Hybrid Approach for COVID-19 Severity Identification.

Research effort have been dedicated to the utilization of deep learning techniques for the detection of COVID-19 and various forms of pneumonia. Although these efforts have produced commendable results in terms of performance, a prevalent focus has been on classifying different pneumonia types, with limited attention paid to assessing infection severity. In this context, our study offers a multifaceted contribution. Firstly, we propose a hybrid framework that combines a Convolutional Neural Network (CNN) with a fuzzy system to identify infection severity. The CNN-based model comprises two stacked layers of CNN models designed to distinguish between normal cases and various pneumonia types. Concurrently, the fuzzy system processes numerical laboratory test data and the CNN model's classification outcomes to determine the degree of infection severity. Experimental findings from experiments conducted on a meticulously curated dataset using diverse CNN models showcase that the VGG16 model achieved an impressive 92%. specificity and 88% accuracy in distinguishing between viral and bacterial pneumonia. Furthermore, our fuzzy model underwent evaluation with twelve actual medical cases, revealing its capacity for accurate assessment of recommended severity levels among patients. This approach facilitates healthcare specialists in quantifying the severity of viral pneumonia through a scoring system, enhancing the diagnostic process.