A person who wants to check for any signs of lung diseases will generally have to book an appointment at the x-ray lab followed by the doctor and then based on the doctors judgement of the X-ray know the condition of his/her health. What if there was an intermediate step which could tell the person whether an appointment with the doctor is necessary. This is made possible by our project.
What it does
In our project the x ray of the patients lung is provided to a ML Based Algorithm which is trained over a set of images having a normal lung and set of lungs displaying signs of pneumonia.
How we built it
We used CNN to build train the machine. We used a data set from Kaggle which contained the test cases as well as the training data.
Challenges we ran into
The trained machine could either predict the lungs with pneumonia with an almost perfect accuracy or could predict the lungs without pneumonia with an almost perfect accuracy. It was a challenge to get these two in a perfect sync. It was a challenge to get the front end working too.
Accomplishments that we're proud of
We got an accuracy of 78% while training the machine for both pneumonia lungs and normal lungs. While training the machine just for the pneumonia lungs, we got an accuracy of 97%.
What we learned
We learnt a lot about CNN as well as designing a front end using Python.
What's next for Pneumonia detection using x-ray of lungs
As there are types of pneumonia, we can train the machine to detect the types. We can also try to train the machine in such a way such that we can increase the accuracy of both pneumonia and normal lungs.