BE 645: Artificial Intelligence (AI) and Radiomics (Summer 2025) - Lecture 10

BE 645: Artificial Intelligence (AI) and Radiomics (Summer 2025)

This tenth lecture focuses on the critical process of feature selection in machine learning within the context of radiomics. The lecture covers various feature selection techniques, including filter, wrapper, and embedded methods, with detailed explanations of approaches like forward selection, backward elimination, and recursive feature elimination. Different normalization techniques (e.g., Normalizer, Standard, MinMax) and machine learning models (e.g., MLP, RF, AB, KNN) are discussed in relation to preprocessing, modeling, training, and performance reporting. The lecture provides extensive performance metrics for different combinations of scalers and models across various feature types (GLCM, GLRLM, GLSZM, Shape, First Order Features). It also compares experiments with and without feature selection, demonstrating how selecting relevant features can improve model performance by reducing overfitting, decreasing training time, and enhancing accuracy. The lecture emphasizes the importance of choosing appropriate feature selection methods and model combinations to optimize radiomics analysis outcomes.

Codes can be accessed using: https://github.com/HossamBalaha/BE-64...

Summer 2025 Playlist:    • BE 645: Artificial Intelligence (AI) and R...  

Spring 2025 Playlist (Old):    • BE 645: Artificial Intelligence (AI) and R...  

Summer 2024 Playlist (Old):    • BE 645: Artificial Intelligence (AI) and R...