BactPNet: A Novel Automated Detection Approach for Bacterial Pneumonia Patients

Every year, a large number of people around the globe, particularly, children die due to pneumonia disease. Approximately, 1.2 million cases of pneumonia have been reported in children of age ranges from 1 to 5. Out of 1.2 million, 880,000 died in 2016. Therefore, pneumonia is considered a major cause of mortality among children, particularly, in South Asia as well as African countries. It is among the top ten causes of mortality in developed countries, namely, the UK, the USA, and other European countries. However, an early diagnosis and treatment can significantly minimize the death rates among children in those countries that have a high prevalence. The research community has worked to diagnose the patients using traditional and deep learning (DL)-based methods; however, the existing approaches have various limitations in terms of accurate detection of the patients. Therefore, to address the above problem, we have presented a novel DL-based framework, BactPNet, for the detection of bacterial pneumonia patients. Our approach has achieved an accuracy of 91.98%, precision is 90%, recall is 84%, and F1-score is 86%. The above results of our approach confirm that it can be utilized to enhance the diagnosis of pneumonia from the chest x-ray images. By adopting the BactPNet, the quality of treatment and correct prediction can further be improved. More specifically, experimental findings and comparative assessment with other techniques show that BactPNet can better detect pneumonia patients and can be adopted by medical experts in hospitals.