Comparing efficiency of an attention-based deep learning network with contemporary radiological workflow for pulmonary embolism detection on CTPA: A retrospective study

Rational and objectives: Pulmonary embolism (PE) is the third most fatal cardiovascular disease in the United States. Currently, Computed Tomography Pulmonary Angiography (CTPA) serves as diagnostic gold standard for detecting PE. However, its efficacy is limited by factors such as contrast bolus ti...

Full description

Saved in:
Bibliographic Details
Main Authors: Gagandeep Singh, Annie Singh, Tejasvi Kainth, Sudhir Suman, Nicole Sakla, Luke Partyka, Tej Phatak, Prateek Prasanna
Format: Article
Language:English
Published: Elsevier 2025-06-01
Series:European Journal of Radiology Open
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S2352047725000243
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850143613173891072
author Gagandeep Singh
Annie Singh
Tejasvi Kainth
Sudhir Suman
Nicole Sakla
Luke Partyka
Tej Phatak
Prateek Prasanna
author_facet Gagandeep Singh
Annie Singh
Tejasvi Kainth
Sudhir Suman
Nicole Sakla
Luke Partyka
Tej Phatak
Prateek Prasanna
author_sort Gagandeep Singh
collection DOAJ
description Rational and objectives: Pulmonary embolism (PE) is the third most fatal cardiovascular disease in the United States. Currently, Computed Tomography Pulmonary Angiography (CTPA) serves as diagnostic gold standard for detecting PE. However, its efficacy is limited by factors such as contrast bolus timing, physician-dependent diagnostic accuracy, and time taken for scan interpretation. To address these limitations, we propose an AI-based PE triaging model (AID-PE) designed to predict the presence and key characteristics of PE on CTPA. This model aims to enhance diagnostic accuracy, efficiency, and the speed of PE identification. Materials and methods: We trained AID-PE on the RSNA-STR PE CT (RSPECT) Dataset, N = 7279 and subsequently tested it on an in-house dataset (n = 106). We evaluated efficiency in a separate dataset (D4, n = 200) by comparing the time from scan to report in standard PE detection workflow versus AID-PE. Results: A comparative analysis showed that AID-PE had an AUC/accuracy of 0.95/0.88. In contrast, a Convolutional Neural Network (CNN) classifier and a CNN-Long Short-Term Memory (LSTM) network without an attention module had an AUC/accuracy of 0.5/0.74 and 0.88/0.65, respectively. Our model achieved AUCs of 0.82 and 0.95 for detecting PE on the validation dataset and the independent test set, respectively. On D4, AID-PE took an average of 1.32 s to screen for PE across 148 CTPA studies, compared to an average of 40 min in contemporary workflow. Conclusion: AID-PE outperformed a baseline CNN classifier and a single-stage CNN-LSTM network without an attention module. Additionally, its efficiency is comparable to the current radiological workflow.
format Article
id doaj-art-ffdf3410a7764581bcf33e1ea9b8c656
institution OA Journals
issn 2352-0477
language English
publishDate 2025-06-01
publisher Elsevier
record_format Article
series European Journal of Radiology Open
spelling doaj-art-ffdf3410a7764581bcf33e1ea9b8c6562025-08-20T02:28:38ZengElsevierEuropean Journal of Radiology Open2352-04772025-06-011410065710.1016/j.ejro.2025.100657Comparing efficiency of an attention-based deep learning network with contemporary radiological workflow for pulmonary embolism detection on CTPA: A retrospective studyGagandeep Singh0Annie Singh1Tejasvi Kainth2Sudhir Suman3Nicole Sakla4Luke Partyka5Tej Phatak6Prateek Prasanna7Department of Radiology, Columbia University Irving Medical Center, NY, USA; Correspondence to: Columbia University Irving Medical Center, 622 W 168th St, NY, New York 10032, USA.Atal Bihari Vajpayee Institute of Medical Sciences, New Delhi, IndiaDepartment of Psychiatry, BronxCare Health System, NY, USADepartment of Biomedical Informatics, Stony Brook University, USADepartment of Radiology, MedStar Washington Hospital Center, DC, USADepartment of Radiology, Rutgers-Newark Beth Israel Medical Center, NJ, USADepartment of Radiology, Rutgers-Newark Beth Israel Medical Center, NJ, USADepartment of Biomedical Informatics, Stony Brook University, USARational and objectives: Pulmonary embolism (PE) is the third most fatal cardiovascular disease in the United States. Currently, Computed Tomography Pulmonary Angiography (CTPA) serves as diagnostic gold standard for detecting PE. However, its efficacy is limited by factors such as contrast bolus timing, physician-dependent diagnostic accuracy, and time