Contents
Academy · Research Data Analytics & AI · AI in medical imaging · Cardiovascular
Artificial Intelligence in Cardiovascular Imaging: Strength of the Published Evidence
Cardiovascular imaging provides the single best-validated published example of rigorously evaluated imaging artificial intelligence (AI), as well as a published example of the reporting gaps that are common elsewhere in the field.
Echocardiography, cardiac computed tomography (CT) and cardiac magnetic resonance (MR) are the areas of cardiology in which AI segmentation and measurement tools are most advanced. Two named studies illustrate both ends of the evidence spectrum.
Functions of AI tools in cardiovascular imaging
The established tasks are measurement rather than diagnosis: outlining the left ventricle on an echocardiogram to estimate ejection fraction, flagging wall-motion abnormalities for review by a sonographer, or segmenting coronary calcium on CT. These are quantification aids that operate within the existing workflow of a cardiologist or sonographer.
EchoNet-Dynamic as a well-validated example
Ouyang D, et al. (Nature. 2020;580:252–256) built a deep learning model that estimates ejection fraction beat-by-beat from standard echocardiogram video, validated it on a held-out Stanford test set (mean absolute error 4.1%, AUC 0.97) and on a separate external dataset from Cedars-Sinai (mean absolute error 6.0%, AUC 0.96), then compared its repeat-measurement variance prospectively with that of sonographers. This is the combination of external validation and prospective checking that the broader evidence standards for imaging AI ask for.
Reporting gaps in cardiac imaging AI studies
Kusunose K, et al. (Biomolecules. 2020;10(5):665) reported a convolutional neural network (CNN) for echocardiographic view classification with a strong headline accuracy (98.1%), but the paper also illustrates a common pattern of weakness: its ejection-fraction accuracy claim rested on a correlation comparison (r=0.80 vs. r=0.82) with no formal agreement analysis, its validation cohort of 189 patients came from the same institution rather than an external site, and the study reported inconsistent error figures for the same model across its own Results and Discussion sections. Authors citing any cardiac-imaging AI paper should look beyond the headline accuracy figure to the manner in which the validation was conducted.
Two cardiac-imaging AI papers, side by side
| EchoNet-Dynamic (Ouyang 2020) | Kusunose et al. 2020 | |
|---|---|---|
| External validation | Yes — separate Cedars-Sinai dataset | No; 189-patient cohort from the same institution |
| EF accuracy measure | Mean absolute error, reported directly | Correlation only (r=0.80 vs. r=0.82), no agreement analysis |
| Error reporting | Consistent across the paper | Inconsistent figures between Results and Discussion |
References
- Ouyang D, He B, Ghorbani A, et al. "Video-based AI for beat-to-beat assessment of cardiac function." read 8 October 2026. Nature. 2020;580:252–256
- Kusunose K, Haga A, Inoue M, et al. "Clinically Feasible and Accurate View Classification of Echocardiographic Images Using Deep Learning." read 8 October 2026. Biomolecules. 2020;10(5):665
Each source carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.
How to cite this article
Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.
Pubrica Academy. (2026, October 8). Artificial Intelligence in Cardiovascular Imaging: Strength of the Published Evidence. Pubrica. https://pubrica.com/academy/medical-writing/artificial-intelligence-in-cardiovascular-imaging/
Pubrica Academy. “Artificial Intelligence in Cardiovascular Imaging: Strength of the Published Evidence.” Pubrica, 8 Oct. 2026, pubrica.com/academy/medical-writing/artificial-intelligence-in-cardiovascular-imaging/.
Pubrica Academy (2026) Artificial Intelligence in Cardiovascular Imaging: Strength of the Published Evidence. Pubrica. Available at: https://pubrica.com/academy/medical-writing/artificial-intelligence-in-cardiovascular-imaging/ (Accessed: 15 October 2026).
Pubrica Academy. “Artificial Intelligence in Cardiovascular Imaging: Strength of the Published Evidence.” Pubrica. October 8, 2026. https://pubrica.com/academy/medical-writing/artificial-intelligence-in-cardiovascular-imaging/
Pubrica Academy, ‘Artificial Intelligence in Cardiovascular Imaging: Strength of the Published Evidence’ (Pubrica, 8 October 2026) <https://pubrica.com/academy/medical-writing/artificial-intelligence-in-cardiovascular-imaging/> accessed 15 October 2026.
Pubrica Academy. Artificial Intelligence in Cardiovascular Imaging: Strength of the Published Evidence. [Internet]. Pubrica; 2026 Oct 8 [cited 2026 Oct 15]. Available from: https://pubrica.com/academy/medical-writing/artificial-intelligence-in-cardiovascular-imaging/
Pubrica Academy, “Artificial Intelligence in Cardiovascular Imaging: Strength of the Published Evidence,” Pubrica, Oct. 8, 2026. [Online]. Available: https://pubrica.com/academy/medical-writing/artificial-intelligence-in-cardiovascular-imaging/
@misc{artificial_intelligence_in_cardiovascu_2026,
author = {{Pubrica Academy}},
title = {Artificial Intelligence in Cardiovascular Imaging: Strength of the Published Evidence},
howpublished = {Pubrica},
year = {2026},
month = {oct},
url = {https://pubrica.com/academy/medical-writing/artificial-intelligence-in-cardiovascular-imaging/},
note = {Accessed: <span data-cite-acc="iso">2026-10-15</span>}
}This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.
Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.
How Pubrica can help
Pubrica assists cardiovascular researchers in assessing the validation evidence for an imaging AI tool and in reporting a new study against the TRIPOD+AI or STARD-AI standard, as the claim requires.