Hybrid Deep Learning Models for Effective COVID -19 Diagnosis with Chest X-Rays

Abstract

The survey on COVID-19 test kits RT-PCR (reverse transcription-polymerase chain reaction) concludes the hit rate of diagnosis and detection is degrading. Manufacturing these RT-PCR kits is very expensive and time-consuming. This work proposed an efficient way for COVID detection using a hybrid convolutional neural network (HCNN) through chest x-rays image analysis. It aids to differentiate non-COVID patient and COVID patients. It makes the medical practitioner to take appropriate treatment and measures. The results outperformed the custom blood and saliva-based RT-PCR test results. A few examinations were carried out over chest X-ray images utilizing ConvNets that produce better accuracy for the recognition of COVID-19. When considering the number of images in the database and the COVID discovery season (testing time = 0.03 s/image), the design reduced the computational expenditure. With mean ROC AUC scores 96.51 & 96.33%, the CNN with minimised convolutional and fully connected layers detects COVID-19 images inside the two-class COVID/Normal and COVID/Pneumonia orders.

Ähnliche Arbeiten