolation parameters on the predictive power of CT radiomic features in a surgical cohort of pancreatic cancer patients

PURPOSE

To explore the variation of the discriminative power of CT (Computed Tomography) radiomic features (RF) against image discretization/interpolation in predicting early distant relapses (EDR) after upfront surgery.

MATERIALS AND METHODS

Data of 144 patients with pre-surgical high contrast CT were processed consistently with IBSI (Image Biomarker Standardization Initiative) guidelines. Image interpolation/discretization parameters were intentionally changed, including cubic voxel size (0.21-27 mm³) and binning (32-128 grey levels) in a 15 parameter's sets. After excluding RF with poor inter-observer delineation agreement (ICC < 0.80) and not negligible inter-scanner variability, the variation of 80 RF against discretization/interpolation was first quantified. Then, their ability in classifying patients with early distant relapses (EDR, < 10 months, previously assessed at the first quartile value of time-to-relapse) was investigated in terms of AUC (Area Under Curve) variation for those RF significantly associated to EDR.

RESULTS

Despite RF variability against discretization/interpolation parameters was large and only 30/80 RF showed %COV < 20 (%COV = 100*STDEV/MEAN), AUC changes were relatively limited: for 30 RF significantly associated with EDR (AUC values around 0.60-0.70), the mean values of SD of AUC variability and AUC range were 0.02 and 0.05 respectively. AUC ranges were between 0.00 and 0.11, with values ≤ 0.05 in 16/30 RF. These variations were further reduced when excluding the extreme values of 32 and 128 for grey levels (Average AUC range 0.04, with values between 0.00 and 0.08).

CONCLUSIONS

The discriminative power of CT RF in the prediction of EDR after upfront surgery for pancreatic cancer is relatively invariant against image interpolation/discretization within a large range of voxel sizes and binning.

Radiomic feature repeatability and its impact on prognostic model generalizability: A multi-institutional study on nasopharyngeal carcinoma patients

BACKGROUND AND PURPOSE

To investigate the radiomic feature (RF) repeatability via perturbation and its impact on cross-institutional prognostic model generalizability in Nasopharyngeal Carcinoma (NPC) patients.

MATERIALS AND METHODS

286 and 183 NPC patients from two institutions were included for model training and validation. Perturbations with random translations and rotations were applied to contrast-enhanced T1-weighted (CET1-w) MR images. RFs were extracted from primary tumor volume under a wide range of image filtering and discretization settings. RF repeatability was assessed by intraclass correlation coefficient (ICC), which was used to equally separate the RFs into low- and high-repeatable groups by the median value. After feature selection, multivariate Cox regression and Kaplan-Meier analysis were independently employed to develop and analyze prognostic models. Concordance index (C-index) and P-value from log-rank test were used to assess model performance.

RESULTS

Most textural RFs from high-pass wavelet-filtered images were susceptible to image perturbations. It was more prominent when a smaller discretization bin number was used (e.g., 8, mean ICC = 0.69). Using high-repeatable RFs for model development yielded a significantly higher C-index (0.63) in the validation cohort than when only low-repeatable RFs were used (0.57, P = 0.024), suggesting higher model generalizability. Besides, significant risk stratification in the validation cohort was observed only when high-repeatable RFs were used (P < 0.001).

CONCLUSION

Repeatability of RFs from high-pass wavelet-filtered CET1-w MR images of primary NPC tumor was poor, particularly when a smaller bin number was used. Exclusive use of high-repeatable RFs is suggested to safeguard model generalizability for wide-spreading clinical utilization.

Machine Learning Radiomics Model for External and Internal Respiratory Motion Correlation Prediction in Lung Tumor

Objectives: The complexity and specificity of lung tumor motion render it necessary to determine the external and internal correlation individually before applying indirect tumor tracking. However, the correlation cannot be determined from patient respiratory and tumor clinical characteristics before treatment. The purpose of this study is to present a machine learning model for an external/internal correlation prediction that is based on computed tomography (CT) radiomic features. Methods: 4-dimensional computed tomography (4DCT) images of 67 patients were collected retrospectively, and the external/internal correlation of lung tumor was calculated based on Spearman's rank correlation coefficient. Radiomic features were extracted from average intensity projection and the light gradient boosting machine (LightGBM)-based cross-validation (the recursive elimination method) was used for feature selection. The LightGBM framework forecasting models with classification thresholds 0.7, 0.8, and 0.9 are established using stratified 5-fold cross-validation. Model performance was assessed using receiver operating characteristics, sensitivity, and specificity. Results: There were 16, 18, and 13 features selected for models 0.7, 0.8, and 0.9, respectively. Texture features are of great importance in external/internal correlation prediction compared to other features in all models. The sensitivities of the predictions in models 0.7, 0.8, and 0.9 were 0.800 ± 0.126, 0.829 ± 0.140, and 0.864 ± 0.086, respectively. The specificities were 0.771 ± 0.114, 0.936 ± 0.0581, and 0.839 ± 0.101, whereas the area under the curve (AUC) was 0.837, 0.946, and 0.877, respectively. Conclusions: Our findings indicate that radiomics is an effective tool for respiratory motion correlation prediction, which can extract tumor motion characteristics. We proposed a machine learning framework for correlation prediction in the motion management strategy for lung tumor patients.

Uniqueness of radiomic features in non-small cell lung cancer

PURPOSE

The uniqueness of radiomic features, combined with their reproducibility, determines the reliability of radiomic studies. This study is to test the hypothesis that radiomic features extracted from a defined region of interest (ROI) are unique to the underlying structure (e.g., tumor).

APPROACH

Two cohorts of non-small cell lung cancer (NSCLC) patients were retrospectively retrieved from a GE and a Siemens CT scanner. The lung nodules (ROI) were delineated manually and radiomic features were extracted using IBEX. The same ROI was then translocated randomly to four other tissue regions of the same set of images: adipose, heart, lung beyond nodule, and muscle for radiomic feature extraction. Coefficient of variation (CV) within different ROIs and concordance correlation coefficient (CCC) between lung nodule and a given tissue region were calculated to test to determine feature uniqueness. The radiomic features were considered nonunique when (1) the CV < 10% and CCC > 0.85 for over 50% of patients; and (2) the CCC > 0.85 appeared in ≥2 tissue regions beyond the defined region.

RESULTS

In total, 14 patients from GE and 18 patients from Siemens are analyzed. The results show that 12 features fall below the 10% CV threshold for over 50% of patients in the GE cohort and 29 features in the Siemens cohort. According to CCC, 18 radiomic features in GE and 16 features in Siemens are identified as nonunique, with 11 overlapping features. Combining CV and CCC, 9 of 123 calculated features (7.3%) are identified as nonunique to a defined ROI.

CONCLUSIONS

Radiomic feature uniqueness should be considered to improve the reliability of radiomics study.

Function-Wise Dual-Omics analysis for radiation pneumonitis prediction in lung cancer patients

Purpose: This study investigates the impact of lung function on radiation pneumonitis prediction using a dual-omics analysis method.

Methods: We retrospectively collected data of 126 stage III lung cancer patients treated with chemo-radiotherapy using intensity-modulated radiotherapy, including pre-treatment planning CT images, radiotherapy dose distribution, and contours of organs and structures. Lung perfusion functional images were generated using a previously developed deep learning method. The whole lung (WL) volume was divided into function-wise lung (FWL) regions based on the lung perfusion functional images. A total of 5,474 radiomics features and 213 dose features (including dosiomics features and dose-volume histogram factors) were extracted from the FWL and WL regions, respectively. The radiomics features (R), dose features (D), and combined dual-omics features (RD) were used for the analysis in each lung region of WL and FWL, labeled as WL-R, WL-D, WL-RD, FWL-R, FWL-D, and FWL-RD. The feature selection was carried out using ANOVA, followed by a statistical F-test and Pearson correlation test. Thirty times train-test splits were used to evaluate the predictability of each group. The overall average area under the receiver operating characteristic curve (AUC), accuracy, precision, recall, and f1-score were calculated to assess the performance of each group.

Results: The FWL-RD achieved a significantly higher average AUC than the WL-RD group in the training (FWL-RD: 0.927 ± 0.031, WL-RD: 0.849 ± 0.064) and testing cohorts (FWL-RD: 0.885 ± 0.028, WL-RD: 0.762 ± 0.053, p < 0.001). When using radiomics features only, the FWL-R group yielded a better classification result than the model trained with WL-R features in the training (FWL-R: 0.919 ± 0.036, WL-R: 0.820 ± 0.052) and testing cohorts (FWL-R: 0.862 ± 0.028, WL-R: 0.750 ± 0.057, p < 0.001). The FWL-D group obtained an average AUC of 0.782 ± 0.032, obtaining a better classification performance than the WL-D feature-based model of 0.740 ± 0.028 in the training cohort, while no significant difference was observed in the testing cohort (FWL-D: 0.725 ± 0.064, WL-D: 0.710 ± 0.068, p = 0.54).

