What question did this study set out to answer?

This research aims to develop and validate a multi-modal Explainable AI (XAI) framework for critical healthcare applications.

March 23, 2026Open Access

Explainable Artificial Intelligence for Critical Healthcare Applications

Puntos clave

This research aims to develop and validate a multi-modal Explainable AI (XAI) framework for critical healthcare applications.
Developed ClinXAI integrating Grad-CAM, SHAP, and LIME for interpretable predictions.
Validated across three domains: sepsis prediction, diabetic retinopathy grading, and lung cancer detection.
Employed numerous datasets, including MIMIC-III ICU, EyePACS, and NLST for analysis.
Achieved AUC-ROC scores of 0.943, 0.961, and 0.938 for respective tasks.
Clinician trust rated at 4.31 out of 5.0, a 41.2% improvement over standard outputs.
Identified creatinine and lactate as key predictors for sepsis, aligning with clinical guidelines.

Resumen

The deployment of artificial intelligence in critical healthcare settings demands not only high predictive accuracy but also transparent, interpretabledecision-support mechanisms that clinicians can audit, trust, and act upon. Existing black-box deep learning models, despite strong performance onclinical benchmarks, lack the explanatory transparency required for regulatory compliance and clinical adoption. This study proposes a multi-modalExplainable AI (XAI) framework — termed ClinXAI — that integrates gradient-weighted class activation mapping (Grad-CAM), SHapley AdditiveexPlanations (SHAP), and Local Interpretable Model-agnostic Explanations (LIME) within a unified clinical decision pipeline. ClinXAI was validatedacross three critical healthcare domains: sepsis onset prediction (MIMIC-III ICU dataset, n = 52,847), diabetic retinopathy grading (EyePACS fundusimage dataset, n = 88,702), and early-stage lung cancer detection (NLST CT scan dataset, n = 26,314). The framework achieved AUC-ROC scoresof 0.943, 0.961, and 0.938 respectively, while clinician trust surveys (n = 84 specialists) rated explanation adequacy at 4.31 out of 5.0 — a 41.2%improvement over standard model outputs without explanations. Temporal feature attribution identified creatinine and lactate trajectories as the mostinfluential sepsis predictors, consistent with established clinical guidelines. ClinXAI provides a scalable, domain-agnostic architecture for transparentAI-assisted diagnosis in high-stakes healthcare environments.

Leer artículo completoexternamente

Me gusta

Guardar

Ver artículo completo