MXRFCS : Development of an enhanced multidomain X-ray feature representation model for the detection and classification of COVID severity levels

Abstract

Global healthcare has been under stress due to the COVID-19 pandemic, necessitating effective methods to determine COVID-19 patient severity and distribute resources accordingly. Current COVID-19 X-ray models are either too complicated or too ineffective to employ in real time. This paper introduces a novel method for assessing COVID-19 severity using an enhanced multimodal X-ray model. A strong patient representation is produced by combining demographics, clinical information, and X-ray results. Multidomain features, including frequency, Gabor, Wavelet, and entropy components, are produced from this data. Using this model, a customized neural network is trained to predict COVID-19 severity levels. A dataset of X-ray pictures and clinical information from COVID-19 patients were used to evaluate the model. The suggested model achieves 98.5% accuracy on several datasets, outperforming existing approaches in the identification of COVID-19 severity levels, according to the results. Performance shows promise in terms of accuracy, memory, and latency, supporting early and efficient management of severe COVID-19 cases and thwarting the pandemic without clinical usage.