Conclusion: The dual-omics features from different lung functional regions can improve the prediction of radiation pneumonitis for lung cancer patients under IMRT treatment. This function-wise dual-omics analysis method holds great promise to improve the prediction of radiation pneumonitis for lung cancer patients.

Deep learning versus iterative reconstruction on image quality and dose reduction in abdominal CT: a live animal study

Objective. While simulated low-dose CT images and phantom studies cannot fully approximate subjective and objective effects of deep learning (DL) denoising on image quality, live animal models may afford this assessment. This study is to investigate the potential of DL in CT dose reduction on image quality compared to iterative reconstruction (IR). Approach. The upper abdomen of a live 4 year old sheep was scanned on a CT scanner at different exposure levels. Images were reconstructed using FBP and ADMIRE with 5 strengths. A modularized DL network with 5 modules was used for image reconstruction via progressive denoising. Radiomic features were extracted from a region over the liver. Concordance correlation coefficient (CCC) was applied to quantify agreement between any two sets of radiomic features. Coefficient of variation was calculated to measure variation in a radiomic feature series. Structural similarity index (SSIM) was used to measure the similarity between any two images. Diagnostic quality, low-contrast detectability, and image texture were qualitatively evaluated by two radiologists. Pearson correlation coefficient was computed across all dose-reconstruction/denoising combinations. Results. A total of 66 image sets, with 405 radiomic features extracted from each, are analyzed. IR and DL can improve diagnostic quality and low-contrast detectability and similarly modulate image texture features. In terms of SSIM, DL has higher potential in preserving image structure. There is strong correlation between SSIM and radiologists’ evaluations for diagnostic quality (0.559) and low-contrast detectability (0.635) but moderate correlation for texture (0.313). There is moderate correlation between CCC of radiomic features and radiologists’ evaluation for diagnostic quality (0.397), low-contrast detectability (0.417), and texture (0.326), implying that improvement of image features may not relate to improvement of diagnostic quality. Conclusion. DL shows potential to further reduce radiation dose while preserving structural similarity, while IR is favored by radiologists and more predictably alters radiomic features.

Radiomic assessment of oesophageal adenocarcinoma: a critical review of 18F-FDG PET/CT, PET/MRI and CT

OBJECTIVES

Radiomic models present an avenue to improve oesophageal adenocarcinoma assessment through quantitative medical image analysis. However, model selection is complicated by the abundance of available predictors and the uncertainty of their relevance and reproducibility. This analysis reviews recent research to facilitate precedent-based model selection for prospective validation studies.

METHODS

This analysis reviews research on 18F-FDG PET/CT, PET/MRI and CT radiomics in oesophageal adenocarcinoma between 2016 and 2021. Model design, testing and reporting are evaluated according to the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) score and Radiomics Quality Score (RQS). Key results and limitations are analysed to identify opportunities for future research in the area.

RESULTS

Radiomic models of stage and therapeutic response demonstrated discriminative capacity, though clinical applications require greater sensitivity. Although radiomic models predict survival within institutions, generalisability is limited. Few radiomic features have been recommended independently by multiple studies.

CONCLUSIONS

Future research must prioritise prospective validation of previously proposed models to further clinical translation.

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

Purpose

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

Materials and Methods

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

Results

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

Conclusion

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

The impact of phantom design and material-dependence on repeatability and reproducibility of CT-based radiomics features

PURPOSE

To understand the design of radiomics phantom and material-dependence on repeatability and reproducibility of CT radiomics features METHODS: : A radiomics phantom consisting of various materials with uniformity, textural and biological components, was constructed. The phantom was scanned with different manufacturer CT scanners and the scans were repeated multiple times on the same scanner with different acquisition settings as kVp, mAs, orientation, field of view (FOV), slice thickness, pitch, reconstruction kernels and acquisition mode. A total of 72 phantom scans were included. For each scan, 18 different regions of interest (ROI) were contoured and 708 radiomics features were extracted from each ROI via an open source radiomics tool, IBEX. To relate the phantom data to patient data, the radiomics features from different phantom materials were compared with those extracted from 50 patients' images of five disease sites as brain, head-and-neck, breast, liver and lung cases using box-plots comparison and principal component analysis (PCA). The temporal stability of imaging features was then evaluated with respect to a controlled scenario (test-retest) via the intra-class correlation coefficient (ICC). The reproducibility of radiomics features with respect to different scanners or acquisition settings were further evaluated with concordance correlation coefficients (CCC).

RESULTS

Among all phantom materials, the biological component had feature values closest to human tissues, especially for tumors in brain and liver. The textural component showed similar ranges of variation to lung lesions, particularly for cartridges of rice, cereal, and the 3D-printed textural phantom with fine and rough-grid. It also showed that certain materials, such as polystyrene foam, plaster and peanuts, did not have comparable values to human tissue and could be excluded for future phantom design. High repeatability was observed in the test-retest study as indicated by an ICC value of 0.998 ± 0.020. All materials were used for feature stability analysis. For the inter-scanner study, shape-related features were the most-reliable category with 94% of features having CCC ≥0.9, while GOH were the least-reliable with only 14.6% meeting the criteria. For the intra-scanner study, the reproducibility of CT-based radiomics features showed material-dependence. In general, the instability of radiomics features introduced by kVp, mAs, pitch, acquisition mode and orientation were relatively mild. However, the homogeneous materials were more vulnerable to those changes compared to materials with textural patterns. Regardless of material compositions, resolution parameters like FOV and slice thickness, could have large impact on feature stability. Switching between standard and bone reconstruction kernels could also result significant changes to feature reproducibility.

CONCLUSION

We have built a radiomics phantom using materials that cover a wide span of tumor textures seen in oncological CT images. The designed phantom presents a preliminary opportunity for investigating reproducibility of radiomics features and the reproducibility can be material dependent. Thus, in the radiomics quality assurance design, it is important to choose appropriate materials that can provide a close range of radiomics features to patients with specific disease sites dependency taken into consideration. This article is protected by copyright. All rights reserved.

Vulnerabilities of radiomic features to respiratory motion on four-dimensional computed tomography-based average intensity projection images: A phantom study

PURPOSE

To evaluate the influence of respiratory motion on the robustness of radiomic features on four-dimensional computed tomography (4DCT)-based average intensity projection (AIP) images by employing an anthropomorphic chest phantom.

METHODS

Three spherical objects (φ30 mm), namely, acrylic (100 Hounsfield unit [HU], homogeneous), rubber (-140 HU, homogeneous), and cork (-630 HU, heterogeneous), were moved with motion amplitudes of 0, 1, 2.5, 4, 6, 8, and 10 mm in the phantom, and 4DCT scans were repeated at four different locations. Thereafter, the AIP images were generated considering the average of the 10 respiratory phases of the 4DCT images. Further, the targets were manually delineated on the AIP images in the lung window setting. A total of 851 radiomic features, including 107 unfiltered features and 744 wavelet filter-based features, were extracted from the region of interest for each material. The feature robustness among the different target motion amplitude (ε) was evaluated by normalizing the feature variability of the target motion relative to the variability of data from 573 patients with early-stage non-small cell lung cancer. The features with absolute ε values ≤0.5 were considered highly robust to target motions.

RESULTS

The percentage of robust unfiltered and wavelet filter-based features with a motion amplitude of 1 mm was greater than 83.2% and 93.4%, respectively; however, the percentage decreased by more than 24.3% and 17.6%, respectively, for motion amplitudes greater than 2.5 mm. The movement of cork had a small effect on the feature robustness compared to that of acrylic and rubber, regardless of the target motion amplitudes.

CONCLUSIONS

Our phantom study demonstrated that target motion amplitudes ≤1 mm led to the robustness of radiomic features on the 4DCT-based AIP images of thoracic regions. The frequency components and directions of the wavelet filters may be essential factors in 4DCT-based radiomic analysis.

The Impact of Resampling and Denoising Deep Learning Algorithms on Radiomics in Brain Metastases MRI

Simple Summary

Due to the central role of magnetic resonance Imaging (MRI) in the management of patients with cancer, waiting lists exceed clinically relevant delays. For this reason, many research groups and MRI manufacturers develop algorithms as resampling and denoising models to allow faster acquisition time without deterioration in image quality. Whereas these algorithms are available in all new MRI, it is not clear how they will impact image features as well as the validity of statistical model of radiomics which use deep images characteristics to predict treatment outcome. The aim of this study was to develop resampling and denoising deep learning (DL) models and evaluate their impact on radiomics from post-Gd-T1w-MRI brain images with brain metastases. We show that resampling and denoising DL models reconstruct low resolution and noised MRI images acquired quickly into high quality images. While fast acquisition loses most of the radiomic-features and invalidates predictive radiomic models, DL models restore these parameters.

