Machine Learning Approaches to the Early Detection of Pancreatic Cancer from Time-Series Primary Care Data

Aug 4, 2026

Pancreatic cancer remains difficult to detect early, since diagnosis often depends on symptoms that emerge only in advanced stages. Routine blood tests can carry predictive signals before symptoms develop, but limited work has evaluated time-varying laboratory information prior to diagnosis. Using UK primary care data, this study compares machine learning methods for detecting pancreatic cancer at multiple lead times despite irregular, sparse real-world time-series.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages11

Reading Time28 min

AccessFree to download

Abstract

Background and problem motivation

Data and methods comparison

Findings across lead times and calibration

Implications

[eprints@whiterose.ac.uk](eprints@whiterose.ac.uk)[ ](eprints@whiterose.ac.uk)[https://eprints.whiterose.ac.uk](https://eprints.whiterose.ac.uk) Universities of Leeds, Sheffield and York Deposited via The University of Leeds. White Rose Research Online URL for this paper: [https://eprints.whiterose.ac.uk/id/eprint/228286/](https://eprints.whiterose.ac.uk/id/eprint/228286/) [Version: Accepted Version](Version: Accepted Version) Proceedings Paper: Moglia, V. , Smith, L. , Cook, G. et al. (2025) Machine Learning Approaches to the Early Detection of Pancreatic Cancer from Time-Series Primary Care Data. In: Artificial Intelligence in Medicine. 23rd International Conference , AIME 2025, 23-26 Jun 2025, Pavia, Italy. Lecture Notes in Computer Science, 15734. Springer, pp. 313-322. ISBN: 978- 3-031-95837-3. [https://doi.org/10.1007/978-3-031-95838-0_31](https://doi.org/10.1007/978-3-031-95838-0_31) © 2025 The Author(s), under exclusive license to Springer Nature Switzerland AG. This version of the article has been accepted for publication, after peer review (when applicable) and is subject to Springer Nature’s AM terms of use ( [https://www.springernature.com/gp/open-research/policies/accepted-manuscript-terms](https://www.springernature.com/gp/open-research/policies/accepted-manuscript-terms)), but is not the Version of Record and does not reflect post-acceptance improvements, or any corrections. The Version of Record is available online at: [https://doi.org/10.1007/978-](https://doi.org/10.1007/978-)[ ](https://doi.org/10.1007/978-)[3-031-95838-0_31.](3-031-95838-0_31.) Reuse Items deposited in White Rose Research Online are protected by copyright, with all rights reserved unless indicated otherwise. They may be downloaded and/or printed for private study, or other acts as permitted by national copyright laws. The publisher or other rights holders may allow further reproduction and re-use of the full text version. This is indicated by the licence information on the White Rose Research Online record for the item. Takedown If you consider content in White Rose Research Online to be in breach of UK law, please notify us by emailing [eprints@whiterose.ac.uk](eprints@whiterose.ac.uk) including the URL of the record and the reason for the withdrawal request. Machine Learning Approaches to the Early Detection of Pancreatic Cancer from Time-Series Primary Care Data Victoria Moglia1[0009􀀀0001􀀀9124􀀀8030], Lesley Smith2[0000􀀀0002􀀀4280􀀀6323], Gordon Cook2 ,3[0000􀀀0003􀀀1717􀀀0412], Marc De Kamps 1[0000 􀀀0001􀀀7162􀀀4425] , and Owen Johnson 1[0000􀀀0003􀀀3998􀀀541X] 1 School of Computing, University of Leeds, Leeds, UK 2 Leeds Institute for Clinical Trials Research, University of Leeds, Leeds, UK 3 NIHR Leeds Biomedical Research Centre, Leeds Teaching Hospitals Trust, Leeds, UK Abstract. Pancreatic cancer is notoriously difficult to detect, with diagnosis often relying on symptoms that only develop at advanced stages of the disease. Routine blood tests may signal a developing cancer before these symptoms appear. Limited research has investigated the use of time-varying information from laboratory tests before diagnosis. This study used UK primary care data to compare machine learning approaches for detecting pancreatic cancer at various time-intervals before diagnosis. The machine learning challenge is that such real-world data is irregular and sparse and therefore difficult to use for model creation. In this study, deep learning time-series models (LSTM and GRU-D) were compared to a feature engineering approach. We found that while predictive performance was strongest at diagnosis date (maximum AUROC of 0.85), cases could be detected 18 months before diagnosis, with GRUD achieving an AUROC of 0.57 . Closer to the diagnosis date, where diagnostic signals are stronger, feature engineering approaches outperformed the deep learning models. However, further from diagnosis, the deep learning models, particul

FAQs

Why is early detection of pancreatic cancer challenging?

Diagnosis typically relies on symptoms that appear only at advanced stages, while early symptoms are often silent and nonspecific.

What data source and lead times are used in this study?

The study uses UK primary care data and evaluates model performance at various time-intervals before diagnosis.

Which machine learning approaches were compared, and how did they perform?

Deep learning time-series models (LSTM and GRU-D) were compared with a feature engineering approach. Performance was strongest at the diagnosis date (maximum AUROC 0.85), and cases could still be detected 18 months before diagnosis, with GRU-D achieving AUROC 0.57.

Aug 4, 2026

Upload by River Wang