Clinically Generalisable End-to-End Graph Learning for CT Image-Based Multitask Stroke Diagnosis
- 1University of Queensland
Abstract
Stroke remains a leading cause of mortality and long-term disability worldwide, yet rapid diagnosis is often limited by the shortage of trained radiologists, particularly in resource-constrained settings. Automated analysis of CT imaging offers a potential solution, but existing methods often struggle to achieve clinically generalisable performance while jointly addressing multiple diagnostic tasks. Here we present the Intelligent Integrated Stroke Diagnosis System (IISDS), an end-to-end deep learning framework built upon StrokeGNN, a graph-based architecture that integrates 3D contextual feature extraction with U-Net-based 2D lesion segmentation to enable comprehensive stroke analysis from non-contrast CT scans. IISDS performs stroke subtype classification, lesion segmentation and lesion volume estimation within a unified pipeline. To develop and validate the system, we collected and curated BGD-ISD through a collaboration between AI researchers, neurologists, radiologists and clinicians, resulting in a large multi-centre dataset comprising 1,507 CT scans from 597 stroke cases acquired across six hospitals and medical centres in Bangladesh. Across BGD-ISD and multiple publicly available datasets, IISDS achieves state-of-the-art performance on all tasks, improving segmentation accuracy by [≥] 0.011 Dice score, reducing lesion volume estimation error by [≥] 0.3 average symmetric surface distance (ASSD), and increasing classification performance by [≥] 0.018 area under the receiver operating characteristic curve (AUC) compared with existing approaches. These results demonstrate the potential of graph-based deep learning to enable clinically generalisable, automated and scalable stroke diagnosis from CT imaging, supporting rapid clinical decision-making, particularly in healthcare environments with limited access to expert radiological interpretation. The BGD-ISD dataset will be made openly available to the research community upon publication at: https://github.com/Zhicheng-Lu/stroke_ct.
Related
- Artificial Intelligence in Gastroenterology: Towards Explainable, Generalisable and Clinically Actionable Decision Support.
- MultiTask learning AI system to assist BCC diagnosis with dual explanation
- Patent foramen ovale-associated stroke: mechanisms, diagnosis, risk stratification, and management
- Inferring clinically relevant molecular subtypes of pancreatic cancer from routine histopathology using deep learning
- Pre-Obesity and Preclinical Obesity: Disease-Specific Concepts or Generalisable Frameworks for Chronic Disease Classification?
BibTeX
@misc{lu2026clinically,
title = {{Clinically Generalisable End-to-End Graph Learning for CT Image-Based Multitask Stroke Diagnosis}},
author = {Z. Lu and S. Uddin and S. Uribe and S. White and R. T. Martins and S. Chau and A. S. M. Mosaddek and M. S. Islam and N. Nahar and A. K. M. Azad and K. M. N. Hossain and H. S. Choudhury and K. M. R. Hasan and N. Mosaddek and S. Rahman and M. M. Hossain and K. M. M. H. Sizar and C. Angione and P. Lio and M. T. Islam and M. A. Moni},
howpublished = {medRxiv},
year = {2026},
month = aug,
doi = {10.64898/2026.08.26.26360026},
url = {https://www.medrxiv.org/content/10.64898/2026.08.26.26360026v1},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - Clinically Generalisable End-to-End Graph Learning for CT Image-Based Multitask Stroke Diagnosis
AU - Z. Lu
AU - S. Uddin
AU - S. Uribe
AU - S. White
AU - R. T. Martins
AU - S. Chau
AU - A. S. M. Mosaddek
AU - M. S. Islam
AU - N. Nahar
AU - A. K. M. Azad
AU - K. M. N. Hossain
AU - H. S. Choudhury
AU - K. M. R. Hasan
AU - N. Mosaddek
AU - S. Rahman
AU - M. M. Hossain
AU - K. M. M. H. Sizar
AU - C. Angione
AU - P. Lio
AU - M. T. Islam
AU - M. A. Moni
PY - 2026
DA - 2026/08/26
PB - medRxiv
DO - 10.64898/2026.08.26.26360026
UR - https://www.medrxiv.org/content/10.64898/2026.08.26.26360026v1
AB - Stroke remains a leading cause of mortality and long-term disability worldwide, yet rapid diagnosis is often limited by the shortage of trained radiologists, particularly in resource-constrained settings. Automated analysis of CT imaging offers a potential solution, but existing methods often struggle to achieve clinically generalisable performance while jointly addressing multiple diagnostic tasks. Here we present the Intelligent Integrated Stroke Diagnosis System (IISDS), an end-to-end deep learning framework built upon StrokeGNN, a graph-based architecture that integrates 3D contextual feature extraction with U-Net-based 2D lesion segmentation to enable comprehensive stroke analysis from non-contrast CT scans. IISDS performs stroke subtype classification, lesion segmentation and lesion volume estimation within a unified pipeline. To develop and validate the system, we collected and curated BGD-ISD through a collaboration between AI researchers, neurologists, radiologists and clinicians, resulting in a large multi-centre dataset comprising 1,507 CT scans from 597 stroke cases acquired across six hospitals and medical centres in Bangladesh. Across BGD-ISD and multiple publicly available datasets, IISDS achieves state-of-the-art performance on all tasks, improving segmentation accuracy by [≥] 0.011 Dice score, reducing lesion volume estimation error by [≥] 0.3 average symmetric surface distance (ASSD), and increasing classification performance by [≥] 0.018 area under the receiver operating characteristic curve (AUC) compared with existing approaches. These results demonstrate the potential of graph-based deep learning to enable clinically generalisable, automated and scalable stroke diagnosis from CT imaging, supporting rapid clinical decision-making, particularly in healthcare environments with limited access to expert radiological interpretation. The BGD-ISD dataset will be made openly available to the research community upon publication at: https://github.com/Zhicheng-Lu/stroke_ct.
ER -