In this chapter, we explore the landscape of artificial intelligence and radiomics in molecular oncology imaging. Advances in molecular imaging methods offer powerful insights into tumor biology but generate complex data beyond the capability of traditional visual interpretation alone. To address this complexity, radiomics emerged as a structured method to quantify imaging features, transforming subtle textures and tumor shapes into meaningful biomarkers. More recently, deep learning approaches—particularly convolutional neural networks—have taken this concept further, automatically learning intricate patterns directly from imaging data. Additionally, we introduce transformer architectures, a newer AI approach, capable of capturing global image context and integrating imaging information seamlessly with other clinical data. Throughout, we discuss the strengths and limitations of each method, highlighting challenges such as large data requirements, standardization, interpretability, and rigorous validation needed for clinical translation. Finally, we emphasize emerging techniques, such as self-supervised and multimodal learning, which promise to overcome these limitations and push molecular imaging toward genuinely personalized cancer care.