,
Assistant Professor (Selection Grade), Department of Computing - Data Science, Coimbatore Institute of Technology, Coimbatore, Tamil Nadu, India India
,
Assistant Professor (Selection Grade), Department of Computing - Artificial Intelligence and Machine Learning, Coimbatore Institute of Technology, Coimbatore, Tamil Nadu, India India
,
Assistant Professor (Selection Grade), Department of Computing- Artificial Intelligence and Machine Learning, Coimbatore Institute of Technology, Coimbatore, Tamil Nadu, India India
Assistant Professor (Senior Grade), Department of Software Systems, Coimbatore Institute of Technology, Coimbatore, Tamil Nadu, India
Clinically-oriented AI-based diagnostic systems have exhibited impressive predictive accuracies; however, their poor interpretability capabilities and incapability to optimize their learning parameters in eal-time pose a major challenge towards implementing such solutions in the actual healthcare settings. To overcome such limitations, in this research, it presents an efficient Adaptive Butterfly Optimization-driven Explainable CNN-Transformer based Clinical Diagnosis Solution (ABO-ClinExplain). This proposed framework uses an innovative hybrid CNN-Transformer approach that incorporates an adaptive butterfly optimization algorithm for optimizing learning parameters, feature extraction weights, and classification criteria in order to deliver high diagnostic accuracies without losing explainability. Additionally, the presented solution uses two new techniques called Explainable Attention Mapping and Shapley Additive explanations to create visual and feature-level explanations of the clinical prediction process. The proposed approach is assessed on a large multimodal clinical database consisting of medical imaging reports, electronic health records, lab reports, and symptomatology data of patients across different diseases gathered through 22,500 patients' records. Results show that the proposed ABO-ClinExplain framework is capable of achieving diagnostic accuracies of 98.6%, 98.1% precision,
97.8% recall, and 97.9% F1 score compared to 90.4%, 85.1%, 89.4%, and 96.3% for CNN, ResNet-50, LSTM, and the baseline ClinExplain-AI models, respectively. The proposed method enhances the feature selection efficiency by 26.8%, while the false diagnosis is decreased by 28.4%. Moreover, the interpretability module achieves a rating of 95.7% and boosts clinicians' confidence levels by 43.6%. Computational assessments demonstrate the prediction lag at only 44 ms, indicating the possibility of implementing the developed technology in real-life scenarios. On the whole, the results indicate that the suggested method makes an effective combination of adaptive optimization, contextual intelligence, and
explainability, resulting in reliable decision support systems. The methodology described above can be deemed a method of developing ethical AI in healthcare.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.
