Decoding Radiomics: A Step-by-Step Guide to Machine Learning Workflow in Hand-Crafted and Deep Learning Radiomics...

Area of Science:

• Radiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Radiomics research has rapidly expanded, extracting diagnostic and prognostic data from medical images (CT, PET, MRI).
• A significant gap exists between radiomics research findings and their clinical implementation, largely due to methodological challenges.
• Radiologists need foundational knowledge to evaluate radiomics workflows and collaborate on clinically impactful models.

Purpose of the Study:

• To provide a systematic guide to the radiomics study pipeline for radiologists.
• To introduce and detail the application of the METhodological RadiomICs Score (METRICS, 2024) for assessing radiomics study quality.
• To support researchers and reviewers in evaluating radiomics study robustness and facilitating clinical translation.

Main Methods:

• Systematic review of the radiomics study pipeline, covering design, preprocessing, feature selection, and model validation.
• Step-by-step guide on applying the METRICS (2024) tool for quality assessment.
• Focus on foundational knowledge for radiologists without specialized machine learning expertise.

Main Results:

• The review outlines a comprehensive radiomics workflow from study design to performance evaluation.
• The METRICS tool is presented as a structured method for appraising the quality and robustness of radiomics studies.
• The paper emphasizes the importance of methodological rigor for clinical translation.

Conclusions:

• Effective evaluation of radiomics study quality is crucial for bridging the gap between research and clinical practice.
• The METRICS tool offers a standardized approach for assessing radiomics research quality.
• This guide empowers radiologists to critically evaluate radiomics studies and contribute to developing reliable clinical tools.