Label-efficient deep semantic segmentation of intracranial hemorrhages in CT-scans
Intracranial hemorrhage (ICH) is a common finding in traumatic brain injury (TBI) and computed tomography (CT) is considered the gold standard for diagnosis. Automated detection of ICH provides clinical value in diagnostics and in the ability to feed robust quantification measures into future predic...
Main Authors: | , , , |
---|---|
Format: | Article |
Language: | English |
Published: |
Frontiers Media S.A.
2023-07-01
|
Series: | Frontiers in Neuroimaging |
Subjects: | |
Online Access: | https://www.frontiersin.org/articles/10.3389/fnimg.2023.1157565/full |
_version_ | 1797772811381506048 |
---|---|
author | Antoine Spahr Antoine Spahr Antoine Spahr Jennifer Ståhle Jennifer Ståhle Chunliang Wang Magnus Kaijser Magnus Kaijser |
author_facet | Antoine Spahr Antoine Spahr Antoine Spahr Jennifer Ståhle Jennifer Ståhle Chunliang Wang Magnus Kaijser Magnus Kaijser |
author_sort | Antoine Spahr |
collection | DOAJ |
description | Intracranial hemorrhage (ICH) is a common finding in traumatic brain injury (TBI) and computed tomography (CT) is considered the gold standard for diagnosis. Automated detection of ICH provides clinical value in diagnostics and in the ability to feed robust quantification measures into future prediction models. Several studies have explored ICH detection and segmentation but the research process is somewhat hindered due to a lack of open large and labeled datasets, making validation and comparison almost impossible. The complexity of the task is further challenged by the heterogeneity of ICH patterns, requiring a large number of labeled data to train robust and reliable models. Consequently, due to the labeling cost, there is a need for label-efficient algorithms that can exploit easily available unlabeled or weakly-labeled data. Our aims for this study were to evaluate whether transfer learning can improve ICH segmentation performance and to compare a variety of transfer learning approaches that harness unlabeled and weakly-labeled data. Three self-supervised and three weakly-supervised transfer learning approaches were explored. To be used in our comparisons, we also manually labeled a dataset of 51 CT scans. We demonstrate that transfer learning improves ICH segmentation performance on both datasets. Unlike most studies on ICH segmentation our work relies exclusively on publicly available datasets, allowing for easy comparison of performances in future studies. To further promote comparison between studies, we also present a new public dataset of ICH-labeled CT scans, Seq-CQ500. |
first_indexed | 2024-03-12T21:56:22Z |
format | Article |
id | doaj.art-b5e7684b33e948e3a6c9756e1de5584b |
institution | Directory Open Access Journal |
issn | 2813-1193 |
language | English |
last_indexed | 2024-03-12T21:56:22Z |
publishDate | 2023-07-01 |
publisher | Frontiers Media S.A. |
record_format | Article |
series | Frontiers in Neuroimaging |
spelling | doaj.art-b5e7684b33e948e3a6c9756e1de5584b2023-07-25T17:05:15ZengFrontiers Media S.A.Frontiers in Neuroimaging2813-11932023-07-01210.3389/fnimg.2023.11575651157565Label-efficient deep semantic segmentation of intracranial hemorrhages in CT-scansAntoine Spahr0Antoine Spahr1Antoine Spahr2Jennifer Ståhle3Jennifer Ståhle4Chunliang Wang5Magnus Kaijser6Magnus Kaijser7Department of Biomedical Engineering and Health Systems, KTH Royal Institute of Technology, Stockholm, SwedenSignal Processing Laboratory (LTS5), École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, SwitzerlandCHUV—Lausanne University Hospital, Lausanne, SwitzerlandDepartment of Clinical Neuroscience, Karolinska Institutet, Solna, Stockholm, SwedenDepartment of Neuroradiology, Karolinska University Hospital, Stockholm, SwedenDepartment of Biomedical Engineering and Health Systems, KTH Royal Institute of Technology, Stockholm, SwedenDepartment of Neuroradiology, Karolinska University Hospital, Stockholm, SwedenInstitute of Environmental Medicine, Karolinska Institutet, Stockholm, SwedenIntracranial hemorrhage (ICH) is a common finding in traumatic brain injury (TBI) and computed tomography (CT) is considered the gold standard for diagnosis. Automated detection of ICH provides clinical value in diagnostics and in the ability to feed robust quantification measures into future prediction models. Several studies have explored ICH detection and segmentation but the research process is somewhat hindered due to a lack of open large and labeled datasets, making validation and comparison almost impossible. The complexity of the task is further challenged by the heterogeneity of ICH patterns, requiring a large number of labeled data to train robust and reliable models. Consequently, due to the labeling cost, there is a need for label-efficient algorithms that can exploit easily available unlabeled or weakly-labeled data. Our aims for this study were to evaluate whether transfer learning can improve ICH segmentation performance and to compare a variety of transfer learning approaches that harness unlabeled and weakly-labeled data. Three self-supervised and three weakly-supervised transfer learning approaches were explored. To be used in our comparisons, we also manually labeled a dataset of 51 CT scans. We demonstrate that transfer learning improves ICH segmentation performance on both datasets. Unlike most studies on ICH segmentation our work relies exclusively on publicly available datasets, allowing for easy comparison of performances in future studies. To further promote comparison between studies, we also present a new public dataset of ICH-labeled CT scans, Seq-CQ500.https://www.frontiersin.org/articles/10.3389/fnimg.2023.1157565/fullcomputer visiontransfer learningICH segmentationtraumatic brain injurydatasetcomputed tomography |
spellingShingle | Antoine Spahr Antoine Spahr Antoine Spahr Jennifer Ståhle Jennifer Ståhle Chunliang Wang Magnus Kaijser Magnus Kaijser Label-efficient deep semantic segmentation of intracranial hemorrhages in CT-scans Frontiers in Neuroimaging computer vision transfer learning ICH segmentation traumatic brain injury dataset computed tomography |
title | Label-efficient deep semantic segmentation of intracranial hemorrhages in CT-scans |
title_full | Label-efficient deep semantic segmentation of intracranial hemorrhages in CT-scans |
title_fullStr | Label-efficient deep semantic segmentation of intracranial hemorrhages in CT-scans |
title_full_unstemmed | Label-efficient deep semantic segmentation of intracranial hemorrhages in CT-scans |
title_short | Label-efficient deep semantic segmentation of intracranial hemorrhages in CT-scans |
title_sort | label efficient deep semantic segmentation of intracranial hemorrhages in ct scans |
topic | computer vision transfer learning ICH segmentation traumatic brain injury dataset computed tomography |
url | https://www.frontiersin.org/articles/10.3389/fnimg.2023.1157565/full |
work_keys_str_mv | AT antoinespahr labelefficientdeepsemanticsegmentationofintracranialhemorrhagesinctscans AT antoinespahr labelefficientdeepsemanticsegmentationofintracranialhemorrhagesinctscans AT antoinespahr labelefficientdeepsemanticsegmentationofintracranialhemorrhagesinctscans AT jenniferstahle labelefficientdeepsemanticsegmentationofintracranialhemorrhagesinctscans AT jenniferstahle labelefficientdeepsemanticsegmentationofintracranialhemorrhagesinctscans AT chunliangwang labelefficientdeepsemanticsegmentationofintracranialhemorrhagesinctscans AT magnuskaijser labelefficientdeepsemanticsegmentationofintracranialhemorrhagesinctscans AT magnuskaijser labelefficientdeepsemanticsegmentationofintracranialhemorrhagesinctscans |