Multimodal AI-based risk stratification for distant metastasis in nasopharyngeal carcinoma

Area of Science:

• Oncology
• Artificial Intelligence
• Medical Imaging

Background:

• Nasopharyngeal carcinoma (NPC) treatment relies on the TNM staging system, but patient outcomes vary.
• Current guidelines for managing distant metastasis in NPC are limited.
• Predicting NPC patient survival remains a challenge.

Purpose of the Study:

• To develop and validate a deep learning-based risk score for predicting nasopharyngeal carcinoma (NPC) survival.
• To improve risk stratification and personalized treatment for NPC patients.

Main Methods:

• Developed a graph for nasopharyngeal carcinoma (GNPC) risk score, a multimodal deep-learning model.
• Integrated haematoxylin and eosin-stained tissue slide images and clinical data.
• Validated the GNPC score on 1949 patients across two independent cohorts.

Main Results:

• The GNPC score significantly stratified patients for distant metastasis (P < 0.001), overall survival (P < 0.01), and local recurrence (P < 0.05).
• Downstream analyses revealed factors associated with GNPC-score-based risk groups, including morphology, molecular, and genomic profiles.

Conclusions:

• The GNPC digital score shows strong predictive performance for distant metastasis, overall survival, and local recurrence in NPC.
• This tool has the potential to enhance personalized treatment strategies and clinical management for NPC.