A small smartwatch study shows uneven detection of known heart disease
Source check: Source checked. Not independent scientific confirmation.
Finding: A trained neural model classified weekly heart-rate patterns in a 12-person test set with known cardiovascular-disease labels. The selected contrastive-loss variant reported mean weekly sensitivity of 67% with a 47-point between-person standard deviation and specificity of 100%. Some affected participants were consistently missed; these results do not establish safe disease exclusion.
Technology: Neural curve classifier; weekly logistic aggregation
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- A trained two-layer neural classifier maps acceleration/deceleration heart-rate curves to disease labels; learned weekly aggregation and contrastive training enable the represented clinical-label test.
- Test and result
- The ME-TIME cohort contained 42 participants. Thirty were used for development and subject-separated tuning; the test comprised six reference participants and six with atrial fibrillation or heart failure. Sensitivity/specificity average per-person weekly scores, not independent weekly patients.
- Limitation / still unknown
- Reference and disease groups differ in age, BMI and comorbidity. Twelve test participants and correlated weeks limit precision and generalization. Variant comparisons are exploratory. Several discussion paragraphs conflict with Table3 or misinterpret high specificity as a rule-out guarantee; neither rule-out safety nor prevention benefit is represented. Prospective data collection does not establish a frozen prospective model evaluation.
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/PMC11566672/ · https://www.ncbi.nlm.nih.gov/research/bionlp/RESTful/pmcoa.cgi/BioC_json/PMC11566672/unicode
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: Official primary BioC original article sections: Curve classifier/weekly aggregation;ME-TIME cohort;subject-separated tuning;Table3 and clinical limitations; no laboratory-method validation.
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 1/5 · 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.