Fuzzy expert system for diagnosing diabetic neuropathy

Existing diagnostic methods for diabetic neuropathy rely on subjective assessments and lengthy evaluations. Prior research has shown that clinical questionnaires and biochemical markers are commonly used, but integrating these into a decision-support system remains a challenge. No prior work had resolved how to combine multiple diagnostic parameters into a single predictive model. This gap motivated the development of a system that could streamline diagnosis. Specialists often face time constraints, and manual evaluation of symptoms and lab results can be error-prone. The need for a more efficient and accurate method is evident. The Michigan questionnaire is widely used, but its interpretation varies. A computational approach could standardize this process. This paper's contribution is a novel fuzzy logic system designed to assist in diabetic neuropathy diagnosis. The system aims to reduce diagnostic delays and improve patient outcomes.

The goal was to create a diagnostic tool using fuzzy logic to identify and assess the severity of diabetic neuropathy. The system was intended to support both specialists and general practitioners in making faster and more accurate diagnoses. The motivation stemmed from the limitations of existing diagnostic methods. Diabetic neuropathy diagnosis is complex due to overlapping symptoms and variable test results. The system aimed to integrate multiple diagnostic parameters into a single framework. The authors proposed using fuzzy logic to handle uncertainty in symptom and lab data. The system was designed to categorize severity levels based on clinical and biochemical indicators. This approach could improve the consistency of diagnoses across different healthcare providers.

The study was divided into two phases: parameter selection and system testing. In the first phase, diagnostic parameters were identified through a literature review and expert consultation. Eight specialists contributed to defining the input variables. The selected parameters included diabetes duration, Michigan questionnaire scores, and biochemical markers. The second phase involved testing the system using 244 patient records from an endocrinology center. The records were collected during the first six months of 2014. The system's performance was evaluated using sensitivity, specificity, and accuracy metrics. The fuzzy logic model was implemented using ASP.Net. The output variable was the severity score, ranging from 0 to 10, divided into four categories.

The system achieved a sensitivity of 89%, indicating a high true positive rate. Specificity was measured at 98%, suggesting strong true negative detection. Overall accuracy reached 93%, combining both positive and negative results. The severity score was divided into four categories: absence, mild, moderate, and severe. The highest accuracy was observed in moderate and severe cases. The lowest accuracy was in mild cases, but still above 90%. The system correctly classified most patients based on clinical and biochemical data. The Michigan questionnaire scores were the most influential variables in determining severity.

The fuzzy expert system proposed in this study can assist healthcare providers in diagnosing diabetic neuropathy more efficiently. The system's high sensitivity and specificity suggest it is a reliable diagnostic aid. The authors propose that integrating such systems into clinical workflows could improve diagnostic consistency. The system's performance was validated using real patient data from an endocrinology center. The severity categories were shown to align with clinical expectations. The use of ASP.Net enabled a user-friendly interface for healthcare professionals. The study's findings suggest that computational models can enhance diagnostic accuracy. The system's design allows for future adaptations to include additional diagnostic parameters.