High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis

Abstract This study proposes a semi-weakly supervised learning approach for pulmonary embolism (PE) detection on CT pulmonary angiography (CTPA) to alleviate the resource-intensive burden of exhaustive medical image annotation. Attention-based CNN-RNN models were trained on the RSNA pulmonary emboli...

Full description

Saved in:
Bibliographic Details
Main Authors: Zixuan Hu, Hui Ming Lin, Shobhit Mathur, Robert Moreland, Christopher D. Witiw, Laura Jimenez-Juan, Matias F. Callejas, Djeven P. Deva, Ervin Sejdić, Errol Colak
Format: Article
Language:English
Published: Nature Portfolio 2025-05-01
Series:npj Digital Medicine
Online Access:https://doi.org/10.1038/s41746-025-01594-2
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1849729002400382976
author Zixuan Hu
Hui Ming Lin
Shobhit Mathur
Robert Moreland
Christopher D. Witiw
Laura Jimenez-Juan
Matias F. Callejas
Djeven P. Deva
Ervin Sejdić
Errol Colak
author_facet Zixuan Hu
Hui Ming Lin
Shobhit Mathur
Robert Moreland
Christopher D. Witiw
Laura Jimenez-Juan
Matias F. Callejas
Djeven P. Deva
Ervin Sejdić
Errol Colak
author_sort Zixuan Hu
collection DOAJ
description Abstract This study proposes a semi-weakly supervised learning approach for pulmonary embolism (PE) detection on CT pulmonary angiography (CTPA) to alleviate the resource-intensive burden of exhaustive medical image annotation. Attention-based CNN-RNN models were trained on the RSNA pulmonary embolism CT dataset and externally validated on a pooled dataset (Aida and FUMPE). Three configurations included weak (examination-level labels only), strong (all examination and slice-level labels), and semi-weak (examination-level labels plus a limited subset of slice-level labels). The proportion of slice-level labels varying from 0 to 100%. Notably, semi-weakly supervised models using approximately one-quarter of the total slice-level labels achieved an AUC of 0.928, closely matching the strongly supervised model’s AUC of 0.932. External validation yielded AUCs of 0.999 for the semi-weak and 1.000 for the strong model. By reducing labeling requirements without sacrificing diagnostic accuracy, this method streamlines model development, accelerates the integration of models into clinical practice, and enhances patient care.
format Article
id doaj-art-fd4acf6f6dc3492da32c2c1efe4c78ee
institution DOAJ
issn 2398-6352
language English
publishDate 2025-05-01
publisher Nature Portfolio
record_format Article
series npj Digital Medicine
spelling doaj-art-fd4acf6f6dc3492da32c2c1efe4c78ee2025-08-20T03:09:21ZengNature Portfolionpj Digital Medicine2398-63522025-05-01811910.1038/s41746-025-01594-2High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosisZixuan Hu0Hui Ming Lin1Shobhit Mathur2Robert Moreland3Christopher D. Witiw4Laura Jimenez-Juan5Matias F. Callejas6Djeven P. Deva7Ervin Sejdić8Errol Colak9The Edward S. Rogers Department of Electrical and Computer Engineering, University of TorontoDepartment of Medical Imaging, St Michael’s Hospital, Unity Health TorontoDepartment of Medical Imaging, St Michael’s Hospital, Unity Health TorontoDepartment of Medical Imaging, St Michael’s Hospital, Unity Health TorontoLi Ka Shing Knowledge Institute, St Michael’s Hospital, Unity Health TorontoDepartment of Medical Imaging, St Michael’s Hospital, Unity Health TorontoDepartment of Medical Imaging, St Michael’s Hospital, Unity Health TorontoDepartment of Medical Imaging, St Michael’s Hospital, Unity Health TorontoThe Edward S. Rogers Department of Electrical and Computer Engineering, University of TorontoDepartment of Medical Imaging, St Michael’s Hospital, Unity Health TorontoAbstract This study proposes a semi-weakly supervised learning approach for pulmonary embolism (PE) detection on CT pulmonary angiography (CTPA) to alleviate the resource-intensive burden of exhaustive medical image annotation. Attention-based CNN-RNN models were trained on the RSNA pulmonary embolism CT dataset and externally validated on a pooled dataset (Aida and FUMPE). Three configurations included weak (examination-level labels only), strong (all examination and slice-level labels), and semi-weak (examination-level labels plus a limited subset of slice-level labels). The proportion of slice-level labels varying from 0 to 100%. Notably, semi-weakly supervised models using approximately one-quarter of the total slice-level labels achieved an AUC of 0.928, closely matching the strongly supervised model’s AUC of 0.932. External validation yielded AUCs of 0.999 for the semi-weak and 1.000 for the strong model. By reducing labeling requirements without sacrificing diagnostic accuracy, this method streamlines model development, accelerates the integration of models into clinical practice, and enhances patient care.https://doi.org/10.1038/s41746-025-01594-2
spellingShingle Zixuan Hu
Hui Ming Lin
Shobhit Mathur
Robert Moreland
Christopher D. Witiw
Laura Jimenez-Juan
Matias F. Callejas
Djeven P. Deva
Ervin Sejdić
Errol Colak
High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis
npj Digital Medicine
title High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis
title_full High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis
title_fullStr High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis
title_full_unstemmed High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis
title_short High performance with fewer labels using semi-weakly supervised learning for pulmonary embolism diagnosis
title_sort high performance with fewer labels using semi weakly supervised learning for pulmonary embolism diagnosis
url https://doi.org/10.1038/s41746-025-01594-2
work_keys_str_mv AT zixuanhu highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis
AT huiminglin highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis
AT shobhitmathur highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis
AT robertmoreland highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis
AT christopherdwitiw highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis
AT laurajimenezjuan highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis
AT matiasfcallejas highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis
AT djevenpdeva highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis
AT ervinsejdic highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis
AT errolcolak highperformancewithfewerlabelsusingsemiweaklysupervisedlearningforpulmonaryembolismdiagnosis