Radiomics

Radiomics is the high-throughput extraction of quantiative features from medical images for textural analysis that spatially characterises regions/volumes of interest and may provide insight into the underlying pathophysiology to support diagnosis, prognosis, and treatment planning (Barry et al., 2025). It bridges radiology and data science, allowing for a more detailed understanding of disease characteristics.

Radiomics analysis workflow (Source)

Learn the Foundations

• Intro to Python (CS50 Week 6)
  Radiomics pipelines are often built in Python, so this course provides a solid foundation in the language.

• PyRadiomics Documentation
  The official documentation for PyRadiomics, a robust open-source Python package for extracting engineered features from medical images.

• Deep Learning Specialization by Andrew Ng
  A comprehensive course that covers the fundamentals of deep learning. Highly relevant for radiomics pipelines that involve neural networks for image feature extraction or outcome prediction.

• Lightning in 15 Minutes
  A concise tutorial introducing PyTorch Lightning—an efficient and reproducible way to structure deep learning projects, especially useful for radiomics model development.

Key Papers and Case Studies

• Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach — Nature Communications
  This seminal paper is widely credited with sparking interest in radiomics by demonstrating how quantitative imaging features can decode tumor phenotype.

• Radiomics strategies for risk assessment of tumour failure in head-and-neck cancer — Scientific Reports
  A strong example of using engineered radiomic features to predict treatment outcomes in head-and-neck cancer.

• Deep learning for lung cancer prognostication: A retrospective multi-cohort radiomics study — PLOS Medicine
  Combines radiomics with deep learning to predict survival outcomes, showcasing the power of hybrid approaches.

• Artificial intelligence in radiology — Nature Reviews Cancer
  A high-level overview of AI’s role in modern radiology, including how radiomics fits into the broader clinical and research landscape.