Abstract

Background: Magnetic resonance imaging (MRI) is predominant in the therapeutic management of cancer patients, unfortunately, patients have to wait a long time to get an appointment for examination. Therefore, new MRI devices include deep-learning (DL) solutions to save acquisition time. However, the impact of these algorithms on intensity and texture parameters has been poorly studied. The aim of this study was to evaluate the impact of resampling and denoising DL models on radiomics. Methods: Resampling and denoising DL model was developed on 14,243 T1 brain images from 1.5T-MRI. Radiomics were extracted from 40 brain metastases from 11 patients (2049 images). A total of 104 texture features of DL images were compared to original images with paired t-test, Pearson correlation and concordance-correlation-coefficient (CCC). Results: When two times shorter image acquisition shows strong disparities with the originals concerning the radiomics, with significant differences and loss of correlation of 79.81% and 48.08%, respectively. Interestingly, DL models restore textures with 46.15% of unstable parameters and 25.96% of low CCC and without difference for the first-order intensity parameters. Conclusions: Resampling and denoising DL models reconstruct low resolution and noised MRI images acquired quickly into high quality images. While fast MRI acquisition loses most of the radiomic features, DL models restore these parameters.

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.

Application of Radiomics and Artificial Intelligence for Lung Cancer Precision Medicine

Medical imaging is the standard-of-care for early detection, diagnosis, treatment planning, monitoring, and image-guided interventions of lung cancer patients. Most medical images are stored digitally in a standardized Digital Imaging and Communications in Medicine format that can be readily accessed and used for qualitative and quantitative analysis. Over the several last decades, medical images have been shown to contain complementary and interchangeable data orthogonal to other sources such as pathology, hematology, genomics, and/or proteomics. As such, "radiomics" has emerged as a field of research that involves the process of converting standard-of-care images into quantitative image-based data that can be merged with other data sources and subsequently analyzed using conventional biostatistics or artificial intelligence (AI) methods. As radiomic features capture biological and pathophysiological information, these quantitative radiomic features have shown to provide rapid and accurate noninvasive biomarkers for lung cancer risk prediction, diagnostics, prognosis, treatment response monitoring, and tumor biology. In this review, radiomics and emerging AI methods in lung cancer research are highlighted and discussed including advantages, challenges, and pitfalls.

Optimising use of 4D-CT phase information for radiomics analysis in lung cancer patients treated with stereotactic body radiotherapy

Purpose. 4D-CT is routine imaging for lung cancer patients treated with stereotactic body radiotherapy. No studies have investigated optimal 4D phase selection for radiomics. We aim to determine how phase data should be used to identify prognostic biomarkers for distant failure, and test whether stability assessment is required. A phase selection approach will be developed to aid studies with different 4D protocols and account for patient differences. Methods. 186 features were extracted from the tumour and peritumour on all phases for 258 patients. Feature values were selected from phase features using four methods: (A) mean across phases, (B) median across phases, (C) 50% phase, and (D) the most stable phase (closest in value to two neighbours), coined personalised selection. Four levels of stability assessment were also analysed, with inclusion of: (1) all features, (2) stable features across all phases, (3) stable features across phase and neighbour phases, and (4) features averaged over neighbour phases. Clinical-radiomics models were built for twelve combinations of feature type and assessment method. Model performance was assessed by concordance index (c-index) and fraction of new information from radiomic features. Results. The most stable phase spanned the whole range but was most often near exhale. All radiomic signatures provided new information for distant failure prediction. The personalised model had the highest c-index (0.77), and 58% of new information was provided by radiomic features when no stability assessment was performed. Conclusion. The most stable phase varies per-patient and selecting this improves model performance compared to standard methods. We advise the single most stable phase should be determined by minimising feature differences to neighbour phases. Stability assessment over all phases decreases performance by excessively removing features. Instead, averaging of neighbour phases should be used when stability is of concern. The models suggest that higher peritumoural intensity predicts distant failure.

Magnetic resonance radiomics signatures for predicting poorly differentiated hepatocellular carcinoma: A SQUIRE-compliant study

Radiomics contributes to the extraction of undetectable features with the naked eye from high-throughput quantitative images. In this study, 2 predictive models were constructed, which allowed recognition of poorly differentiated hepatocellular carcinoma (HCC). In addition, the effectiveness of the as-constructed signature was investigated in HCC patients.A retrospective study involving 188 patients (age, 29-85 years) enrolled from November 2010 to April 2018 was carried out. All patients were divided randomly into 2 cohorts, namely, the training cohort (n = 141) and the validation cohort (n = 47). The MRI images (DICOM) were collected from PACS before ablation; in addition, the radiomics features were extracted from the 3D tumor area on T1-weighted imaging (T1WI) scans, T2-weighted imaging (T2WI) scans, arterial images, portal images and delayed phase images. In total, 200 radiomics features were extracted. t test and Mann-Whitney U test were performed to exclude some radiomics signatures. Afterwards, a radiomics signature model was built through LASSO regression by RStudio Software. We constructed 2 support vector machine (SVM)-based models: 1 with a radiomics signature only (model 1) and 1 that integrated clinical and radiomics signatures (model 2). Then, the diagnostic performance of the radiomics signature was evaluated through receiver operating characteristic (ROC) analysis.The classification accuracy in the training and validation cohorts was 80.9% and 72.3%, respectively, for model 1. In the training cohort, the area under the ROC curve (AUC) was 0.623, while it was 0.576 in the validation cohort. The classification accuracy in the training and validation cohorts were 79.4% and 74.5%, respectively, for model 2. In the training cohort, the AUC was 0.721, while it was 0.681 in the validation cohort.The MRI-based radiomics signature and clinical model can distinguish HCC patients that belong in a low differentiation group from other patients, which helps in the performance of personal medical protocols.

The impact of the variation of imaging parameters on the robustness of Computed Tomography radiomic features: A review

The field of radiomics is at the forefront of personalized medicine. However, there is concern that high variation in imaging parameters will impact robustness of radiomic features and subsequently the performance of the predictive models built upon them. Therefore, our review aims to evaluate the impact of imaging parameters on the robustness of radiomic features. We also provide insights into the validity and discrepancy of different methodologies applied to investigate the robustness of radiomic features. We selected 47 papers based on our predefined inclusion criteria and grouped these papers by the imaging parameter under investigation: (i) scanner parameters, (ii) acquisition parameters and (iii) reconstruction parameters. Our review highlighted that most of the imaging parameters are disruptive parameters, and shape along with First order statistics were reported as the most robust radiomic features against variation in imaging parameters. This review identified inconsistencies related to the methodology of the reviewed studies such as the metrics used for robustness, the feature extraction techniques, the reporting style, and their outcome inclusion. We hope this review will aid the scientific community in conducting research in a way that is more reproducible and avoids the pitfalls of previous analyses.

Understanding Sources of Variation to Improve the Reproducibility of Radiomics

Radiomics is the method of choice for investigating the association between cancer imaging phenotype, cancer genotype and clinical outcome prediction in the era of precision medicine. The fast dispersal of this new methodology has benefited from the existing advances of the core technologies involved in radiomics workflow: image acquisition, tumor segmentation, feature extraction and machine learning. However, despite the rapidly increasing body of publications, there is no real clinical use of a developed radiomics signature so far. Reasons are multifaceted. One of the major challenges is the lack of reproducibility and generalizability of the reported radiomics signatures (features and models). Sources of variation exist in each step of the workflow; some are controllable or can be controlled to certain degrees, while others are uncontrollable or even unknown. Insufficient transparency in reporting radiomics studies further prevents translation of the developed radiomics signatures from the bench to the bedside. This review article first addresses sources of variation, which is illustrated using demonstrative examples. Then, it reviews a number of published studies and progresses made to date in the investigation and improvement of feature reproducibility and model performance. Lastly, it discusses potential strategies and practical considerations to reduce feature variability and improve the quality of radiomics study. This review focuses on CT image acquisition, tumor segmentation, quantitative feature extraction, and the disease of lung cancer.

Identification of robust and reproducible CT-texture metrics using a customized 3D-printed texture phantom

OBJECTIVE

The objective of this study was to evaluate the robustness and reproducibility of computed tomography-based texture analysis (CTTA) metrics extracted from CT images of a customized texture phantom built for assessing the association of texture metrics to three-dimensional (3D) printed progressively increasing textural heterogeneity.

