Dr. Chandrashekhar Uppin Profile Dr. Chandrashekhar Uppin

Enhancing viral pneumonia diagnosis accuracy using transfer learning and ensemble technique from chest x-ray images

  • Authors Details :  
  • Chandrashekar Uppin,  
  • Usman Bello Abubakar

201 Views Original Article

Pneumonia is an acute pulmonary infection that can be caused by bacteria, viruses, or fungi. It infects the lungs, causing inflammation of the air sacs and pleural effusion: a condition in which the lung is filled with fluid. The diagnosis of pneumonia is tasking as it requires a review of Chest X-ray (CXR) by specialists, laboratory tests, vital signs, and clinical history. Utilizing CXR is an important pneumonia diagnostic method for the evaluation of the airways, pulmonary parenchyma, and vessels, chest walls among others. It can also be used to show changes in the lungs caused by pneumonia. This study aims to employ transfer learning, and ensemble approach to help in the detection of viral pneumonia in chest radiographs. The transfer learning model used was Inception network, ResNet-50, and InceptionResNetv2. With the help of our research, we were able to show how well the ensemble technique, which uses InceptionResNetv2 and the utilization of the Non-local Means Denoising algorithm, works. By utilizing these techniques, we have significantly increased the accuracy of pneumonia classification, opening the door for better diagnostic abilities and patient care. For objective labeling, we obtained a selection of patient chest X-ray images. In this work, the model was assessed using state-of-the-art metrics such as accuracy, sensitivity, and specificity. From the statistical analysis and scikit learn python analysis, the accuracy of the ResNet-50 model was 84%, the accuracy of the inception model was 91% and lastly, the accuracy of the InceptionResNetv2 model was 96%.

Article Subject Details


Article Keywords Details



Article File

Full Text PDF





More Article by Dr. Chandrashekhar Uppin

A natural language processing approach to determine the polarity and subjectivity of iphone 12 twitter feeds using textblob

Sentiment analysis and opinion mining is a branch of computer science that has gained considerable growth over the last decade. this branch of computer science deals with determini...

A proactive approach to network forensics intrusion (denial of service flood attack) using dynamic features, selection and convolution neural network

Currently, the use of internet-connected applications for storage by different organizations have rapidly increased with the vast need to store data, cybercrimes are also increasin...

A comprehensive review for security analysis of iot platforms

Due to the rapid growth in the field of science and technology, iot (internet of things) has become emerging technique for connecting heterogeneous technologies related to our dail...

Smartphone based ischemic heart disease (heart attack) risk prediction using clinical data and data mining approaches

We designed a mobile application to deal with ischemic heart disease (ihd) (heart attack) an android based mobile application has been used for coordinating clinical information ta...