Image biomarkers and explainable AI

Feature extraction and selection in medical data are crucial for radiomics and image biomarker discovery, particularly using convolutional neural networks (CNNs). The process involves feature extraction, dimensionality reduction, and addressing the curse of dimensionality. While deep learning (DL) techniques perform well, handcrafted features are important for certain studies and need to be considered. Dataset size and diversity are also key factors in model stability, and high-performance computational resources are vital for training deep architectures. Collaborative learning, large language models, and model explainability are also discussed in this study. Non-DL methods often offer superior explainability, which can be enhanced through explainable AI and post hoc mechanisms. Key points: The key concepts for processing the imaging features to extract reliable and robust image biomarkers are provided. The main differences between radiomics and representation learning approaches are highlighted. The advantages and disadvantages of handcrafted versus learned features are given without losing sight of the clinical purpose of artificial intelligence models. Article: Image biomarkers and explainable AI: handcrafted features versus deep learned features Authors: Leonardo Rundo & Carmelo Militello