Cine cardiac magnetic resonance segmentation using temporal-spatial adaptation of prompt-enabled segment-anything-model: a feasibility study

ABSTRACT: Background: We propose an approach to adapt a segmentation foundation model, segment-anything-model (SAM), for cine cardiovascular magnetic resonance (CMR) segmentation and evaluate its generalization performance on unseen datasets. Methods: We present our model, cineCMR-SAM, which introd...

Full description

Saved in:
Bibliographic Details
Main Authors: Zhennong Chen, Sekeun Kim, Hui Ren, Sunghwan Kim, Siyeop Yoon, Quanzheng Li, Xiang Li
Format: Article
Language:English
Published: Elsevier 2025-01-01
Series:Journal of Cardiovascular Magnetic Resonance
Subjects:
Online Access:http://www.sciencedirect.com/science/article/pii/S1097664725000717
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850124611565387776
author Zhennong Chen
Sekeun Kim
Hui Ren
Sunghwan Kim
Siyeop Yoon
Quanzheng Li
Xiang Li
author_facet Zhennong Chen
Sekeun Kim
Hui Ren
Sunghwan Kim
Siyeop Yoon
Quanzheng Li
Xiang Li
author_sort Zhennong Chen
collection DOAJ
description ABSTRACT: Background: We propose an approach to adapt a segmentation foundation model, segment-anything-model (SAM), for cine cardiovascular magnetic resonance (CMR) segmentation and evaluate its generalization performance on unseen datasets. Methods: We present our model, cineCMR-SAM, which introduces a temporal-spatial attention mechanism to produce segmentation across one cardiac cycle. We freeze the pre-trained SAM’s weights to leverage SAM’s generalizability while fine-tuning the rest of the model on two public cine CMR datasets. Our model also enables text prompts to specify the view type (short-axis or long-axis) of the input slices and box prompts to guide the segmentation region. We evaluated our model’s generalization performance on three external testing datasets including a public multi-center, multi-vendor testing dataset of 136 cases and 2 retrospectively collected in-house datasets from 2 different centers with specific pathologies: aortic stenosis (40 cases) and heart failure with preserved ejection fraction (HFpEF) (53 cases). Results: Our approach achieved superior generalization in both the public testing dataset (Dice for LV=0.94 and for myocardium=0.86) and two in-house datasets (Dice ≥0.90 for LV and ≥0.82 for myocardium) compared to existing CMR deep learning segmentation methods. Clinical parameters derived from automatic and manual segmentations showed a strong correlation (r ≥0.90). The use of both text prompts and box prompts enhanced the segmentation accuracy. Conclusion: cineCMR-SAM effectively adapts SAM for cine CMR segmentation, achieving high generalizability and superior accuracy on unseen datasets.
format Article
id doaj-art-cb532442e05a44839e09bdcf7c4b437e
institution OA Journals
issn 1097-6647
language English
publishDate 2025-01-01
publisher Elsevier
record_format Article
series Journal of Cardiovascular Magnetic Resonance
spelling doaj-art-cb532442e05a44839e09bdcf7c4b437e2025-08-20T02:34:16ZengElsevierJournal of Cardiovascular Magnetic Resonance1097-66472025-01-0127110190910.1016/j.jocmr.2025.101909Cine cardiac magnetic resonance segmentation using temporal-spatial adaptation of prompt-enabled segment-anything-model: a feasibility studyZhennong Chen0Sekeun Kim1Hui Ren2Sunghwan Kim3Siyeop Yoon4Quanzheng Li5Xiang Li6From Center for Advanced Medical Computing and Analysis, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USAFrom Center for Advanced Medical Computing and Analysis, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USAFrom Center for Advanced Medical Computing and Analysis, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USAFrom Center for Advanced Medical Computing and Analysis, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USAFrom Center for Advanced Medical Computing and Analysis, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USAFrom Center for Advanced Medical Computing and Analysis, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USACorresponding author.; From Center for Advanced Medical Computing and Analysis, Massachusetts General Hospital