Integrating artificial intelligence into cancers of unknown primary diagnosis and treatment.

Authors

Abstract

Cancers of unknown primary (CUP) refer to a highly heterogeneous group of metastatic tumors whose primary site remains undetectable despite comprehensive conventional evaluations. Characterized by obscure primary origins and extremely poor prognosis, the pathogenesis of CUP remains incompletely elucidated. Empirical chemotherapy, the traditional mainstay of treatment, yields limited efficacy, while emerging therapeutic strategies lack sufficient evidence from randomized controlled trials. Consequently, CUP continues to pose a formidable challenge in clinical practice. However, advances in artificial intelligence (AI), particularly in deep learning, have enabled reliable performance in oncology-related tasks, including tumor diagnosis, treatment response prediction, and prognostic assessment. In CUP research, AI is predominantly applied to develop diagnostic models for predicting tumor tissue of origin (TOO) or molecular subtypes, with a small number of studies focusing on the prediction of treatment response and survival. Current AI models that integrate multi-modal data (e.g., molecular data, medical imaging) leverage their advantages in high-throughput data processing and in-depth feature mining to overcome the limitations of traditional CUP diagnosis and treatment, providing new avenues to better understand this complex disease. Given the substantial progress of AI in CUP, this review systematically summarizes the current research status and latest breakthroughs of AI in CUP from three perspectives: AI-based diagnosis using molecular data, AI-based diagnosis using medical imaging, and AI-assisted prediction of treatment response and prognosis. The aim is to promote precise diagnosis and treatment of CUP and improve patient outcomes.

Tags

Topics