Pneumonia detection using Transfer Learning

  • S.S.Saranya, Shubham Jha, Gangesh Bharadwaj


Pneumonia is a deadly disease and it accounts death of millions of people worldwide, its fatality rate is higher in children. Therefore we have introduced a mechanism which will detect the disease, and the tool provided by us is usable without paying a penny. The tool is based on the image processing and neural network mechanism which predicts the chances of a person suffering from pneumonia.There is always a high chance of person suffering from it who work in factories, resin labs or even the furnaces. The smoke produced from these places is inhaled by the workers, which in turn causes inflammation. Thus it is mandatory for them to have a regular checkup, while our tool serves well with its use as it can be used numerous times and at any instance.Chest Xrays are used for diagnosing pneumonia.This requires highly skilled radiologist.Thus an automated system for pneumonia detection is beneficial for people living in remote areas.Convolutional Neural Networks(CNNs) have been quite successful in disease identification.Also by using  pretrained CNN models on large scale datasets important features can be learned which can be used in image classification .In the following work, we make use of pretrained CNN models .These are used for extracting features.After this step has various classifying algorithms for differentiating pneumonia affected person and normal person chest radiographs.We use analytical techniques for finding best pretrained CNN model.Statistical analysis proves that the use of pretrained CNN models and various other supervised classification algorithms can be used to detect Pneumonia.Users are asked to submit their X-ray image on our too, website in this case and the prediction result follows further. The tool will also give suggestion to visit the nearby hospital when diagnosed with pneumonia. 

How to Cite
Gangesh Bharadwaj, S. S. J. (2020). Pneumonia detection using Transfer Learning. International Journal of Advanced Science and Technology, 29(3), 986 - 994. Retrieved from