Design and Development of an Intelligent Medical System

Abstract

Electrocardiogram (ECG) interpretation plays a crucial role in the diagnosis of cardiovascular diseases. Recent advances in artificial intelligence (AI), particularly in machine learning, deep learning, and large language models (LLMs), have created new opportunities for developing intelligent ECG analysis systems. This study proposes an integrated framework for ECG inter pretation that combines feature optimization, semantic representation, multilabel classification, and automatic report generation using the PTB-XL and PTB-XL+ datasets. Traditional machine learning models, deep learning architectures, and biomedical language models were evaluated to investigate the effectiveness of different ECG representations for diagnostic classification, while a FLAN-T5-based generation module was adapted to produce structured clinical reports from textual ECGdescriptions. Experimental results showed that the proposed optimization and classification strategy achieved classification performance approaching 80% Macro-F1 using structured ECG information.In addition, biomedical language models demonstrated the abil ity to effectively exploit semantic ECG representations for diagnostic prediction, achieving a Macro-F1 score of 71%, while The report generation module also showed promising potential for producing coherent and clinically meaningful ECG interpretations. Overall, the obtained results highlight the potential of combining machine learning, biomedical language models, and generative AI techniques to supportaccurate, interpretable, andclinically relevant ECGanalysis. This work contributes toward the development of intelligent clinical decision-support systems capable of assisting ECG interpretation through both diagnostic prediction and textual explana tion.