Foundation versus domain-specific models for left ventricular segmentation on cardiac ultrasound

doi:10.1038/s41746-025-01730-y

. 2025 Jun 6;8(1):341.

doi: 10.1038/s41746-025-01730-y.

Foundation versus domain-specific models for left ventricular segmentation on cardiac ultrasound

Chieh-Ju Chao^{1

2}, Yunqi Richard Gu², Wasan Kumar², Tiange Xiang², Lalith Appari^{3

4}, Justin Wu², Juan M Farina⁵, Rachael Wraith⁵, Jiwoon Jeong^{3

4}, Reza Arsanjani⁵, Garvan C Kane¹, Jae K Oh¹, Curtis P Langlotz^{6

7}, Imon Banerjee^{3

4}, Li Fei-Fei^{8

9}, Ehsan Adeli^{10

11

12}

Affiliations

¹ Department of Cardiovascular Medicine, Mayo Clinic Rochester, Rochester, MN, USA.
² Stanford Institute for Human-Centered Artificial Intelligence, Stanford University, Stanford, CA, USA.
³ Department of Radiology, Mayo Clinic Arizona, Scottsdale, AZ, USA.
⁴ School of Computing and Augmented Intelligence, Arizona State University, Phoenix, AZ, USA.
⁵ Department of Cardiovascular Medicine, Mayo Clinic Arizona, Scottsdale, AZ, USA.
⁶ Center of Artificial Intelligence in Medical Imaging, Stanford University, Stanford, CA, USA.
⁷ Department of Radiology, Stanford University, Stanford, CA, USA.
⁸ Stanford Institute for Human-Centered Artificial Intelligence, Stanford University, Stanford, CA, USA. feifeili@cs.stanford.edu.
⁹ Department of Computer Science, Stanford University, Stanford, CA, USA. feifeili@cs.stanford.edu.
¹⁰ Stanford Institute for Human-Centered Artificial Intelligence, Stanford University, Stanford, CA, USA. eadeli@stanford.edu.
¹¹ Department of Computer Science, Stanford University, Stanford, CA, USA. eadeli@stanford.edu.
¹² Department of Psychiatry and Behavioral Sciences, Stanford University, Stanford, CA, USA. eadeli@stanford.edu.

PMID: 40481190
PMCID: PMC12144204
DOI: 10.1038/s41746-025-01730-y

Foundation versus domain-specific models for left ventricular segmentation on cardiac ultrasound

Chieh-Ju Chao et al. NPJ Digit Med. 2025.

. 2025 Jun 6;8(1):341.

doi: 10.1038/s41746-025-01730-y.

Authors

Affiliations

¹ Department of Cardiovascular Medicine, Mayo Clinic Rochester, Rochester, MN, USA.
² Stanford Institute for Human-Centered Artificial Intelligence, Stanford University, Stanford, CA, USA.
³ Department of Radiology, Mayo Clinic Arizona, Scottsdale, AZ, USA.
⁴ School of Computing and Augmented Intelligence, Arizona State University, Phoenix, AZ, USA.
⁵ Department of Cardiovascular Medicine, Mayo Clinic Arizona, Scottsdale, AZ, USA.
⁶ Center of Artificial Intelligence in Medical Imaging, Stanford University, Stanford, CA, USA.
⁷ Department of Radiology, Stanford University, Stanford, CA, USA.
⁸ Stanford Institute for Human-Centered Artificial Intelligence, Stanford University, Stanford, CA, USA. feifeili@cs.stanford.edu.
⁹ Department of Computer Science, Stanford University, Stanford, CA, USA. feifeili@cs.stanford.edu.
¹⁰ Stanford Institute for Human-Centered Artificial Intelligence, Stanford University, Stanford, CA, USA. eadeli@stanford.edu.
¹¹ Department of Computer Science, Stanford University, Stanford, CA, USA. eadeli@stanford.edu.
¹² Department of Psychiatry and Behavioral Sciences, Stanford University, Stanford, CA, USA. eadeli@stanford.edu.

PMID: 40481190
PMCID: PMC12144204
DOI: 10.1038/s41746-025-01730-y

Abstract

The Segment Anything Model (SAM) was fine-tuned on the EchoNet-Dynamic dataset and evaluated on external transthoracic echocardiography (TTE) and Point-of-Care Ultrasound (POCUS) datasets from CAMUS (University Hospital of St Etienne) and Mayo Clinic (99 patients: 58 TTE, 41 POCUS). Fine-tuned SAM was superior or comparable to MedSAM. The fine-tuned SAM also outperformed EchoNet and U-Net models, demonstrating strong generalization, especially on apical 2-chamber (A2C) images (fine-tuned SAM vs. EchoNet: CAMUS-A2C: DSC 0.891 ± 0.040 vs. 0.752 ± 0.196, p < 0.0001) and POCUS (DSC 0.857 ± 0.047 vs. 0.667 ± 0.279, p < 0.0001). Additionally, SAM-enhanced workflow reduced annotation time by 50% (11.6 ± 4.5 sec vs. 5.7 ± 1.7 sec, p < 0.0001) while maintaining segmentation quality. We demonstrated an effective strategy for fine-tuning a vision foundation model for enhancing clinical workflow efficiency and supporting human-AI collaboration.

