Radiomics

This lecture provides a comprehensive introduction to radiomics analysis and its applications in neuro-oncological research, specifically focusing on how medical images can be transformed into mineable, high-dimensional data to support clinical decision-making. The lecture details the standard radiomics workflow, which includes image preprocessing, extraction of intensity-, shape-, and texture-based features, and the use of machine learning for feature selection and model training. Using a practical example involving low-grade gliomas and the PyRadiomics Python package, the video demonstrates how these quantitative features can be used to predict molecular subtypes and conduct survival analysis, effectively bridging the gap between medical imaging and personalized medicine.

Learning objectives:
By the end of this lecture, students will be able to:

• Define Radiomics, and understand how medical images are transformed into mineable, high-dimensional quantitative data.
• Identify the essential steps from image acquisition and segmentation to feature extraction and model training.
• Differentiate between first-order (intensity), shape-based, and texture-based features (e.g., GLCM, GLRLM).
• Understand how to use feature selection and classification algorithms (like SVM or Logistic Regression) to predict tumor phenotypes.