Deep Learning-Based Analysis of Gene Expression Data and Gene-Related Information in Pediatric Surgical Oncology: A Scoping Review.

Deep learning (DL) methods may enhance analysis of complex gene expression data to aid in diagnosis and treatment planning for pediatric extracranial tumors. However, the literature regarding the application of DL to gene expression data in this field remains limited. This scoping review was based on the question "What is the current status of research in gene expression, gene-related information and deep learning-based analyses for pediatric surgical oncology". We conducted a scoping review in accordance with the PRISMA-ScR guidelines. A systematic search of PubMed, Scopus, and Embase was performed to identify studies applying DL models to gene-related data in pediatric extracranial solid tumors. After deduplication, title and abstract screening and full-text screening, nine studies met the inclusion criteria. Neuroblastoma was the most commonly studied tumor type (n = 6), with classification and survival prediction as applications. In general, the studies reported strong performance; however, external validation was rarely reported. Although the application of DL to gene-related data in pediatric solid tumors remains in its infancy, current studies highlight the diversity and potential of approaches that could improve classification, prognostication, and the treatment of patients. The large variety of technical approaches reflects the ongoing process of adaptation to gene-related data. Advancing this field will require larger datasets, consistent methodology, external, and prospective validation within a cross-disciplinary setting.