DnnARs: An Artificial Intelligence Technique for Prediction of Antimicrobial Resistant Strains in E. coli Bacteria Causing Urine Tract Infection

The rise of antimicrobial resistance (AMR) in bacteria that cause infectious diseases poses a significant global health challenge, necessitating advanced prediction technology for effective management and response. This study aims to evaluate the predictive capabilities of several machine learning (ML) and deep learning (DL) models in identifying patterns of antibiotic resistance. We have proposed a comprehensive experimental framework that combines multiple machine learning algorithms and deep learning architectures to evaluate their efficacy and accuracy in predicting antibiotic resistance. We have developed DnnARs, an artificial intelligence pipeline designed to identify antibiotic-resistant strains of E. coli bacteria, which are mostly responsible for urinary tract infections (UTIs). DnnARs utilizes Synthetic Minority Over-sampling Technique (SMOTE) and applies feature engineering using principal component analysis (PCA) and t-distributed Stochastic neighbor embedding (t-SNE) on our dataset to evaluate the effectiveness of various AI models. Our model comparison demonstrates that DL models outperform ML models when the dataset is fine-tuned using SMOTE and feature engineering. Our DnnARs achieve a classification accuracy of 97%, a specificity of 91%, and a remarkable sensitivity of 100%. The DL models that were examined exhibit an average classification accuracy that is 3.66% higher than that of the ML models. The findings demonstrate the capability of deep learning methodology to enhance predictive analytics for antimicrobial resistance, offering a robust and efficient alternative to traditional machine learning techniques. This research aims to enhance diagnostic tools and medications, improve the management of infectious diseases, reduce antibiotic resistance, and integrate deep learning models into clinical workflows to enable immediate decision-making.