AI Science Tracker
EN

Narayana Health

1 discovery from 2025

Discoveries

Narayana Health

ECG-image model identifies concurrent low ejection fraction in archived records

Source check: Source checked. Not independent scientific confirmation.

Finding: A DenseNet121 model learned to identify echocardiography-referenced ejection fraction below50% from ECG images. It reported ROC areas of0.92 on an internal ECG test set and0.88 at another healthcare centre. The study tests concurrent diagnostic-label discrimination, without establishing early warning or clinical benefit.

Technology: DenseNet121 convolutional neural network

Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.

AI contribution
A trained convolutional network maps cropped ECG trace images to a concurrent echocardiographic low-ejection-fraction label.
Test and result
Development used119,281 same-day ECG/echo pairs from104,697 people. ECGs were randomly split into107,147 training,2,020 validation and10,114 test records. Another centre supplied roughly24,300 pairs; the internal-test Youden threshold0.18 was carried to that cohort.
Limitation / still unknown
Random ECG-level splitting does not establish separation of repeated patients. The external denominator is24,323 in text versus24,319 in the table. The internal threshold was selected on test outcomes; the external data are archived, not a prospectively frozen workflow. Standardized device images and same-day selected imaging pairs limit generalization; no screening benefit or diagnosis of future dysfunction is established.
Evidence sources and annotation context

Evidence context (descriptions follow the cited sources; test type is a catalog classification):

Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC12223524/ · https://doi.org/10.1016/j.ihj.2025.03.009

Source checked · editorial details

The source page was checked against this entry. Independent confirmation is not assessed by this label.

Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.

Review date: 21 Nov 2026. Review scope: Original primary clinical/computational Methods, reference definitions, model evaluation and Results independently read; no operational laboratory sections.

No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.

Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.

Open the interactive profile for the timeline, or browse all organizations and model families.