From Accuracy to Interpretability: SHAP-Based Explainable Deep Learning for CHD Prediction in Type 2 Diabetes
International Journal of Drug Delivery Technology · 29.05.2026 · 16 (37s)
DOI: 10.25258/ijddt.16.37s.5 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Drug Delivery Technology, online erschienen am 29.05.2026. Verfasst von K. Sathyaseelan und 3 weiteren Autor:innen.
Themen: Diabetes mellitus
Abstract (Originalsprache)
Coronary Heart Disease (CHD) remains a lifethreatening complication for patients with Type 2 Diabetes Mellitus (T2DM), demanding prediction models that are not only accurate but also trustworthy and transparent. While deep learning has revolutionized medical risk prediction through exceptional performance, its opaque decision-making process often limits clinical confidence. This study introduces a SHAP-based explainable deep learning framework that transforms high-performing models into clinically interpretable decision-support tools for CHD prediction in T2DM patients. The proposed model leverages demographic, clinical, and biochemical data to deliver precise CHD risk estimates using a deep neural network. To illuminate the reasoning behind predictions, SHAP Additive explanations (SHAP) are employed, providing both global and individual-level insights into feature contributions. The framework successfully uncovers dominant risk drivers such as age, glycaemic control, blood pressure, lipid levels, body mass index, and duration of diabetes—findings that strongly align with established clinical evidence. By coupling predictive power with meaningful explanations, this approach empowers clinicians to move beyond blind reliance on algorithms toward informed, confident decision-making. The integration of explainability enhances transparency, accountability, and real-world usability, paving the way for personalized risk assessment and proactive intervention. Overall, this work highlights the transformative potential of explainable deep learning in delivering reliable, interpretable, and clinically impactful CHD prediction for improved T2DM care.
Zitieren
APA
Sathyaseelan, K., Allwin, T., Bavithra, S. A., & Miruthika, R. (2026). From Accuracy to Interpretability: SHAP-Based Explainable Deep Learning for CHD Prediction in Type 2 Diabetes. International Journal of Drug Delivery Technology, 16(37s). https://doi.org/10.25258/ijddt.16.37s.5
Vancouver
Sathyaseelan K, Allwin T, Bavithra SA, Miruthika R. From Accuracy to Interpretability: SHAP-Based Explainable Deep Learning for CHD Prediction in Type 2 Diabetes. International Journal of Drug Delivery Technology. 2026;16(37s). doi:10.25258/ijddt.16.37s.5
BibTeX
@article{Sathyaseelan2026,
author = {Sathyaseelan, K. and Allwin, T. and Bavithra, S. Anu and Miruthika, R.},
title = {From Accuracy to Interpretability: SHAP-Based Explainable Deep Learning for CHD Prediction in Type 2 Diabetes},
journal = {International Journal of Drug Delivery Technology},
year = {2026},
volume = {16},
number = {37s},
doi = {10.25258/ijddt.16.37s.5}
}
EUVIMED