matRadiomics: A Novel and Complete Radiomics Framework, from Image Visualization to Predictive Model.

Area of Science:

• Medical imaging analysis
• Computational pathology
• Radiology informatics

Background:

• Radiomics analysis involves multiple steps: segmentation, feature extraction, selection, and model fitting.
• Current radiomics tools often require users to switch between different software, hindering workflow efficiency.

Purpose of the Study:

• To introduce matRadiomics, a novel, free, and user-friendly framework that integrates the entire radiomics workflow.
• To provide a unified platform for biomedical image analysis, feature extraction, selection, and predictive model development.

Main Methods:

• Development of the matRadiomics framework, integrating image visualization, segmentation, and feature extraction via Pyradiomics.
• Implementation of a hybrid descriptive-inferential method for feature selection.
• Training and validation of three machine learning classifiers (LDA, k-NN, SVM) using k-fold cross-validation.

Main Results:

• matRadiomics enables seamless import, inspection, segmentation of biomedical images, and extraction of radiomic features.
• The framework successfully integrates feature selection and trains predictive models with performance metrics displayed in a graphical interface.
• Demonstration of the workflow's efficacy through a real case study.

Conclusions:

• matRadiomics offers a comprehensive, integrated solution for radiomics analysis, from image visualization to predictive model implementation.
• The framework aims to establish a reference standard for radiomics analysis, enhancing clinical decision support.
• Future developments will further expand the capabilities and applications of the matRadiomics platform.