01
Publication and research question
Published in 2024 in the Journal of Prapokklao Hospital Clinical Medical Education Center, 41(2):123–132, the article is titled Assessing the Efficacy of Artificial Intelligence in Left Ventricular Function Screening Using Parasternal Long Axis View Cardiac Ultrasound Video Clips. It asks how well AI classifies cardiac function from application-recorded clips against a reference assessment.
The result should be understood as an evaluation of image classification rather than a treatment-outcome study. Whatever design terminology is used in the article, the unit underlying the principal accuracy result is the test video clip.
Source: 2024 paper, pp. 123–124, 127
02
Dataset and split
The paper dates video collection to 1 May–31 July 2023. Of 1,336 submitted clips, 923 remained after screening. Included clips were five-second parasternal long-axis recordings with a reference LVEF from a cardiologist report.
The system used 739 clips for training and 184 for testing. Training performance must not substitute for test performance, and clip counts must not be relabeled as patient counts without evidence linking clips to unique individuals.
Source: 2024 paper, pp. 127–130
| Stage | Count | Meaning |
|---|---|---|
| Submitted videos | 1,336 clips | Before screening |
| Included dataset | 923 clips | After screening |
| Training | 739 clips | Data used for learning |
| Testing | 184 clips | Denominator of the 96.2% overall result |
03
What the included data covered
Exclusions included valvular abnormalities, pericardial disease, congenital heart disease, atrial fibrillation, and unsuitable clips, such as excessive motion, insufficient duration, poor lighting, or incomplete ultrasound framing. These criteria limit the settings to which the accuracy result can be applied.
Good performance on included data does not establish performance in every condition or image-quality setting. Reference LVEF came from clinician reports, with the paper accepting the measurement method used in those reports. Variation in reference assessment is therefore relevant to future evaluation design.
Source: 2024 paper, p. 128
04
From video preparation to three-category classification
The published system includes quality screening, preprocessing, neural-network classification, and postprocessing. It first classifies five categories, then consolidates them into Reduced EF, Mildly Reduced EF, and Preserved LV for the three-category result.
The article reports 68.5% accuracy for five categories and 96.2% for three. Changing the number of categories changes the evaluation task, so 96.2% does not describe equally strong performance across all five finer categories.
Source: 2024 paper, pp. 128–131
05
The main result and its limitation
The reported overall result is 177 correct classifications among 184 clips, or 96.2%. Mildly Reduced EF was correct in 12 of 17 clips, approximately 71%. This intermediate category had a limited test sample and more errors than the headline figure alone suggests.
Preserved LV details are inconsistent between the Thai and English abstracts, so this page does not present that subgroup statistic as a confirmed fact. Readers should consult the original article and seek author clarification before reusing the disputed detail.
Source: 2024 paper, pp. 123–124, 130–132
| Measure | Correct / tested | Reported value |
|---|---|---|
| Overall three-category accuracy | 177 / 184 clips | 96.2% |
| Mildly Reduced EF | 12 / 17 clips | Approximately 71% |
06
What this study supports and what remains
The study supports the feasibility of screening from smartphone recordings of ultrasound displays and explicitly identifies further work on the intermediate category. Its future-research discussion includes data from other hospitals and better explanations of AI outputs for users.
This study does not establish that EasyEF replaces clinicians, reduces mortality, or has SaMD approval. The later multisite evaluation goal is a further evidence-building step whose results must be reported separately from the 2024 work.
Source: 2024 paper, pp. 131–132; EasyEF project brief, 22 June 2026
SOURCE NOTES
Sources for this page
- บทความวิจัยปี 2567 / 2024 original article — J Prapokklao Hosp Clin Med Educat Center 41(2):123–132 ↗
- ข้อเสนอโครงการ EasyEF ฉบับ 22 มิถุนายน 2569: ขอบเขตและแผนงาน / EasyEF project brief, 22 June 2026: scope and planned work
Presentations reflect information at the time they were prepared. Explore context and evidence status in the resource library.
Explore the source library →