An Innovative AI-Enabled System for COPD Exacerbation Diagnosis Using ClinicalBERT

Abstract

Aim: This work presents a diagnostic model based on ClinicalBERT for predicting COPD exacerbations using clinical texts and compares its performance with a Random Forest approach. Materials and Methods: In this study, Group 1 refers to Random Forest, where Group 2 refers to ClinicalBERT and both are examined with 2644 samples. The Statistical power is 80 %, significance threshold is 0.05 and a confidence interval of 95 % and evaluated using parameters. Result: ClinicalBERT has significantly higher performance than Random Forest. Accuracy of the ClinicalBERT is 96.17% and Random Forest is 92.89%. In COPD diagnosis, the ClinicalBERT has produced the highest accuracy with a statistical significance of 0.0056. Conclusion: In this work, it is noted that ClinicalBERT has higher accuracy compared to Random Forest in exacerbation prediction.