P2.067: Artificial Intelligence in Pediatric Solid Tumor Research and Clinical Care: A Scoping Review

Background: Clinical and research advances for children with solid tumors are limited by the rare nature of the tumors and increasing complexity of molecular and biologic information available. Artificial intelligence (AI) and machine learning (ML) are often used to analyze large datasets; however, they hold the potential for advancing diagnosis, risk stratification, and treatment in rare diseases where individual patients contribute high-dimensional data. This review characterizes current AI/ML applications in rare pediatric extra-cranial solid tumors.

Methods: PubMed, Embase, and Web-of-Science Core Collection databases were queried for relevant articles published before February 10, 2025. Eligible studies applied AI/ML methods to study rare tumors in pediatric populations (≤19 years). Two reviewers independently screened all articles. Due to the limited number of studies identified of rare non-solid tumors, only studies of solid tumors were included in this review. Eligible articles were evaluated for tumor diagnosis and AI/ML methods, clinical applications, comparators, effectiveness, and interpretability.

Results: The search identified 4,687 studies. Eighty-two studies underwent full text review and 23 underwent final analysis. Most were published after 2020 (19/23) and originated from China (10/23). Liver tumors were the most frequently studied tumor (16/23), followed by differentiated thyroid carcinoma (6/23), extragonadal germ cell tumors (2/23), pheochromocytoma (2/23), and adrenocortical carcinoma (2/23). Multiple studies analyzed more than one tumor type (9/23). The AI approaches most used included: supervised learning (20/23), unsupervised learning (11/23), hybrid models (9/23), and deep learning (4/23). Applications included diagnostics (16/23), prognostics (10/23), and risk stratification (9/23), with additional applications in imaging interpretation (8/23), clinical decision support (5/23), integration of multimodal data (4/23), and treatment-related tasks (5/23). Twenty studies reported effectiveness measures, with many models achieving AUCs >0.85. Among the 16 studies with comparators, AI/ML methods were often equal or superior compared to expert consensus, traditional statistical models, or alternative algorithms.

Conclusion: Current AI/ML research in rare pediatric extra-cranial solid tumors focuses on diagnosis, risk stratification, and prognosis, often outperforming traditional methods. Future work should prioritize external validation and clinical applicability.