Radiomics and Its Feature Selection: A Review

Abstract

Medical imaging plays an indispensable role in evaluating, predicting, and monitoring a range of medical conditions. Radiomics, a specialized branch of medical imaging, utilizes quantitative features extracted from medical images to describe underlying pathologies, genetic information, and prognostic indicators. The integration of radiomics with artificial intelligence presents innovative avenues for cancer diagnosis, prognosis evaluation, and therapeutic choices. In the context of oncology, radiomics offers significant potential. Feature selection emerges as a pivotal step, enhancing the clinical utility and precision of radiomics. It achieves this by purging superfluous and unrelated features, thereby augmenting model performance and generalizability. The goal of this review is to assess the fundamental radiomics process and the progress of feature selection methods, explore their applications and challenges in cancer research, and provide theoretical and methodological support for future investigations. Through an extensive literature survey, articles pertinent to radiomics and feature selection were garnered, synthesized, and appraised. The paper provides detailed descriptions of how radiomics is applied and challenged in different cancer types and their various stages. The review also offers comparative insights into various feature selection strategies, including filtering, packing, and embedding methodologies. Conclusively, the paper broaches the limitations and prospective trajectories of radiomics.

A distributed feature selection pipeline for survival analysis using radiomics in non-small cell lung cancer patients

“…In their study, feature selection is performed through intra-feature correlation calculation and the application of hierarchical clustering, while in our work we also consider the correlation of features with respect to the survival outcome using the CFS algorithm. We think that the CFS was particularly suited for the dataset due to the correlated nature of some radiomics features 34 . Moreover, we further reduce the feature set by applying LASSO regularization, that is a very well-established feature selection method 34 , to the training of Cox Regression Model to select the best feature set for the model, taking into account also the time dependent nature of the outcome.…”

“…We think that the CFS was particularly suited for the dataset due to the correlated nature of some radiomics features 34 . Moreover, we further reduce the feature set by applying LASSO regularization, that is a very well-established feature selection method 34 , to the training of Cox Regression Model to select the best feature set for the model, taking into account also the time dependent nature of the outcome. LASSO alone would have been impractical to train on a distributed network with such a large dataset, due to both the communication overhead involved in federated learning and the slow convergence of constrained optimization methods in high dimensional settings such as LASSO itself 11 , 35 .…”

Radiomics in oncology: A mini-review of principles, applications and challenges

“…These features can be categorized into different types, such as first-order statistics (e.g., mean, variance), second-order statistics (e.g., texture features), and higher-order statistics (e.g., fractal dimensions). These features capture various aspects of the underlying tissue properties and can provide insights into the tissue microenvironment and disease state [2,8]. With the abundance of extracted features, selecting the most relevant and informative ones for further analysis is crucial.…”

“…Consequently, radiomics focuses on extracting numerous features from radiological images to find hidden features of specific regions. These features include shape, texture, and intensity patterns, which are not readily visible to the human eye [2]. By utilizing computational techniques, radiomics aims to transform standard medical images, such as computed tomography (CT) scans, magnetic resonance imaging (MRI), and positron emission tomography (PET) scans, into high-dimensional data that can be analyzed to extract valuable information about tissue characteristics, disease progression, and treatment response [3,4].…”

Multimodality radiomics prediction of radiotherapy-induced the early proctitis and cystitis in rectal cancer patients: a machine learning study

“…The analysis of radiation toxicity in rectal cancer can be enhanced through the use of image-based features, which aid physicians in mitigating radiation risks and determining the feasibility of local tumor control [6,7]. Radiomics, a novel imaging analysis approach, involves the quantification of high-dimensional data extracted from medical images, providing valuable information about pathophysiological properties [8][9][10]. In the context of radiotherapy, radiomics feature analysis of the target volume and organs at risk (OARs) can have various applications, such as diagnostics, risk stratification, disease-free survival prediction, automatic segmentation, target volume definition, toxicity prognosis, treatment plan optimization, adaptive re-planning, decision support, treatment response assessment, and follow-up [8,[11][12][13].…”