MATERIALS AND METHODS

A custom-built 3D-printed texture phantom comprising of six texture patterns was used to evaluate the robustness and reproducibility of a radiomics panel under a variety of routine abdominal imaging protocols. The phantom was scanned on four CT scanners (Philips, Canon, GE, and Siemens) to assess reproducibility. The robustness assessment was conducted by imaging the texture phantom across different CT imaging parameters such as slice thickness, field of view (FOV), tube voltage, and tube current for each scanner. The texture panel comprised of 387 features belonging to 15 subgroups of texture extraction methods (e.g., Gray-level Co-occurrence Matrix: GLCM). Twelve unique image settings were tested on all the four scanners (e.g., FOV125). Interclass correlation two-way mixed with absolute agreement (ICC3) was used to assess the robustness and reproducibility of radiomic features. Linear regression was used to test the association between change in radiomic features and increased texture heterogeneity. Results were summarized in heat maps.

RESULTS

A total of 5612 (23.2%) of 24 090 features showed excellent robustness and reproducibility (ICC ≥ 0.9). Intensity, GLCM 3D, and gray-level run length matrix (GLRLM) 3D features showed best performance. Among imaging variables, changes in slice thickness affected all metrics more intensely compared to other imaging variables in reducing the ICC3. From the analysis of linear trend effect of the CTTA metrics, the top three metrics with high linear correlations across all scanners and scanning settings were from the GLRLM 2D/3D and discrete cosine transform (DCT) texture family.

CONCLUSION

The choice of scanner and imaging protocols affect texture metrics. Furthermore, not all CTTA metrics have a linear association with linearly varying texture patterns.

The impact of respiratory motion and CT pitch on the robustness of radiomics feature extraction in 4DCT lung imaging

PURPOSE/OBJECTIVE(S)

The precise radiomics analysis on thoracic 4DCT data is easily compromised by the respiratory motion and CT scan parameter setting, thus leading to the risk of overfitting and/or misinterpretation of data in AI-enabled therapeutic model building. In this study, we investigated the impact of respiratory amplitudes, frequencies and CT scan pitch settings within the thoracic 4DCT scan on robust radiomics feature selection.

MATERIALS/METHODS

A Three-dimensional QUSAR^TM lung tumor phantom was used to simulate different respiratory amplitudes and frequencies along with different CT scan pitch settings. A total of 43 tumor respiratory patterns extracted from 43 patients with non-small cell lung cancer were used to drive the QUSAR^TM lung tumor phantom to mimic the human tumor motion. The 4DCT images of the QUSAR^TM lung tumor phantom with different respiratory patterns and different CT scan pitch setups were acquired for radiomics feature extraction. A static high-quality CT images of the phantom acquired were also used as a reference for radiomics feature extraction. The range of respiratory amplitudes was mimicked at 3mm at left and right (LR) and anterior and posterior (AP) directions and 3mm - 15 mm at the superior and inferior (SI) direction with an interval of 2 mm. The respiratory frequencies were set at 10, 11, 12, 13, 14, 15 and 20 beats per minute (BPMs), respectively. The CT scan pitches were set at 0.025, 0.048, 0.071, 0.93, 0.108, 0.14, 0.16, 0.18, 0.21, 0.23, and 0.25, respectively, which was based on a procedure described in Med. Phys. 30(1):88-97. The pairwise Concordance Correlation Coefficient (CCC) was used to determine the robustness of radiomics feature extraction via comparing the agreement in feature values between 1766 radiomics features extracted from each image acquired under different combinations of respiratory amplitudes and frequencies and CT scan pitches of 4DCT and those extracted from the static CT images.

RESULTS

(1) When the respiratory amplitudes were at 3, 5, 7, 9, 12 and 15mm in the SI direction, the maximum CCC index could be achieved at the reconstructed 4DCT phase images of 60%, 70%, 30%, 20%, 60%~70% and 10%, respectively. Under these six amplitudes, the maximum intensity projection (MIP) and average intensity projection (AIP) images reconstructed show mean CCC values of 0.778 and 0.609, respectively, in pairwise radiomics feature extraction comparison between 4DCT and static CT. (2) When the respiratory amplitude was set at 12 mm in the SI direction, the maximum CCC index could be consistently achieved at the reconstructed 4DCT phase of 90% for the seven respiratory frequencies of 10, 11, 12, 13, 14, 15 and 20 BPMs, respectively. Under these respiratory states, the MIP and AIP images reconstructed show mean CCC values of 0.702 and 0.562, respectively. (3) When the respiratory amplitude was set at 12 mm and the respiratory frequency was set at 13 BPM, the maximum CCC index could be obtained at the reconstructed 4DCT phase of 90% for all scan pitches used except the 0% phase which was obtained at the pitch setting of 0.048. Under these CT scan pitch settings, the MIP and AIP images reconstructed show mean CCC values of 0.558 and 0.782, respectively. (4) The total number of robust features were 50, 34 and 35 with different respiratory amplitudes and phases and CT scanning pitch used (CCC values ≥ 0.99).

CONCLUSION

In 4DCT, the respiratory amplitude, frequency and CT scan pitch are three limiting factors that greatly affect the robustness of radiomics feature extraction. The reconstructed 4DCT phases with better robustness along with suitable respiratory amplitude, frequency and CT scan pitch determined could be used to guide the breathing training for patients with lung cancer for radiation therapy to improve the robust radiomics feature extraction process.

Impact of quantization algorithm and number of gray level intensities on variability and repeatability of low field strength magnetic resonance image-based radiomics texture features

PURPOSE

The purpose of this work was to investigate the impact of quantization preprocessing parameter selection on variability and repeatability of texture features derived from low field strength magnetic resonance (MR) images.

METHODS

Texture features were extracted from low field strength images of a daily image QA phantom with four texture inserts. Feature variability over time was quantified using all combinations of three quantization algorithms and four different numbers of gray level intensities. In addition, texture features were extracted using the same combinations from the low field strength MR images of the gross tumor volume (GTV) and left kidney of patients with repeated set up scans. The impact of region of interest (ROI) preprocessing on repeatability was investigated with a test-retest study design.

RESULTS

The phantom ROIs quantized to 64 Gy level intensities using the histogram equalization method resulted in the greatest number of features with the least variability. There was no clear method that resulted in the highest repeatability in the GTV or left kidney. However, eight texture features extracted from the GTV were repeatable regardless of ROI processing combination.

CONCLUSION

Low field strength MR images can provide a stable basis for texture analysis with ROIs quantized to 64 Gy levels using histogram equalization, but there is no clear optimal combination for repeatability.

Performance of radiomics models for survival prediction in non-small-cell lung cancer: influence of CT slice thickness

OBJECTIVES

To investigate whether CT slice thickness influences the performance of radiomics prognostic models in non-small-cell lung cancer (NSCLC) patients.

METHODS

CT images including 1-, 3-, and 5-mm slice thicknesses acquired from 311 patients who underwent surgical resection for NSCLC between May 2014 and December 2015 were evaluated. Tumor segmentation was performed on CT of each slice thickness and total 94 radiomics features (shape, tumor intensity, and texture) were extracted. The study population was temporally split into development (n = 185) and validation sets (n = 126) for prediction of disease-free survival (DFS). Three radiomics models were built from three different slice thickness datasets (Rad-1, Rad-3, and Rad-5), respectively. Model performance was assessed and compared in three slice thickness datasets and mixed slice thickness dataset using C-indices.

RESULTS

In corresponding slice thickness datasets, the C-indices of Rad-1, Rad-3, and Rad-5 for prediction of DFS were 0.68, 0.70, and 0.68 in the development set, and 0.73, 0.73, and 0.76 in the validation set (p = 0.40-0.89 and 0.27-0.90, respectively). Performance of the models was not significantly changed when they were applied to different slice thicknesses data in the validation set (C-index, 0.73-0.76, 0.72-0.73, 0.75-0.76; p = 0.07-0.92). In the mixed slice thickness dataset, performances of the models were similar to or slightly lower than their performances in the corresponding slice thickness datasets (C-index, 0.72-0.75 vs. 0.73-0.76) in the validation set.

CONCLUSIONS

The performance of radiomics models for predicting DFS in NSCLC patients was not significantly affected by CT slice thickness.

KEY POINTS

• Three radiomics models based on 1-, 3-, and 5-mm CT datasets showed C-indices for predicting disease-free survival of 0.68-0.70 in the development set and 0.73-0.76 in the validation set, without statistical differences (p = 0.27-0.90). • Application of the radiomics models to different slice thickness datasets showed no significant differences in performance between the values in the prediction of disease-free survival (p = 0.07-0.99). • Three radiomics models based on 1-, 3-, and 5-mm CT datasets performed well in mixed slice thickness datasets, showing similar or slightly lower performances.

