Radiomics, with its transformative potential for predicting colorectal cancer (CRC) prognosis, encounters challenges in clinical translation due to the unclear biological basis of risk stratification [1]. Bridging this gap is pivotal, and the emerging field of radiogenomics, situated at the intersection of radiomics and genomics, presents an opportunity to unravel the intricate relationship between imaging representations and molecular pathway dysregulation [2,3]. However, one-way association analyses have provided insufficient evidence, prompting the need for a multi-phase radiogenomics strategy. This strategy encompasses (1) the employment of tightly integrated forward and reverse radiogenomics engineering approaches [4], (2) the verification of micro-macro level biological associations [5], and (3) the validation of the radiomics signature across different datasets [6].