EUVIMEDEUVIMED entdecken
Crossref

Cardio-MTL: Multi-Task Cardiovascular Disease Prediction Using Large-Scale Multi-Disease Health Records

R. Auxilia Anitha Mary, T. Ramaprabha

International Journal of Drug Delivery Technology · 02.06.2026 · 16 (45s)

DOI: 10.25258/ijddt.16.45s.10 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, online erschienen am 02.06.2026. Verfasst von R. Auxilia Anitha Mary und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

The rapid expansion of national health surveillance systems, cloud-integrated medical repositories, and AI-driven chronic disease monitoring platforms has created unprecedented opportunities for population-scale cardiovascular analytics. However, existing predictive systems rely heavily on single-disease modeling, static feature pipelines, and imbalanced learning strategies that fail to capture cross-disease interactions, leading to reduced accuracy, poor minority-disease sensitivity, and limited clinical interpretability. To overcome these constraints, this work proposes CARDIO-MTL, a unified multi-task learning framework designed to jointly predict six major cardiovascular and metabolic conditions using large-scale BRFSS health records. The architecture integrates four complementary intelligence layers: UNI-ENCODE, which constructs survey-aware shared–private latent representations to stabilize heterogeneous inputs; HYBRID-MAP, an adaptive expert-routing engine combining deep neural, gradient-boosted, and Bayesian predictors; RISK-BAL, a prevalence-aware optimization layer addressing class imbalance and gradient conflict; and EXPLAIN-MED, a clinically grounded interpretability module generating attribution scores, comorbidity graphs, disease progression trajectories, and intervention priorities. Evaluated on more than 400,000 population samples, CARDIO-MTL achieves AUC values up to 0.98, average precision above 0.94, and consistent gains of 8–20% over single-task and traditional machine learning baselines. Ablation results confirm the essential contribution of each module, while confusion matrix analysis demonstrates low false-negative rates for life-critical diseases. CARDIO-MTL provides a scalable, interpretable, and population-ready intelligence layer for nextgeneration cardiovascular risk assessment.

Zitieren

APA

Mary, R. A. A., & Ramaprabha, T. (2026). Cardio-MTL: Multi-Task Cardiovascular Disease Prediction Using Large-Scale Multi-Disease Health Records. International Journal of Drug Delivery Technology, 16(45s). https://doi.org/10.25258/ijddt.16.45s.10

Vancouver

Mary RAA, Ramaprabha T. Cardio-MTL: Multi-Task Cardiovascular Disease Prediction Using Large-Scale Multi-Disease Health Records. International Journal of Drug Delivery Technology. 2026;16(45s). doi:10.25258/ijddt.16.45s.10

BibTeX
@article{Mary2026,
  author = {Mary, R. Auxilia Anitha and Ramaprabha, T.},
  title = {Cardio-MTL: Multi-Task Cardiovascular Disease Prediction Using Large-Scale Multi-Disease Health Records},
  journal = {International Journal of Drug Delivery Technology},
  year = {2026},
  volume = {16},
  number = {45s},
  doi = {10.25258/ijddt.16.45s.10}
}
Kostenlos starten →