Radiomics in medical imaging-"how-to" guide and critical reflection

Radiomics is a quantitative approach to medical imaging, which aims at enhancing the existing data available to clinicians by means of advanced mathematical analysis. Through mathematical extraction of the spatial distribution of signal intensities and pixel interrelationships, radiomics quantifies textural information by using analysis methods from the field of artificial intelligence. Various studies from different fields in imaging have been published so far, highlighting the potential of radiomics to enhance clinical decision-making. However, the field faces several important challenges, which are mainly caused by the various technical factors influencing the extracted radiomic features.The aim of the present review is twofold: first, we present the typical workflow of a radiomics analysis and deliver a practical "how-to" guide for a typical radiomics analysis. Second, we discuss the current limitations of radiomics, suggest potential improvements, and summarize relevant literature on the subject.

[Formula: see text]: deep learning-based radiomics for the time-to-event outcome prediction in lung cancer

Hand-crafted radiomics has been used for developing models in order to predict time-to-event clinical outcomes in patients with lung cancer. Hand-crafted features, however, are pre-defined and extracted without taking the desired target into account. Furthermore, accurate segmentation of the tumor is required for development of a reliable predictive model, which may be objective and a time-consuming task. To address these drawbacks, we propose a deep learning-based radiomics model for the time-to-event outcome prediction, referred to as DRTOP that takes raw images as inputs, and calculates the image-based risk of death or recurrence, for each patient. Our experiments on an in-house dataset of 132 lung cancer patients show that the obtained image-based risks are significant predictors of the time-to-event outcomes. Computed Tomography (CT)-based features are predictors of the overall survival (OS), with the hazard ratio (HR) of 1.35, distant control (DC), with HR of 1.06, and local control (LC), with HR of 2.66. The Positron Emission Tomography (PET)-based features are predictors of OS and recurrence free survival (RFS), with hazard ratios of 1.67 and 1.18, respectively. The concordance indices of [Formula: see text], [Formula: see text], and [Formula: see text] for predicting the OS, DC, and RFS show that the deep learning-based radiomics model is as accurate or better in predicting predefined clinical outcomes compared to hand-crafted radiomics, with concordance indices of [Formula: see text], [Formula: see text], and [Formula: see text], for predicting the OS, DC, and RFS, respectively. Deep learning-based radiomics has the potential to offer complimentary predictive information in the personalized management of lung cancer patients.

Robustness of CT radiomic features against image discretization and interpolation in characterizing pancreatic neuroendocrine neoplasms

PURPOSE

To explore the variation of the discriminative power of CT radiomic features (RF) against image discretization/interpolation in characterizing pancreatic neuro-endocrine (PanNEN) neoplasms.

MATERIALS AND METHODS

Thirty-nine PanNEN patients with pre-surgical high contrast CT available were considered. Image interpolation and discretization parameters were intentionally changed, including pixel size (0.73-2.19 mm²), slice thickness (2-5 mm) and binning (32-128 grey levels) and their combination generated 27 parameter's set. The ability of 69 RF in discriminating post-surgically assessed tumor grade (>G1), positive nodes, metastases and vascular invasion was tested: AUC changes when changing the parameters were quantified for selected RF, significantly associated to each end-point. The analysis was repeated for the corresponding images with contrast medium and in a sub-group of 29/39 patients scanned on a single scanner.

RESULTS

The median tumor volume was 1.57 cm³ (16%-84% percentiles: 0.62-34.58 cm³). RF variability against discretization/interpolation parameters was large: only 21/69 RF showed %COV < 20%. Despite this variability, AUC changes were limited for all end-points: with typical AUC values around 0.75-0.85, AUC ranges for the 27 parameter's set were on average 0.062 (1SD:0.037) for all end-points with maximum %COV equal to 5.5% (mean:2.3%). Performances significantly improved when excluding the 5 mm thickness case and fixing the binning to 64 (mean AUC range: 0.036, 1SD:0.019). Using contrast images or limiting the population to single-scanner patients had limited impact on AUC variability.

CONCLUSIONS

The discriminative power of CT RF for panNEN is relatively invariant against image interpolation/discretization within a large range of voxel sizes and binning.

Radiomics as a personalized medicine tool in lung cancer: Separating the hope from the hype

Radiomics has become a popular image analysis method in the last few years. Its key hypothesis is that medical images harbor biological, prognostic and predictive information that is not revealed upon visual inspection. In contrast to previous work with a priori defined imaging biomarkers, radiomics instead calculates image features at scale and uses statistical methods to identify those most strongly associated to outcome. This builds on years of research into computer aided diagnosis and pattern recognition. While the potential of radiomics to aid personalized medicine is widely recognized, several technical limitations exist which hinder biomarker translation. Aspects of the radiomic workflow lack repeatability or reproducibility under particular circumstances, which is a key requirement for the translation of imaging biomarkers into clinical practice. One of the most commonly studied uses of radiomics is for personalized medicine applications in Non-Small Cell Lung Cancer (NSCLC). In this review, we summarize reported methodological limitations in CT based radiomic analyses together with suggested solutions. We then evaluate the current NSCLC radiomics literature to assess the risk associated with accepting the published conclusions with respect to these limitations. We review different complementary scoring systems and initiatives that can be used to critically appraise data from radiomics studies. Wider awareness should improve the quality of ongoing and future radiomics studies and advance their potential as clinically relevant biomarkers for personalized medicine in patients with NSCLC.

Radiomics: from qualitative to quantitative imaging

Historically, medical imaging has been a qualitative or semi-quantitative modality. It is difficult to quantify what can be seen in an image, and to turn it into valuable predictive outcomes. As a result of advances in both computational hardware and machine learning algorithms, computers are making great strides in obtaining quantitative information from imaging and correlating it with outcomes. Radiomics, in its two forms "handcrafted and deep," is an emerging field that translates medical images into quantitative data to yield biological information and enable radiologic phenotypic profiling for diagnosis, theragnosis, decision support, and monitoring. Handcrafted radiomics is a multistage process in which features based on shape, pixel intensities, and texture are extracted from radiographs. Within this review, we describe the steps: starting with quantitative imaging data, how it can be extracted, how to correlate it with clinical and biological outcomes, resulting in models that can be used to make predictions, such as survival, or for detection and classification used in diagnostics. The application of deep learning, the second arm of radiomics, and its place in the radiomics workflow is discussed, along with its advantages and disadvantages. To better illustrate the technologies being used, we provide real-world clinical applications of radiomics in oncology, showcasing research on the applications of radiomics, as well as covering its limitations and its future direction.

Robust, independent and relevant prognostic 18F-fluorodeoxyglucose positron emission tomography radiomics features in non-small cell lung cancer: Are there any?

In locally advanced lung cancer, established baseline clinical variables show limited prognostic accuracy and ¹⁸F-fluorodeoxyglucose positron emission tomography (FDG PET) radiomics features may increase accuracy for optimal treatment selection. Their robustness and added value relative to current clinical factors are unknown. Hence, we identify robust and independent PET radiomics features that may have complementary value in predicting survival endpoints. A 4D PET dataset (n = 70) was used for assessing the repeatability (Bland-Altman analysis) and independence of PET radiomics features (Spearman rank: |ρ|<0.5). Two 3D PET datasets combined (n = 252) were used for training and validation of an elastic net regularized generalized logistic regression model (GLM) based on a selection of clinical and robust independent PET radiomics features (GLM_all). The fitted model performance was externally validated (n = 40). The performance of GLM_all (measured with area under the receiver operating characteristic curve, AUC) was highest in predicting 2-year overall survival (0.66±0.07). No significant improvement was observed for GLM_all compared to a model containing only PET radiomics features or only clinical variables for any clinical endpoint. External validation of GLM_all led to AUC values no higher than 0.55 for any clinical endpoint. In this study, robust independent FDG PET radiomics features did not have complementary value in predicting survival endpoints in lung cancer patients. Improving risk stratification and clinical decision making based on clinical variables and PET radiomics features has still been proven difficult in locally advanced lung cancer patients.

