2024 study: screening from ultrasound clips ↗
We tested how well AI could classify ventricular function from smartphone-recorded ultrasound videos. Three-category classification was correct in 177 of 184 test clips, or 96.2%, with a clear limitation in the Mildly Reduced EF group.
Study summary • peer review not verifiedRayong study: learning from 218 participants ↗
Moving from clip testing into a care setting, the Rayong Hospital study compared EasyEF outputs with clinician assessments during December 2025–January 2026, examining both overall and subgroup performance.
Evidence-reading guideUnderstanding EasyEF research results ↗
Accuracy is a starting point for understanding a system. We also need to know who was studied, what data was used, and how performance varies by group. This guide connects the numbers with their meaning for technology evaluation.
12-month project planWhat comes next for EasyEF ↗
The next step is to connect the app, processing, and reporting, then evaluate them across different care settings. The 12-month plan targets technology readiness from TRL 7 to TRL 9 through development, evaluation in at least five hospitals, and preparation of medical-software documentation.
01
Two studies, two perspectives
The first study evaluated classification of quality-screened clips. The Rayong study compared outputs with clinicians in participants with different assessment indications. Keeping these settings distinct shows what has actually been evaluated.
| Evidence | Unit and size | Overall result | Material limitation |
|---|---|---|---|
| 2024 publication | 184 test clips; 739 training clips | 177/184 = 96.2% (three categories) | Mildly Reduced EF: 12/17 ≈ 71% |
| Rayong study summary | 218 participants | 93.6% | fair LV: 27.3%; peer review not verified |
| Multisite evaluation plan | Target of at least five hospitals | No completed result reported | Planned in the project brief |
02
A foundation of 184 test clips
The published 2024 study included 923 clips: 739 for training and 184 for testing. Three-category classification was correct for 177 test clips, or 96.2%.
Mildly Reduced EF was correct in 12 of 17 clips, approximately 71%. This highlights why overall accuracy needs to be read alongside category-level performance.
References and further reading
03
Evaluation in the Rayong care setting
A retrospective Rayong Hospital study covered 218 participants in December 2025–January 2026, including pre-chemotherapy and general cardiac-function assessment. Overall accuracy was 93.6%.
Fair LV accuracy was 27.3%, a material limitation. Full publication and peer-review status for this study remain unverified.
References and further reading
04
The intermediate category needs more work
Both studies show a challenge in the intermediate category between reduced and better-preserved function. Errors in this group matter when interpreting and using AI output.
Further research needs to focus on this group, image quality, and reference assessment, with clear reporting of subgroup results.
References and further reading
05
The next questions are in practice
We plan evaluation in at least five hospitals to study variation in users, devices, and populations, alongside how the app fits care workflows.
Practical use raises questions beyond accuracy: time, workload, interpretation, and deployment feasibility. These are part of the next development phase.
References and further reading
SOURCE NOTES
Sources and published project summaries
- 2024 published study — original journal article ↗
- EasyEF team: Rayong study report
- The cardiac-assessment journey — by the EasyEF team
- EasyEF team: project scope and development roadmap
Project content is compiled and published by the EasyEF team from development records, presentations and study reports. Internal links lead to the team’s summaries, not additional independent evidence. Published research links to its original journal; study status and limitations are stated in the relevant articles.
Explore the source library →