taken for scan interpretation. To address these limitations, we propose an AI-based PE triaging model (AID-PE) designed to predict the presence and key characteristics of PE on CTPA. This model aims to enhance diagnostic accuracy, efficiency, and the speed of PE identification. Materials and methods: We trained AID-PE on the RSNA-STR PE CT (RSPECT) Dataset, N = 7279 and subsequently tested it on an in-house dataset (n = 106). We evaluated efficiency in a separate dataset (D4, n = 200) by comparing the time from scan to report in standard PE detection workflow versus AID-PE. Results: A comparative analysis showed that AID-PE had an AUC/accuracy of 0.95/0.88. In contrast, a Convolutional Neural Network (CNN) classifier and a CNN-Long Short-Term Memory (LSTM) network without an attention module had an AUC/accuracy of 0.5/0.74 and 0.88/0.65, respectively. Our model achieved AUCs of 0.82 and 0.95 for detecting PE on the validation dataset and the independent test set, respectively. On D4, AID-PE took an average of 1.32 s to screen for PE across 148 CTPA studies, compared to an average of 40 min in contemporary workflow. Conclusion: AID-PE outperformed a baseline CNN classifier and a single-stage CNN-LSTM network without an attention module. Additionally, its efficiency is comparable to the current radiological workflow.http://www.sciencedirect.com/science/article/pii/S2352047725000243Pulmonary EmbolismDeep learningNeural NetworkComputed Tomography Pulmonary Angiogram (CTPA)Medical imagesRadiology
spellingShingle Gagandeep Singh
Annie Singh
Tejasvi Kainth
Sudhir Suman
Nicole Sakla
Luke Partyka
Tej Phatak
Prateek Prasanna
Comparing efficiency of an attention-based deep learning network with contemporary radiological workflow for pulmonary embolism detection on CTPA: A retrospective study
European Journal of Radiology Open
Pulmonary Embolism
Deep learning
Neural Network
Computed Tomography Pulmonary Angiogram (CTPA)
Medical images
Radiology
title Comparing efficiency of an attention-based deep learning network with contemporary radiological workflow for pulmonary embolism detection on CTPA: A retrospective study
title_full Comparing efficiency of an attention-based deep learning network with contemporary radiological workflow for pulmonary embolism detection on CTPA: A retrospective study
title_fullStr Comparing efficiency of an attention-based deep learning network with contemporary radiological workflow for pulmonary embolism detection on CTPA: A retrospective study
title_full_unstemmed Comparing efficiency of an attention-based deep learning network with contemporary radiological workflow for pulmonary embolism detection on CTPA: A retrospective study
title_short Comparing efficiency of an attention-based deep learning network with contemporary radiological workflow for pulmonary embolism detection on CTPA: A retrospective study
title_sort comparing efficiency of an attention based deep learning network with contemporary radiological workflow for pulmonary embolism detection on ctpa a retrospective study
topic Pulmonary Embolism
Deep learning
Neural Network
Computed Tomography Pulmonary Angiogram (CTPA)
Medical images
Radiology
url http://www.sciencedirect.com/science/article/pii/S2352047725000243
work_keys_str_mv AT gagandeepsingh comparingefficiencyofanattentionbaseddeeplearningnetworkwithcontemporaryradiologicalworkflowforpulmonaryembolismdetectiononctpaaretrospectivestudy
AT anniesingh comparingefficiencyofanattentionbaseddeeplearningnetworkwithcontemporaryradiologicalworkflowforpulmonaryembolismdetectiononctpaaretrospectivestudy
AT tejasvikainth comparingefficiencyofanattentionbaseddeeplearningnetworkwithcontemporaryradiologicalworkflowforpulmonaryembolismdetectiononctpaaretrospectivestudy
AT sudhirsuman comparingefficiencyofanattentionbaseddeeplearningnetworkwithcontemporaryradiologicalworkflowforpulmonaryembolismdetectiononctpaaretrospectivestudy
AT nicolesakla comparingefficiencyofanattentionbaseddeeplearningnetworkwithcontemporaryradiologicalworkflowforpulmonaryembolismdetectiononctpaaretrospectivestudy
AT lukepartyka comparingefficiencyofanattentionbaseddeeplearningnetworkwithcontemporaryradiologicalworkflowforpulmonaryembolismdetectiononctpaaretrospectivestudy
AT tejphatak comparingefficiencyofanattentionbaseddeeplearningnetworkwithcontemporaryradiologicalworkflowforpulmonaryembolismdetectiononctpaaretrospectivestudy
AT prateekprasanna comparingefficiencyofanattentionbaseddeeplearningnetworkwithcontemporaryradiologicalworkflowforpulmonaryembolismdetectiononctpaaretrospectivestudy