Opportunistic Screening for Pancreatic Cancer using Computed Tomography Imaging and Radiology Reports

2025-03-31

AI Summary

Pancreatic ductal adenocarcinoma (PDAC) poses significant clinical challenges due to its late diagnosis and poor five-year survival rate. Method: We propose a multimodal deep learning framework that integrates routine contrast-enhanced CT scans with unstructured radiology reports for opportunistic screening in asymptomatic populations and personalized survival risk prediction. Our approach uniquely fuses BERT-encoded textual features with 3D ResNet–extracted volumetric CT features via cross-modal attention, and incorporates a survival analysis–based loss function to optimize risk stratification. Contribution/Results: The model achieves C-indices of 0.675 (internal validation) and 0.644 (external validation), with statistically significant separation of Kaplan–Meier survival curves (p < 0.0001). Crucially, it operates without dedicated screening protocols, leveraging routinely acquired clinical data. This work establishes a clinically deployable paradigm for early PDAC detection and precision intervention.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: ML for personalized search and recommendations
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG

Abstract

Pancreatic ductal adenocarcinoma (PDAC) is a highly aggressive cancer, with most cases diagnosed at stage IV and a five-year overall survival rate below 5%. Early detection and prognosis modeling are crucial for improving patient outcomes and guiding early intervention strategies. In this study, we developed and evaluated a deep learning fusion model that integrates radiology reports and CT imaging to predict PDAC risk. The model achieved a concordance index (C-index) of 0.6750 (95% CI: 0.6429, 0.7121) and 0.6435 (95% CI: 0.6055, 0.6789) on the internal and external dataset, respectively, for 5-year survival risk estimation. Kaplan-Meier analysis demonstrated significant separation (p<0.0001) between the low and high risk groups predicted by the fusion model. These findings highlight the potential of deep learning-based survival models in leveraging clinical and imaging data for pancreatic cancer.

Problem

• Early detection of pancreatic cancer using CT and reports
• Predicting PDAC risk with deep learning fusion model
• Improving survival prognosis by integrating imaging and clinical data

Innovation

• Deep learning fusion model for PDAC risk prediction
• Integrates radiology reports and CT imaging data
• Achieves significant survival risk group separation