Accuracy Improvement in Detection of COVID-19 in Chest Radiography

Abstract

From late 2019 to early 2020, the coronavirus outbreak affected 213 countries and territories around the world. This respiratory virus seriously affects human lung functionality. One way to diagnose this illness and find out if the lungs are infected is to evaluate chest X-ray. The evaluation of X-rays is challenging because corona has minor effects on the lungs in the early stages, and other diseases can have a similar effect. In this condition, Computer-Aided Diagnosis (CAD) can make a huge contribution and help decision support for healthcare professionals. Deep learning has obtained great results in data analysis recently, but the requirement for a large amount of training data prevents the use deep learning in medical data analysis since it is difficult to obtain a large amount of data from the medical field. This paper proposes an effective deep transfer learning-based model that improves current state-of-the-art systems in COVID-19 detection in chest radiographs. The weights of the DesneNet121 and ResNet50 on the Imagenet have transferred as initial weights, and then the two models have been fine-tuned with a deep classifier with data augmentation to detect three classes of COVID-19, Viral Pneumonia and normal radiographs. The proposed models obtained 97.83% accuracy with minimal false-negative results on the only public available COVID-19 radiography dataset. The Image-Level Accuracy (ILA) of the results outperform the results of previous studies, together with sensitivity and recall performance. Moreover, the proposed methods are scalable, and can be expanded to cover the detection of other types of diseases in the future and be integrated with more CNNs to increase their generalization capabilities.

An Effective Deep Learning Approach Based On CNN to Predict COVID-19 Rapidly Using Chest Images

“…This algorithm can classified the input medical images as benign, malignant or normal patient with accuracy, specificity, sensitivity and AUC of 90.59%, 90.67%, 90.53%, and 0.906 ± 0.0227 respectively. XiaofeErik at el [13] worked with the 2D mammograms and 3D tomosynthesis images. Used CNN for classification.…”

Pneumonia Transfer Learning Deep Learning Model from Segmented X-rays

“…Nevertheless, it is a time-intense and biased procedure. Pneumonia is misdiagnosed as other pulmonary diseases of comparable radiologic forms [ 5 , 6 , 7 , 8 ]. This can lead to incorrect prognosis and deterioration of the patient case.…”

“…Large-size labelled databases and deep learning models yield correct X-ray diagnoses. Deep learning permits hierarchical feature extraction from sufficient training inputs [ 8 , 9 , 10 , 11 ]. The medical sector is also completely unlike other arenas as it has not satisfied the ambitions of humanity, while it engages a large proportion of countries’ budgets [ 12 ].…”