The Technome - A Predictive Internal Calibration Approach for Quantitative Imaging Biomarker Research

The goal of radiomics is to convert medical images into a minable data space by extraction of quantitative imaging features for clinically relevant analyses, e.g. survival time prediction of a patient. One problem of radiomics from computed tomography is the impact of technical variation such as reconstruction kernel variation within a study. Additionally, what is often neglected is the impact of inter-patient technical variation, resulting from patient characteristics, even when scan and reconstruction parameters are constant. In our approach, measurements within 3D regions-of-interests (ROI) are calibrated by further ROIs such as air, adipose tissue, liver, etc. that are used as control regions (CR). Our goal is to derive general rules for an automated internal calibration that enhance prediction, based on the analysed features and a set of CRs. We define qualification criteria motivated by status-quo radiomics stability analysis techniques to only collect information from the CRs which is relevant given a respective task. These criteria are used in an optimisation to automatically derive a suitable internal calibration for prediction tasks based on the CRs. Our calibration enhanced the performance for centrilobular emphysema prediction in a COPD study and prediction of patients' one-year-survival in an oncological study.

Reproducibility of CT Radiomic Features within the Same Patient: Influence of Radiation Dose and CT Reconstruction Settings

Background Results of recent phantom studies show that variation in CT acquisition parameters and reconstruction techniques may make radiomic features largely nonreproduceable and of limited use for prognostic clinical studies. Purpose To investigate the effect of CT radiation dose and reconstruction settings on the reproducibility of radiomic features, as well as to identify correction factors for mitigating these sources of variability. Materials and Methods This was a secondary analysis of a prospective study of metastatic liver lesions in patients who underwent staging with single-energy dual-source contrast material-enhanced staging CT between September 2011 and April 2012. Technique parameters were altered, resulting in 28 CT data sets per patient that included different dose levels, section thicknesses, kernels, and reconstruction algorithm settings. By using a training data set (n = 76), reproducible intensity, shape, and texture radiomic features (reproducibility threshold, R2 ≥ 0.95) were selected and correction factors were calculated by using a linear model to convert each radiomic feature to its estimated value in a reference technique. By using a test data set (n = 75), the reproducibility of hierarchical clustering based on 106 radiomic features measured with different CT techniques was assessed. Results Data in 78 patients (mean age, 60 years ± 10; 33 women) with 151 liver lesions were included. The percentage of radiomic features deemed reproducible for any variation of the different technical parameters was 11% (12 of 106). Of all technical parameters, reconstructed section thickness had the largest impact on the reproducibility of radiomic features (12.3% [13 of 106]) if only one technical parameter was changed while all other technical parameters were kept constant. The results of the hierarchical cluster analysis showed improved clustering reproducibility when reproducible radiomic features with dedicated correction factors were used (ρ = 0.39-0.71 vs ρ = 0.14-0.47). Conclusion Most radiomic features are highly affected by CT acquisition and reconstruction settings, to the point of being nonreproducible. Selecting reproducible radiomic features along with study-specific correction factors offers improved clustering reproducibility. © RSNA, 2019 Online supplemental material is available for this article. See also the editorial by Sosna in this issue.

A radiomic approach to predicting nodal relapse and disease-specific survival in patients treated with stereotactic body radiation therapy for early-stage non-small cell lung cancer

PURPOSE

To describe the possibility of building a classifier for patients at risk of lymph node relapse and a predictive model for disease-specific survival in patients with early stage non-small cell lung cancer.

METHODS

A cohort of 102 patients who received stereotactic body radiation treatment was retrospectively investigated. A set of 45 textural features was computed for the tumor volumes on the treatment planning CT images. Patients were split into two independent cohorts (70 patients, 68.9%, for training; and 32 patients, 31.4%, for validation). Three different models were built in the study. A stepwise backward linear discriminant analysis was applied to identify patients at risk of lymph node progression. The performance of the model was assessed by means of standard metrics derived from the confusion matrix. Furthermore, all textural features were correlated to survival data to build two separate predictive models for progression-free survival (PFS) and disease-specific survival (DS-OS). These models were built from the features/predictors found significant in univariate analysis and elastic net regularization by means of a multivarate Cox regression with backward selection. Low- and high-risk groups were identified by maximizing the separation by means of the Youden method.

RESULTS

In the total cohort (77, 75.5%, males; and 25, 24.5%, females; median age 76.6 years), 15 patients presented nodal progression at the time of analysis; 19 patients (18.6%) died because of disease-specific causes, 25 (24.5%) died from other reasons, 28 (27.5%) were alive without disease, and 30 (29.4%) with either local or distant progression. The specificity, sensitivity, and accuracy of the classifier resulted 83.1 ± 24.5, 87.4 ± 1.2, and 85.4 ± 12.5 in the validation group (coherent with the findings in the training). The area under the curve for the classifier resulted in 0.84 ± 0.04 and 0.73 ± 0.05 for training and validation, respectively. The mean time for DS-OS and PFS for the low- and high-risk subgroups of patients (in the validation groups) were 88.2 month ± 9.0 month vs. 84.1 month ± 7.8 month (low risk) and 52.7 month ± 5.9 month vs. 44.6 month ± 9.2 month (high risk), respectively.

CONCLUSION

Radiomics analysis based on planning CT images allowed a classifier and predictive models capable of identifying patients at risk of nodal relapse and high-risk of bad prognosis to be built. The radiomics signatures identified were mostly related to tumor heterogeneity.

Assessing radiomic feature robustness to interpolation in 18F-FDG PET imaging

Radiomic studies link quantitative imaging features to patient outcomes in an effort to personalise treatment in oncology. To be clinically useful, a radiomic feature must be robust to image processing steps, which has made robustness testing a necessity for many technical aspects of feature extraction. We assessed the stability of radiomic features to interpolation processing and categorised features based on stable, systematic, or unstable responses. Here, ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET images for 441 oesophageal cancer patients (split: testing = 353, validation = 88) were resampled to 6 isotropic voxel sizes (1.5 mm, 1.8 mm, 2.0 mm, 2.2 mm, 2.5 mm, 2.7 mm) and 141 features were extracted from each volume of interest (VOI). Features were categorised into four groups with two statistical tests. Feature reliability was analysed using an intraclass correlation coefficient (ICC) and patient ranking consistency was assessed using a Spearman’s rank correlation coefficient (ρ). We categorised 93 features robust and 6 limited robustness (stable responses), 34 potentially correctable (systematic responses), and 8 not robust (unstable responses). We developed a correction technique for features with potential systematic variation that used surface fits to link voxel size and percentage change in feature value. Twenty-nine potentially correctable features were re-categorised to robust for the validation dataset, after applying corrections defined by surface fits generated on the testing dataset. Furthermore, we found the choice of interpolation algorithm alone (spline vs trilinear) resulted in large variation in values for a number of features but the response categorisations remained constant. This study attempted to quantify the diverse response of radiomics features commonly found in ¹⁸F-FDG PET clinical modelling to isotropic voxel size interpolation.

Radiomic feature stability across 4D respiratory phases and its impact on lung tumor prognosis prediction

Radiomic analysis has recently demonstrated versatile uses in improving diagnostic and prognostic prediction accuracy for lung cancer. However, since lung tumors are subject to substantial motion due to respiration, the stability of radiomic features over the respiratory cycle of the patient needs to be investigated to better evaluate the robustness of the inter-patient feature variability for clinical applications, and its impact in such applications needs to be assessed. A full panel of 841 radiomic features, including tumor intensity, shape, texture, and wavelet features, were extracted from individual phases of a four-dimensional (4D) computed tomography on 20 early-stage non-small-cell lung cancer (NSCLC) patients. The stability of each radiomic feature was assessed across different phase images of the same patient using the coefficient of variation (COV). The relationship between individual COVs and tumor motion magnitude was inspected. Population COVs, the mean COVs of all 20 patients, were used to evaluate feature motion stability and categorize the radiomic features into 4 different groups. The two extremes, the Very Small group (COV≤5%) and the Large group (COV>20%), each accounted for about a quarter of the features. Shape features were the most stable, with COV≤10% for all features. A clinical study was subsequently conducted using 140 early-stage NSCLC patients. Radiomic features were employed to predict the overall survival with a 500-round bootstrapping. Identical multiple regression model development process was applied, and the model performance was compared between models with and without a feature pre-selection step based on 4D COV to pre-exclude unstable features. Among the systematically tested cutoff values, feature pre-selection with 4D COV≤5% achieved the optimal model performance. The resulting 3-feature radiomic model significantly outperformed its counterpart with no 4D COV pre-selection, with P = 2.16x10^-27 in the one-tailed t-test comparing the prediction performances of the two models.

