Development and Multicenter Validation of a Deep Learning Model Using Whole Slide Imaging and MRI to Predict Distant Metastases in Nasopharyngeal Carcinoma

What this Trial Is About

Researchers are evaluating an AI model designed to predict the risk of distant metastasis in patients with nasopharyngeal carcinoma (NPC). This model uses whole slide imaging of pathology sections and MRI scans of the primary tumor and lymph nodes. The study aims to validate the AI model's accuracy across multiple centers and to determine its potential to guide PET/CT scan recommendations for patients with advanced NPC stages, helping to optimize medical resource use and reduce patient costs. The AI model is applied to patients with advanced NPC stages T3-4 or N2-3 who are recommended for PET/CT scans based on established NCCN and CSCO guidelines. The study involves prospective enrollment of patients to validate the diagnostic efficacy of this AI model. The model's threshold is set to ensure a negative predictive value of at least 95% for predicting absence of distant metastasis, supporting decisions on whether patients can be exempt from PET/CT examinations. Participants will undergo standard imaging including MRI of the nasopharynx and neck, and either PET/CT or conventional examinations to screen for distant metastasis. Researchers will measure the negative predictive value as the primary outcome over an average follow-up of two years, along with sensitivity, specificity, and positive predictive value as secondary outcomes. The study involves continuous follow-up through study completion to evaluate the AI model's performance in real-world clinical settings.

Development and Validation of a Deep Learning Model to Predict Distant Metastases in Nasopharyngeal Carcinoma Using Whole Slide Imaging and MRI