AI-Powered NLP Systems for Differential Diagnosis: Improving COPD Exacerbation Detection

Chronic Obstructive Pulmonary Disease (COPD) exacerbation is a life-threatening condition which manifests with overlapping symptoms that require early and correct differential diagnosis to ensure effective treatment. To this end, in this study, we propose a Natural Language Processing (NLP) enabled AI system to enhance COPD exacerbation identification and classification from EHRs and clinical notes. Based on large language models (LLMs) and AI based multi agent architectures, we automate data extraction and analysis of patients' data such as symptoms, patient's medical background and lifestyle factors. We perform unstructured clinical text processing for the cases of COPD exacerbation vs other respiratory conditions using multiple machine learning classifiers and NLP pipelines. It combines structured and unstructured data sources to increase the precision of conjunctive diagnostics and thus shorten the time needed for medical professionals to assess critical cases. In the study, different AI models are evaluated by random forest, logistic regression and Naïve Bayes classifiers, and results indicate that the random forest model is able to detect COPD exacerbation with the highest accuracy (80%) and F1-score (73%). This AI powered NLP system can automate the Medical History analysis and differential diagnosis process to be more efficient in diagnosis, reduce misclassification and improve patient outcomes. The model will be developed to include real-time patient monitoring and has been integrated with hospital information systems for clinical decision supports.