Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

David Le, Ramon Correa-Medero, Amara Tariq, Bhavik Patel, Motoyo Yano, Imon Banerjee

TL;DR

Pancreatic ductal adenocarcinoma (PDAC) remains highly lethal due to late diagnosis. The authors develop a multimodal survival model that opportunistically leverages pre-diagnostic CT imaging and radiology reports to estimate PDAC risk up to five years before diagnosis. The fusion model combining SBERT-derived text embeddings and radiomics features achieves the strongest prognostic performance, with internal and external concordance indices of $0.6750$ and $0.6435$, respectively, and significant Kaplan-Meier separation ($p<0.0001$). This approach demonstrates the potential for AI-driven, pre-diagnostic risk scoring to enable earlier intervention and targeted screening for PDAC, with plans to build an end-to-end pipeline incorporating volumetric imaging and text data.

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.

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

TL;DR

Pancreatic ductal adenocarcinoma (PDAC) remains highly lethal due to late diagnosis. The authors develop a multimodal survival model that opportunistically leverages pre-diagnostic CT imaging and radiology reports to estimate PDAC risk up to five years before diagnosis. The fusion model combining SBERT-derived text embeddings and radiomics features achieves the strongest prognostic performance, with internal and external concordance indices of and , respectively, and significant Kaplan-Meier separation (). This approach demonstrates the potential for AI-driven, pre-diagnostic risk scoring to enable earlier intervention and targeted screening for PDAC, with plans to build an end-to-end pipeline incorporating volumetric imaging and text data.

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.

Paper Structure

This paper contains 11 sections, 2 figures, 3 tables.

Table of Contents

  1. Abstract
  2. Introduction
  3. Methods
  4. Text-Only Model: SBERT Embedding
  5. Image-Only Model: PyRadiomics + MLP
  6. Multimodality Fusion Model
  7. Implementation
  8. Results
  9. Ablation Study
  10. Conclusion
  11. Acknowledgements

Figures (2)

  • Figure 1: Pipeline for (a) text-only model, (b) image-only model, and (c) fusion model integrating text-based and CT volumetric information into a survival model. Sentence-BERT was used to extract embeddings from clinical report sentences, while pancreas segmentation was performed using TotalSegmentator, followed by feature extraction with PyRadiomics.
  • Figure 2: Kaplan-Meier curves for risk stratification between low and high risk for the image only, text only and fusion models for the internal and external validation dataset.