Repeatability of texture features derived from magnetic resonance and computed tomography imaging and use in predictive models for non-small cell lung cancer outcome

PURPOSE

To evaluate the repeatability of MRI and CT derived texture features and to investigate the feasibility of use in predictive single and multi-modality models for radiotherapy of non-small cell lung cancer. Methods: Fifty-nine texture features were extracted from unfiltered and wavelet filtered images. Repeatability of test-retest features from helical 4D CT scans, true fast MRI with steady state precession (TRUFISP), and volumetric interpolation breath-hold examination (VIBE) was determined by the concordance correlation coefficient (CCC). A workflow was developed to predict overall survival at 12, 18, and 24 months and tumour response at end of treatment for tumour features, and normal muscle tissue features as a control. Texture features were reduced to repeatable and stable features before clustering. Cluster representative feature selection was performed by univariate or medoid analysis before model selection. P-values were corrected for false discovery rate. Results: Repeatable (CCC ≥ 0.9) features were found for both tumour and normal muscle tissue: CT: 54.4% for tumour and 78.5% for normal tissue, TRUFISP: 64.4% for tumour and 67.8% for normal tissue, and VIBE: 52.6% for tumour and 72.9% for normal muscle tissue. Muscle tissue control analysis found 7 significant models with 6 of 7 models utilizing the univariate representative feature selection technique. Tumour analysis revealed 12 significant models for overall survival and none for tumour response at end of treatment. The accuracy of significant single modality was about the same for MR and CT. Multi-modality tumour models had comparable performance to single modality models. Conclusion: MR derived texture features may add value to predictive models and should be investigated in a larger cohort. Control analysis demonstrated that the medoid representative feature selection method may result in more robust models. .

Stability of radiomics features in apparent diffusion coefficient maps from a multi-centre test-retest trial

Quantitative radiomics features, extracted from medical images, characterize tumour-phenotypes and have been shown to provide prognostic value in predicting clinical outcomes. Stability of radiomics features extracted from apparent diffusion coefficient (ADC)-maps is essential for reliable correlation with the underlying pathology and its clinical applications. Within a multicentre, multi-vendor trial we established a method to analyse radiomics features from ADC-maps of ovarian (n = 12), lung (n = 19), and colorectal liver metastasis (n = 30) cancer patients who underwent repeated (<7 days) diffusion-weighted imaging at 1.5 T and 3 T. From these ADC-maps, 1322 features describing tumour shape, texture and intensity were retrospectively extracted and stable features were selected using the concordance correlation coefficient (CCC > 0.85). Although some features were tissue- and/or respiratory motion-specific, 122 features were stable for all tumour-entities. A large proportion of features were stable across different vendors and field strengths. By extracting stable phenotypic features, fitting-dimensionality is reduced and reliable prognostic models can be created, paving the way for clinical implementation of ADC-based radiomics.

Radiomic analysis of planning computed tomograms for predicting radiation-induced lung injury and outcome in lung cancer patients treated with robotic stereotactic body radiation therapy

OBJECTIVES

To predict radiation-induced lung injury and outcome in non-small cell lung cancer (NSCLC) patients treated with robotic stereotactic body radiation therapy (SBRT) from radiomic features of the primary tumor.

METHODS

In all, 110 patients with primary stage I/IIa NSCLC were analyzed for local control (LC), disease-free survival (DFS), overall survival (OS) and development of local lung injury up to fibrosis (LF). First-order (histogram), second-order (GLCM, Gray Level Co-occurrence Matrix) and shape-related radiomic features were determined from the unprocessed or filtered planning CT images of the gross tumor volume (GTV), subjected to LASSO (Least Absolute Shrinkage and Selection Operator) regularization and used to construct continuous and dichotomous risk scores for each endpoint.

RESULTS

Continuous scores comprising 1-5 histogram or GLCM features had a significant (p = 0.0001-0.032) impact on all endpoints that was preserved in a multifactorial Cox regression analysis comprising additional clinical and dosimetric factors. At 36 months, LC did not differ between the dichotomous risk groups (93% vs. 85%, HR 0.892, 95%CI 0.222-3.590), while DFS (45% vs. 17%, p < 0.05, HR 0.457, 95%CI 0.240-0.868) and OS (80% vs. 37%, p < 0.001, HR 0.190, 95%CI 0.065-0.556) were significantly lower in the high-risk groups. Also, the frequency of LF differed significantly between the two risk groups (63% vs. 20% at 24 months, p < 0.001, HR 0.158, 95%CI 0.054-0.458).

CONCLUSION

Radiomic analysis of the gross tumor volume may help to predict DFS and OS and the development of local lung fibrosis in early stage NSCLC patients treated with stereotactic radiotherapy.

Gray-level discretization impacts reproducible MRI radiomics texture features

OBJECTIVES

To assess the influence of gray-level discretization on inter- and intra-observer reproducibility of texture radiomics features on clinical MR images.

MATERIALS AND METHODS

We studied two independent MRI datasets of 74 lacrymal gland tumors and 30 breast lesions from two different centers. Two pairs of readers performed three two-dimensional delineations for each dataset. Texture features were extracted using two radiomics softwares (Pyradiomics and an in-house software). Reproducible features were selected using a combination of intra-class correlation coefficient (ICC) and concordance and coherence coefficient (CCC) with 0.8 and 0.9 as thresholds, respectively. We tested six absolute and eight relative gray-level discretization methods and analyzed the distribution and highest number of reproducible features obtained for each discretization. We also analyzed the number of reproducible features extracted from computer simulated delineations representative of inter-observer variability.

RESULTS

The gray-level discretization method had a direct impact on texture feature reproducibility, independent of observers, software or method of delineation (simulated vs. human). The absolute discretization consistently provided statistically significantly more reproducible features than the relative discretization. Varying the bin number of relative discretization led to statistically significantly more variable results than varying the bin size of absolute discretization.

CONCLUSIONS

When considering inter-observer reproducible results of MRI texture radiomics features, an absolute discretization should be favored to allow the extraction of the highest number of potential candidates for new imaging biomarkers. Whichever the chosen method, it should be systematically documented to allow replicability of results.

Interchangeability of radiomic features between [18F]-FDG PET/CT and [18F]-FDG PET/MR

PURPOSE

Radiomics is a promising tool for identification of new prognostic biomarkers. However, image reconstruction settings and test-retest variability may influence the absolute values of radiomic features. Unstable radiomic features cannot be used as reliable biomarkers. PET/MR is becoming increasingly available and often replaces PET/CT for different indications. The aim of this study was to quantify to what extend [18F]-FDG PET/CT radiomics models can be transferred to [18F]-FDG PET/MR and thereby to investigate the feasibility of combined PET/CT-PET/MR models. For this purpose, we compared PET radiomic features calculated on PET/MR and PET/CT and on a 4D-gated PET/MR dataset to select radiomic features that are robust to attenuation correction differences and test-retest variability, respectively.

METHODS

Two cohorts of patients with lung lesions were studied. In the first cohort (n = 10), inhale and exhale phases of a 4D [18F]-FDG PET/MR (4DPETMR) scan were used as a surrogate for a test-retest dataset. In the second cohort (n = 9), patients underwent first an [18F]-FDG PET/MR scan (SIGNA PET/MR, GE Healthcare, Waukesha) followed by an [18F]-FDG PET/CT scan (Discovery 690, GE Healthcare) with a delay of 33 ± 5 min (PETCT-PETMR). Lesions were segmented on inhale and exhale 4D-PET phases and on the individual PET scans from PET/CT and PET/MR with two semi-automated methods (gradient-based and threshold-based). The scan resolution was 2.73 × 2.73 × 3.27 mm and 2.34 × 2.34 × 2.78 mm for the PET/CT and PET/MR, respectively. In total, 1355 radiomic features were calculated, i.e., shape (n = 18), intensity (n = 17), texture (n = 136), and wavelet (n = 1184). The intraclass correlation coefficient (ICC) was calculated to compare the radiomic features of the 4DPETMR (ICC(1,1)) and PETCT-PETMR (ICC(3,1)) datasets. An ICC > 0.9 was considered stable among both types of PET scans.

RESULTS AND CONCLUSION

The 4DPETMR showed highest stability for shape, intensity, and texture (>80%) and lower stability for wavelet features (40%). Gradient-based method showed higher stability compared to threshold-based method except from shape features. In PETCT-PETMR, more than 61% of shape and intensity features were stable for both segmentation methods. However, a reduced stability was observed for texture (50%) and wavelet (<30%) features. More wavelet features were robust in the smoothed images (low-pass filtering) compared to images with emphasized heterogeneity (high-pass filtering). Comparing stable features of both investigations, highest agreement was found for intensity and lower agreement for shape, texture, and wavelet features. Only 53.6% of stable texture features in 4DPETMR were also stable in PETCT-PETMR, and even less in case of wavelet features (40.4%). Approximately 16.9% (texture) and 43.2% (wavelet) of stable PETCT-PETMR features are unstable in 4DPETMR. To conclude, shape and intensity features were robust when comparing two types of [18F]-FDG PET scans (PET/CT and PET/MR). Reduced stability was observed for texture and wavelet features. We identified multiple origins of instability of radiomic features, such as attenuation correction differences, different uptake times, and spatial resolution. This needs to be considered when models based on PET/CT are transferred PET/MR models or when combined models are used.

