Measurement ( IF 3.364 ) Pub Date : 2021-02-16 , DOI: 10.1016/j.measurement.2021.109153 S.S. Sai Koushik; K.G. Srinivasa
Recent developments in the field of machine learning have led to drastic improvements in medical diagnosis. Identification of different medical conditions with high accuracy is possible through machine learning, specifically deep learning. Convolutional Neural Networks are a subset of deep neural networks, used in investigating visual images.
In this study, a method to identify bacterial pneumonia, viral pneumonia and COVID-19 from chest X-rays is proposed using convolutional neural networks. Training accuracy of 0.9440 and validation accuracy of 0.9356 was obtained using this model. The test accuracy was found to be 0.8753. As a matter of fact, COVID-19 diagnosing precision and recall of the proposed method are 0.95 and 1.00 respectively. Significant improvements are seen when compared to other approaches.