Radiomic and radiogenomic modeling for radiotherapy: strategies, pitfalls, and challenges

Abstract

The power of predictive modeling for radiotherapy outcomes has historically been limited by an inability to adequately capture patient-specific variabilities; however, nextgeneration platforms together with imaging technologies and powerful bioinformatic tools have facilitated strategies and provided optimism. Integrating clinical, biological, imaging, and treatment-specific data for more accurate prediction of tumor control probabilities or risk of radiation-induced side effects are high-dimensional problems whose solutions could have widespread benefits to a diverse patient population-we discuss technical approaches toward this objective. Increasing interest in the above is specifically reflected by the emergence of two nascent fields, which are distinct but complementary: radiogenomics, which broadly seeks to integrate biological risk factors together with treatment and diagnostic information to generate individualized patient risk profiles, and radiomics, which further leverages large-scale imaging correlates and extracted features for the same purpose. We review classical analytical and datadriven approaches for outcomes prediction that serve as antecedents to both radiomic and radiogenomic strategies. Discussion then focuses on uses of conventional and deep machine learning in radiomics. We further consider promising strategies for the harmonization of highdimensional, heterogeneous multiomics datasets (panomics) and techniques for nonparametric validation of best-fit models. Strategies to overcome common pitfalls that are unique to dataintensive radiomics are also discussed. © The Authors. Published by SPIE under a Creative Commons Attribution 4.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI.

Applications of Radiomics and Radiogenomics in High-Grade Gliomas in the Era of Precision Medicine

“…Most of the existing radiogenomics studies have adopted an exploratory analysis approach to investigate the relationships between molecular dynamics and tumor characteristics reflected by specific radiographic phenotypes (radiophenotypes) [ 11 ]. For example, radiophenotypes, including tumor enhancement, nonenhancing tumor, necrosis, infiltrated edema, neo-angiogenesis, microstructural changes, and tumor location, have been associated with genomic profiles of the tumors to provide a better understanding of the underlying tumor biology [ 11 , 46 , 47 , 48 , 49 , 50 , 51 , 52 , 53 ]. Along these lines, a few radiogenomics studies have stratified high-grade glioma patients based on their risk, i.e., into groups of high, intermediate, and low-risk based on radiomic features that were predictive of overall or progression-free survival, and explored associations of these predictive radiomic features with gene expression profiles [ 54 ].…”

Impact of image filtering and assessment of volume-confounding effects on CT radiomic features and derived survival models in non-small cell lung cancer

“…Similar to other big data-based approaches (e.g., genomics, proteomics, metabolomics), radiomics could be used to refine outcome modeling, to assist auto-segmentation tasks and to identify novel predictors of treatment response ( 14 , 15 ). The radiomic workflow is structured into a well-defined pipeline, which generally includes the following steps: image segmentation, preprocessing, features extraction and selection, model construction and validation ( 11 ).…”

Assessment of tissue toxicity risk in breast radiotherapy using Bayesian networks

“…MR algorithms, and BNs specifically, attract less attention in the contemporary literature than ML 35,36 . This lack of attention can largely be attributed to the complexity of implementation, requisite expert guidance in constructing network topology, and higher data availability.…”

“…MR algorithms, and BNs specifically, attract less attention in the contemporary literature than ML. 35,36 This lack of attention can largely be attributed to the complexity of implementation, requisite expert guidance in constructing network topology, and higher data availability. However, BNs have been shown to be very effective in clinical applications due to access to expert guidance and limited data availability.…”