Adaptive Neuro-Fuzzy Inference System (ANFIS) for rapid diagnosis of COVID-19 cases based on routine blood tests

This article presents an Adaptive Neuro-Fuzzy Inference System (ANFIS) approach to rapidly detect COVID-19 cases using commonly available laboratory blood tests. Current Reverse transcription-polymerase chain reaction (RT-PCR) tests for COVID-19 suffer from several limitations including false-negative results as large as 1520%, the need for certified laboratories, expensive equipment, and trained personnel; hence the development of an efficient diagnosis system that provides prompt and accurate results is of great importance to control the spread of the virus. Therefore, it was aimed to develop an intelligent system to analyze blood tests and identify significant hematological indicators to support COVID-19 diagnosis. This study interpreted the ANFIS model performance by shapely values to identify the most important and decisive parameters that could assist clinicians in making effective patient management decisions. The findings of this study revealed that WBC (White blood cells) & Platelet counts can act as relevant and significant indicators for the diagnosis of COVID-19 patients. Moreover, the proposed ANFIS model achieved a high prediction accuracy as it was able to discriminate between positive and negative COVID-19 patients with an Accuracy, Sensitivity, and Specificity rates of 95%, 75%, and 97.25% respectively even though 10 % only of the data was positive. Therefore by combining available and low-cost blood test results to analysis based on the ANFIS model, we were able to provide an efficient and robust system to diagnose COVID-19.