Interpretability of radiomics models is improved when using feature group selection strategies for predicting molecular and clinical targets in clear-cell renal cell carcinoma: insights from the TRACERx Renal study

Cancer Imaging

Research Article

Evan Hann¹  Matthew R. Orton¹  Dow-Mu Koh²  Christina Messiou²  Charlotte E. Spencer³  José I. López⁴  Scott T. C. Shepherd⁵  Samra Turajlic⁵  Joshua Shur⁶  Derfel Ap Dafydd⁶  Simon J. Doran⁷  Francesca Comito⁸  Víctor Albarrán-Artahona⁹  James Larkin^1,10  Hannah Warren^1,11 

[1] Artificial Intelligence Imaging Hub, Royal Marsden NHS Foundation Trust, London, UK;Artificial Intelligence Imaging Hub, Royal Marsden NHS Foundation Trust, London, UK;Division of Radiotherapy and Imaging, Institute of Cancer Research, London, UK;Department of Radiology, Royal Marsden NHS Foundation Trust, London, UK;Cancer Dynamics Laboratory, The Francis Crick Institute, London, UK;Cancer Dynamics Laboratory, The Francis Crick Institute, London, UK;Biomarkers in Cancer Unit, Biocruces-Bizkaia Health Research Institute, Barakaldo, Spain;Cancer Dynamics Laboratory, The Francis Crick Institute, London, UK;Renal and Skin Units, Royal Marsden Hospital NHS Foundation Trust, London, UK;Melanoma and Kidney Cancer Team, Institute of Cancer Research, London, UK;Department of Radiology, Royal Marsden NHS Foundation Trust, London, UK;Division of Radiotherapy and Imaging, Institute of Cancer Research, London, UK;Medical Oncology, IRCCS Azienda Ospedaliero-Universitaria di Bologna, Bologna, Italy;Department of Experimental, Diagnostic and Specialty Medicine (DIMES), University of Bologna, Bologna, Italy;Renal and Skin Units, Royal Marsden Hospital NHS Foundation Trust, London, UK;Medical Oncology Department, Hospital Clinic de Barcelona, Barcelona, Spain;Renal and Skin Units, Royal Marsden Hospital NHS Foundation Trust, London, UK;Melanoma and Kidney Cancer Team, Institute of Cancer Research, London, UK;Urology Centre, Guy’s and St. Thomas’ NHS Foundation Trust, SE1 9RT, London, UK;Division of Surgery and Interventional Science, University College London, London, UK;

关键词: Radiomics;    Radiogenomics;    Histology;    Interpretable;    Machine learning;    Feature selection;    Group selection;    Renal cancer;    Nested validation;    Molecular subtyping;   

DOI  :  10.1186/s40644-023-00594-3

 received in 2023-01-10, accepted in 2023-07-12,  发布年份 2023

来源: Springer

【 摘 要 】

BackgroundThe aim of this work is to evaluate the performance of radiomics predictions for a range of molecular, genomic and clinical targets in patients with clear cell renal cell carcinoma (ccRCC) and demonstrate the impact of novel feature selection strategies and sub-segmentations on model interpretability.MethodsContrast-enhanced CT scans from the first 101 patients recruited to the TRACERx Renal Cancer study (NCT03226886) were used to derive radiomics classification models to predict 20 molecular, histopathology and clinical target variables. Manual 3D segmentation was used in conjunction with automatic sub-segmentation to generate radiomics features from the core, rim, high and low enhancing sub-regions, and the whole tumour. Comparisons were made between two classification model pipelines: a Conventional pipeline reflecting common radiomics practice, and a Proposed pipeline including two novel feature selection steps designed to improve model interpretability. For both pipelines nested cross-validation was used to estimate prediction performance and tune model hyper-parameters, and permutation testing was used to evaluate the statistical significance of the estimated performance measures. Further model robustness assessments were conducted by evaluating model variability across the cross-validation folds.ResultsClassification performance was significant (p < 0.05, H0:AUROC = 0.5) for 11 of 20 targets using either pipeline and for these targets the AUROCs were within ± 0.05 for the two pipelines, except for one target where the Proposed pipeline performance increased by > 0.1. Five of these targets (necrosis on histology, presence of renal vein invasion, overall histological stage, linear evolutionary subtype and loss of 9p21.3 somatic alteration marker) had AUROC > 0.8. Models derived using the Proposed pipeline contained fewer feature groups than the Conventional pipeline, leading to more straightforward model interpretations without loss of performance. Sub-segmentations lead to improved performance and/or improved interpretability when predicting the presence of sarcomatoid differentiation and tumour stage.ConclusionsUse of the Proposed pipeline, which includes the novel feature selection methods, leads to more interpretable models without compromising prediction performance.Trial registrationNCT03226886 (TRACERx Renal)

【 授权许可】

CC BY   
© International Cancer Imaging Society (ICIS) 2023

【 图 表 】

13690_2023_1154_Article_IEq11.gif

Fig. 4

Fig. 1

Fig. 1

Fig. 2

42004_2023_982_Article_IEq18.gif

Fig. 1

Fig. 1