AI-Driven Radiomics: Bridging Quantitative Imaging and Precision Medicine

Abstract

Radiomics has transformed medical imaging by converting CT, MRI, PET, and ultrasound scans into quantitative features that capture tumor heterogeneity and inform clinical decision-making. While traditional handcrafted radiomics has shown promise in predicting survival and treatment response, it is limited by variability and lack of reproducibility. The integration of artificial intelligence, particularly deep learning, has expanded radiomics by enabling automated feature extraction, improved accuracy, and multimodal data fusion. Emerging approaches such as explainable AI and federated learning address interpretability and privacy challenges, supporting broader adoption. However, real-world translation requires harmonization of imaging protocols, validation in multi-center studies, and integration into clinical workflows. By combining quantitative imaging, computational modeling, and robust statistical methods, AI-driven radiomics has the potential to reshape precision oncology, offering more reliable tools for diagnosis, prognosis, and personalized therapy. Its future impact will depend on interdisciplinary collaboration and sustained investment to ensure reproducibility, transparency, and equity in clinical applications.