Deep Learning Approach for Renal Tumor from the Viewpoint of Secondary Prevention: A Narrative Review

We conducted a review of deep-learning articles focusing on renal tumors from the perspective of secondary prevention using a search expression constructed by the Japan Pharmaceutical Manufacturers Association. Twenty-seven articles were found, of which computed tomography (CT), histopathological images, and genomic data were analyzed in 18, 7, and 2 articles, respectively. Tasks were categorized into classification, segmentation, generation of synthetic contrast-enhanced CT, and scoring. These articles aimed to contribute, at least in part, to secondary prevention, defined as the reduction of the serious consequences of disease through early diagnosis and treatment. As future directions, this review identifies two key issues in the field of renal tumors. First, there is a lack of multimodal analysis that reflects a comprehensive understanding of patient information such as age, sex, medical imaging, and laboratory tests as used in clinical practice. Second, there are limited studies employing generative artificial intelligence from the viewpoint of secondary prevention. Future deep-learning research on renal tumors should address these points to better support secondary prevention.