Review of machine and deep learning models and data sets for the diagnosis of chronic diseases
Sushma Balaji Adsul, Kishor K. BhoyarPurpose
The purpose of the study is to identify systemic challenges in current AI healthcare models and synthesize predictive methodologies, and identify gaps in medical data. Over the recent years, there has been a substantial increase in chronic diseases (CDs) due to environmental conditions and lifestyle of humans. Developing an early diagnosis tool and an efficient treatment is extremely difficult due to the complexity of the different disease mechanisms and underlying symptoms of the patient population. Recently, the artificial intelligence (AI) methods, including deep learning and machine learning, have demonstrated exceptional performance for improving the medical domain. Despite the promising results, the existing diagnosis or detection systems are facing challenges and pose certain constraints for accurate and timely disease diagnosis.
Design/methodology/approach
Consequently, this study presents a review of around 60 existing literatures for CD diagnosis based on machine learning and deep learning approaches. Specifically, the review explores the merits and challenges of the cutting-edge techniques available in CD diagnosis. In addition, this review provides a brief overview of different methods used in CD diagnosis, particularly heart disease, kidney disease and diabetes, highlighting the tools, challenges and opportunities. Focusing on the recent CD diagnosis methods, the review explores the research gaps, data sets and metrics of such methods, offering significant insights that can guide future improvements rather than providing new experimental results.
Findings
The research significance is highlighted through data set analysis and achievements of recent methods. Significantly, the review offers a systematic evaluation and comparison, identification of best practices and benchmarks, enhance credibility and understand the applicability in real-world applications.
Originality/value
Overall, the systematic review analyzes the cutting-edge CD diagnosis methods for clinical deployment based on benchmarking model performance and discusses the opportunities to improve the field of CD diagnosis.