and Harvard Medical School, Boston, Massachusetts, USAABSTRACT: Background: We propose an approach to adapt a segmentation foundation model, segment-anything-model (SAM), for cine cardiovascular magnetic resonance (CMR) segmentation and evaluate its generalization performance on unseen datasets. Methods: We present our model, cineCMR-SAM, which introduces a temporal-spatial attention mechanism to produce segmentation across one cardiac cycle. We freeze the pre-trained SAM’s weights to leverage SAM’s generalizability while fine-tuning the rest of the model on two public cine CMR datasets. Our model also enables text prompts to specify the view type (short-axis or long-axis) of the input slices and box prompts to guide the segmentation region. We evaluated our model’s generalization performance on three external testing datasets including a public multi-center, multi-vendor testing dataset of 136 cases and 2 retrospectively collected in-house datasets from 2 different centers with specific pathologies: aortic stenosis (40 cases) and heart failure with preserved ejection fraction (HFpEF) (53 cases). Results: Our approach achieved superior generalization in both the public testing dataset (Dice for LV=0.94 and for myocardium=0.86) and two in-house datasets (Dice ≥0.90 for LV and ≥0.82 for myocardium) compared to existing CMR deep learning segmentation methods. Clinical parameters derived from automatic and manual segmentations showed a strong correlation (r ≥0.90). The use of both text prompts and box prompts enhanced the segmentation accuracy. Conclusion: cineCMR-SAM effectively adapts SAM for cine CMR segmentation, achieving high generalizability and superior accuracy on unseen datasets.http://www.sciencedirect.com/science/article/pii/S1097664725000717CMRSegmentationDeep learningSegment-anything-model
spellingShingle Zhennong Chen
Sekeun Kim
Hui Ren
Sunghwan Kim
Siyeop Yoon
Quanzheng Li
Xiang Li
Cine cardiac magnetic resonance segmentation using temporal-spatial adaptation of prompt-enabled segment-anything-model: a feasibility study
Journal of Cardiovascular Magnetic Resonance
CMR
Segmentation
Deep learning
Segment-anything-model
title Cine cardiac magnetic resonance segmentation using temporal-spatial adaptation of prompt-enabled segment-anything-model: a feasibility study
title_full Cine cardiac magnetic resonance segmentation using temporal-spatial adaptation of prompt-enabled segment-anything-model: a feasibility study
title_fullStr Cine cardiac magnetic resonance segmentation using temporal-spatial adaptation of prompt-enabled segment-anything-model: a feasibility study
title_full_unstemmed Cine cardiac magnetic resonance segmentation using temporal-spatial adaptation of prompt-enabled segment-anything-model: a feasibility study
title_short Cine cardiac magnetic resonance segmentation using temporal-spatial adaptation of prompt-enabled segment-anything-model: a feasibility study
title_sort cine cardiac magnetic resonance segmentation using temporal spatial adaptation of prompt enabled segment anything model a feasibility study
topic CMR
Segmentation
Deep learning
Segment-anything-model
url http://www.sciencedirect.com/science/article/pii/S1097664725000717
work_keys_str_mv AT zhennongchen cinecardiacmagneticresonancesegmentationusingtemporalspatialadaptationofpromptenabledsegmentanythingmodelafeasibilitystudy
AT sekeunkim cinecardiacmagneticresonancesegmentationusingtemporalspatialadaptationofpromptenabledsegmentanythingmodelafeasibilitystudy
AT huiren cinecardiacmagneticresonancesegmentationusingtemporalspatialadaptationofpromptenabledsegmentanythingmodelafeasibilitystudy
AT sunghwankim cinecardiacmagneticresonancesegmentationusingtemporalspatialadaptationofpromptenabledsegmentanythingmodelafeasibilitystudy
AT siyeopyoon cinecardiacmagneticresonancesegmentationusingtemporalspatialadaptationofpromptenabledsegmentanythingmodelafeasibilitystudy
AT quanzhengli cinecardiacmagneticresonancesegmentationusingtemporalspatialadaptationofpromptenabledsegmentanythingmodelafeasibilitystudy
AT xiangli cinecardiacmagneticresonancesegmentationusingtemporalspatialadaptationofpromptenabledsegmentanythingmodelafeasibilitystudy