Data-efficient machine learning for the detection of duodenal neoplasia: a multi-centre study
- 1School of Advanced Studies Sant'Anna
Abstract
BackgroundMachine Learning (ML) has shown promise in histopathological diagnosis, but rare tumours, like duodenal epithelial neoplasms, are challenging because large, labelled datasets are difficult to assemble. We propose a data-efficient ML pipeline for detecting duodenal epithelial neoplasia in whole-slide images.
MethodsWe assembled a five-hospital dataset of haematoxylin and eosin-stained duodenal biopsy whole slide images (WSIs) at 10x magnification, comprising 8,027 normal and 401 neoplastic biopsies, including adenomas, carcinomas and neuroendocrine tumours. WSIs were divided into patches and encoded using the Hibou-B foundation model. We compared global standardisation with unsupervised site-specific standardisation of patch embeddings. Our pipeline combined principal component analysis, a Gaussian mixture model and a Random Forest classifier. We assessed generalisability using leave-one-hospital-out cross-validation across four hospitals and on an independent test set from a fifth hospital.
ResultsIn leave-one-hospital-out cross-validation, the pipeline achieved a mean accuracy of 90%, sensitivity of 96%, specificity of 83% and an AUC of 0.96. On the external test hospital, it achieved 94% accuracy, 93% sensitivity, 95% specificity, and AUC of 0.97. Site-specific standardisation improved mean cross-validation accuracy by 20 percentage points compared with global standardisation, while the Gaussian mixture pipeline improved accuracy by 12 percentage points compared with mean pooling of patch embeddings.
ConclusionsWe present a data-efficient ML approach for detecting duodenal epithelial neoplasia from WSIs. Strong performance was maintained across hospitals and on an independent external test cohort despite the limited number of neoplastic training cases, supporting the feasibility of foundation-model-based approaches for rare histopathological diagnoses.
Related
- WHAT IS NORMAL? MULTIMODAL CHARACTERIZATION OF NON-DISEASED PEDIATRIC DUODENAL BIOPSIES USING MACHINE LEARNING IMAGE ANALYSIS AND TRANSCRIPTOMICS
- Advancing Lumbar Disc Herniation Classification from MRI Using an Efficient and Reproducible Classical Machine Learning Framework
- Machine Learning and Deep Learning for Early Disease Detection Across Modalities: A Comprehensive Review of Methods, Applications, and Clinical Translation
- Machine Learning-Based Fatigue Detection from EEG Signals: A Methodological Pilot Framework with Potential for Nursing Applications
- Machine Learning Detection of Peritoneal Entry from Vibroacoustic Signals During Veress Needle Insertion
BibTeX
@misc{lisio2026dataefficient,
title = {{Data-efficient machine learning for the detection of duodenal neoplasia: a multi-centre study}},
author = {F. Di Lisio and E. J. Soilleux and F. Jaeckle},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.08.26365053},
url = {https://www.medrxiv.org/content/10.64898/2026.10.08.26365053v1},
note = {CC BY-NC},
}
RIS
TY - UNPB
TI - Data-efficient machine learning for the detection of duodenal neoplasia: a multi-centre study
AU - F. Di Lisio
AU - E. J. Soilleux
AU - F. Jaeckle
PY - 2026
DA - 2026/10/08
PB - medRxiv
DO - 10.64898/2026.10.08.26365053
UR - https://www.medrxiv.org/content/10.64898/2026.10.08.26365053v1
AB - BackgroundMachine Learning (ML) has shown promise in histopathological diagnosis, but rare tumours, like duodenal epithelial neoplasms, are challenging because large, labelled datasets are difficult to assemble. We propose a data-efficient ML pipeline for detecting duodenal epithelial neoplasia in whole-slide images. MethodsWe assembled a five-hospital dataset of haematoxylin and eosin-stained duodenal biopsy whole slide images (WSIs) at 10x magnification, comprising 8,027 normal and 401 neoplastic biopsies, including adenomas, carcinomas and neuroendocrine tumours. WSIs were divided into patches and encoded using the Hibou-B foundation model. We compared global standardisation with unsupervised site-specific standardisation of patch embeddings. Our pipeline combined principal component analysis, a Gaussian mixture model and a Random Forest classifier. We assessed generalisability using leave-one-hospital-out cross-validation across four hospitals and on an independent test set from a fifth hospital. ResultsIn leave-one-hospital-out cross-validation, the pipeline achieved a mean accuracy of 90%, sensitivity of 96%, specificity of 83% and an AUC of 0.96. On the external test hospital, it achieved 94% accuracy, 93% sensitivity, 95% specificity, and AUC of 0.97. Site-specific standardisation improved mean cross-validation accuracy by 20 percentage points compared with global standardisation, while the Gaussian mixture pipeline improved accuracy by 12 percentage points compared with mean pooling of patch embeddings. ConclusionsWe present a data-efficient ML approach for detecting duodenal epithelial neoplasia from WSIs. Strong performance was maintained across hospitals and on an independent external test cohort despite the limited number of neoplastic training cases, supporting the feasibility of foundation-model-based approaches for rare histopathological diagnoses.
ER -