PubMed Disclaimer

Conflict of interest statement

Competing interests: The authors declare no competing interests.

Figures

**Fig. 1. Qualitative performance of fine-tuned SAM on representative cases against ground truth on the EchoNet-dynamic test dataset.**
From top to bottom: 97.5th to 2.5th percentile of DSC. Panels a, b, and c are end-diastolic frames, and Panels d, e, and f are end-systolic frames. We observed that many of the poor-performance cases had suboptimal image qualities, such as weak LV endocardial borders or off-axis views (Panels c and f, suggesting the importance of good input image quality on model performance. Additionally, end-diastolic frames usually have a better delineation of borders than end-systolic frames, which is consistent with the model performance (end-diastolic slightly better than end-systolic).

**Fig. 2. Zero-shot and fine-tuned SAM performance on a representative POCUS case.**
Panel a. end-diastolic frame, Panel b. end-systolic frame. From left to right are the ground truth, zero-shot, and fine-tuned mask, with an overlay of bounding boxes (green-colored) and mask (blue-colored), on the original POCUS image. Fine-tuned masks were more consistent with the anticipated left ventricular geometry on visualization. Note that POCUS images generally had worse quality compared to transthoracic echocardiography images.

**Fig. 3. SAM-assisted workflow decreased the time for annotation while maintaining the quality of echocardiography image segmentation.**
a With SAM’s assistance, the average annotation time significantly decreased by 50%. b Among experienced annotators, the workflow decreased annotation time by 39.2%; (c) It also reduced 66.0% annotation time for an inexperienced annotator. d–f No significant difference was observed in segmentation quality when measured by the Dice Similarity Score. ***p < 0.0001.

See this image and copyright information in PMC

References

1. Antoine, C. et al. Clinical outcome of degenerative mitral regurgitation. Circulation138, 1317–1326 (2018). - PubMed
1. Matulevicius, S. A. et al. Appropriate use and clinical impact of transthoracic echocardiography. JAMA Intern. Med.173, 1600–1607 (2013). - PubMed
1. Ouyang, D. et al. Video-based AI for beat-to-beat assessment of cardiac function. Nature580, 252–256 (2020). - PMC - PubMed
1. Lang, R. M. et al. Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American society of echocardiography and the European association of cardiovascular imaging. J. Am. Soc. Echocardiogr.28, 1–39.e14 (2015). - PubMed
1. Liu, J. et al. Contemporary role of echocardiography for clinical decision making in patients during and after cancer therapy. JACC Cardiovasc Imaging11, 1122–1131 (2018). - PubMed

LinkOut - more resources

Full Text Sources
- Nature Publishing Group
- PubMed Central

[1] Antoine, C. et al. Clinical outcome of degenerative mitral regurgitation. Circulation138, 1317–1326 (2018). - PubMed

[2] Antoine, C. et al. Clinical outcome of degenerative mitral regurgitation. Circulation138, 1317–1326 (2018). - PubMed

[3] Matulevicius, S. A. et al. Appropriate use and clinical impact of transthoracic echocardiography. JAMA Intern. Med.173, 1600–1607 (2013). - PubMed

[4] Matulevicius, S. A. et al. Appropriate use and clinical impact of transthoracic echocardiography. JAMA Intern. Med.173, 1600–1607 (2013). - PubMed

[5] Ouyang, D. et al. Video-based AI for beat-to-beat assessment of cardiac function. Nature580, 252–256 (2020). - PMC - PubMed

[6] Ouyang, D. et al. Video-based AI for beat-to-beat assessment of cardiac function. Nature580, 252–256 (2020). - PMC - PubMed

[7] Lang, R. M. et al. Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American society of echocardiography and the European association of cardiovascular imaging. J. Am. Soc. Echocardiogr.28, 1–39.e14 (2015). - PubMed

[8] Lang, R. M. et al. Recommendations for cardiac chamber quantification by echocardiography in adults: an update from the American society of echocardiography and the European association of cardiovascular imaging. J. Am. Soc. Echocardiogr.28, 1–39.e14 (2015). - PubMed

[9] Liu, J. et al. Contemporary role of echocardiography for clinical decision making in patients during and after cancer therapy. JACC Cardiovasc Imaging11, 1122–1131 (2018). - PubMed

[10] Liu, J. et al. Contemporary role of echocardiography for clinical decision making in patients during and after cancer therapy. JACC Cardiovasc Imaging11, 1122–1131 (2018). - PubMed

Save citation to file

Email citation

Add to Collections

Add to My Bibliography

Your saved search

Create a file for external citation management software

Your RSS Feed

Foundation versus domain-specific models for left ventricular segmentation on cardiac ultrasound

Affiliations

Foundation versus domain-specific models for left ventricular segmentation on cardiac ultrasound

Authors

Affiliations

Abstract

Conflict of interest statement

Figures

Similar articles

References

LinkOut - more resources

Full Text Sources

Abstract

Conflict of interest statement

Figures

Similar articles

References

Related information

LinkOut - more resources

Full Text Sources