Investigation of thoracic four-dimensional CT-based dimension reduction technique for extracting the robust radiomic features

Robust feature selection in radiomic analysis is often implemented using the RIDER test-retest datasets. However, the CT Protocol between the facility and test-retest datasets are different. Therefore, we investigated possibility to select robust features using thoracic four-dimensional CT (4D-CT) scans that are available from patients receiving radiation therapy. In 4D-CT datasets of 14 lung cancer patients who underwent stereotactic body radiotherapy (SBRT) and 14 test-retest datasets of non-small cell lung cancer (NSCLC), 1170 radiomic features (shape: n = 16, statistics: n = 32, texture: n = 1122) were extracted. A concordance correlation coefficient (CCC) > 0.85 was used to select robust features. We compared the robust features in various 4D-CT group with those in test-retest. The total number of robust features was a range between 846/1170 (72%) and 970/1170 (83%) in all 4D-CT groups with three breathing phases (40%-60%); however, that was a range between 44/1170 (4%) and 476/1170 (41%) in all 4D-CT groups with 10 breathing phases. In test-retest, the total number of robust features was 967/1170 (83%); thus, the number of robust features in 4D-CT was almost equal to that in test-retest by using 40-60% breathing phases. In 4D-CT, respiratory motion is a factor that greatly affects the robustness of features, thus by using only 40-60% breathing phases, excessive dimension reduction will be able to be prevented in any 4D-CT datasets, and select robust features suitable for CT protocol of your own facility.

Combining Multiple Magnetic Resonance Imaging Sequences Provides Independent Reproducible Radiomics Features

To evaluate the relative contribution of different Magnetic Resonance Imaging (MRI) sequences for the extraction of radiomics features in a cohort of patients with lacrimal gland tumors. This prospective study was approved by the Institutional Review Board and signed informed consent was obtained from all participants. From December 2015 to April 2017, 37 patients with lacrimal gland lesions underwent MRI before surgery, including axial T1-WI, axial Diffusion-WI, coronal DIXON-T2-WI and coronal post-contrast DIXON-T1-WI. Two readers manually delineated both lacrimal glands to assess inter-observer reproducibility, and one reader performed two successive delineations to assess intra-observer reproducibility. Radiomics features were extracted using an in-house software to calculate 85 features per region-of-interest (510 features/patient). Reproducible features were defined as features presenting both an intra-class correlation coefficient ≥0.8 and a concordance correlation coefficient ≥0.9 across combinations of the three delineations. Among these features, the ones yielding redundant information were identified as clusters using hierarchical clustering based on the Spearman correlation coefficient. All the MR sequences provided reproducible radiomics features (range 14(16%)−37(44%)) and non-redundant clusters (range 5–14). The highest numbers of features and clusters were provided by the water and in-phase DIXON T2-WI and water and in-phase post-contrast DIXON T1-WI (37, 26, 26 and 26 features and 14,12, 9 and 11 clusters, respectively). A total of 145 reproducible features grouped into 51 independent clusters was provided by pooling all the MR sequences. All MRI sequences provided reproducible radiomics features yielding independent information which could potentially serve as biomarkers.

Feasibility of CT radiomics to predict treatment response of individual liver metastases in esophagogastric cancer patients

In this study we investigate a CT radiomics approach to predict response to chemotherapy of individual liver metastases in patients with esophagogastric cancer (EGC). In eighteen patients with metastatic EGC treated with chemotherapy, all liver metastases were manually delineated in 3D on the pre-treatment and evaluation CT. From the pre-treatment CT scans 370 radiomics features were extracted per lesion. Random forest (RF) models were generated to discriminate partial responding (PR, >65% volume decrease, including 100% volume decrease), and complete remission (CR, only 100% volume decrease) lesions from other lesions. RF-models were build using a leave one out strategy where all lesions of a single patient were removed from the dataset and used as validation set for a model trained on the lesions of the remaining patients. This process was repeated for all patients, resulting in 18 trained models and one validation set for both the PR and CR datasets. Model performance was evaluated by receiver operating characteristics with corresponding area under the curve (AUC). In total 196 liver metastases were delineated on the pre-treatment CT, of which 99 (51%) lesions showed a decrease in size of more than 65% (PR). From the PR set a total of 47 (47% of RL, 24% of initial) lesions were no longer detected in CT scan 2 (CR). The RF-model for PR lesions showed an average training AUC of 0.79 (range: 0.74-0.83) and 0.65 (95% ci: 0.57-0.73) for the combined validation set. The RF-model for CR lesions had an average training AUC of 0.87 (range: 0.83-0.90) and 0.79 (95% ci 0.72-0.87) for the validation set. Our findings show that individual response of liver metastases varies greatly within and between patients. A CT radiomics approach shows potential in discriminating responding from non-responding liver metastases based on the pre-treatment CT scan, although further validation in an independent patient cohort is needed to validate these findings.

Spatial-temporal variability of radiomic features and its effect on the classification of lung cancer histology

The purpose of this research was to study the sensitivity of Computed Tomography (CT) radiomic features to motion blurring and signal-to-noise ratios (SNR), and investigate its downstream effect regarding the classification of non-small cell lung cancer (NSCLC) histology. Forty-three radiomic features were considered and classified into one of four categories: Morphological, Intensity, Fine Texture, and Coarse Texture. First, a series of simulations were used to study feature-sensitivity to changes in spatial-temporal resolution. A dynamic digital phantom was used to generate images with different breathing amplitudes and SNR, from which features were extracted and characterized relative to initial simulation conditions. Stage I NSCLC patients were then retrospectively identified, from which three different acquisition-specific feature-spaces were generated based on free-breathing (FB), average-intensity-projection (AIP), and end-of-exhalation (EOE) CT images. These feature-spaces were derived to cover a wide range of spatial-temporal tradeoff. Normalized percent differences and concordance correlation coefficients (CCC) were used to assess the variability between the 3D and 4D acquisition techniques. Subsequently, three corresponding acquisition-specific logistic regression models were developed to classify lung tumor histology. Classification performance was compared between the different data-dependent models. Simulation results demonstrated strong linear dependences (p  >  0.95) between respiratory motion and morphological features, as well as between SNR and texture features. The feature Short Run Emphasis was found to be particularly stable to both motion blurring and changes in SNR. When comparing FB-to-EOE, 37% of features demonstrated high CCC agreement (CCC  >  0.8), compared to only 30% for FB-to-AIP. In classifying tumor histology, EoE images achieved an average AUC, Accuracy, Sensitivity, and Specificity of [Formula: see text], respectively. FB images achieved respective values of [Formula: see text], and AIP images achieved respective values of [Formula: see text]. Several radiomic features have been identified as being relatively robust to spatial-temporal variations based on both simulation data and patient data. In general, features that were sensitive to motion blurring were not necessarily the same features that were sensitive to changes in SNR. Our modeling results suggest that the EoE phase of a 4DCT acquisition may provide useful radiomic information, particularly for features that are highly sensitive to respiratory motion.

A review on radiomics and the future of theranostics for patient selection in precision medicine

The growing complexity and volume of clinical data and the associated decision-making processes in oncology promote the advent of precision medicine. Precision (or personalised) medicine describes preventive and/or treatment procedures that take individual patient variability into account when proscribing treatment, and has been hindered in the past by the strict requirements of accurate, robust, repeatable and preferably non-invasive biomarkers to stratify both the patient and the disease. In oncology, tumour subtypes are traditionally measured through repeated invasive biopsies, which are taxing for the patient and are cost and labour intensive. Quantitative analysis of routine clinical imaging provides an opportunity to capture tumour heterogeneity non-invasively, cost-effectively and on large scale. In current clinical practice radiological images are qualitatively analysed by expert radiologists whose interpretation is known to suffer from inter- and intra-operator variability. Radiomics, the high-throughput mining of image features from medical images, provides a quantitative and robust method to assess tumour heterogeneity, and radiomics-based signatures provide a powerful tool for precision medicine in cancer treatment. This study aims to provide an overview of the current state of radiomics as a precision medicine decision support tool. We first provide an overview of the requirements and challenges radiomics currently faces in being incorporated as a tool for precision medicine, followed by an outline of radiomics' current applications in the treatment of various types of cancer. We finish with a discussion of possible future advances that can further develop radiomics as a precision medicine tool.