This Project uses ARTNETS – Real time learning Deep neural networks (CNN + Linear Models) for predicting pneumonia from X-rays.
Detecting infection in early stages is crucial for curing pneumonia. Chest X-rays are currently the best available method for detecting this. However, using chest X-rays is a challenging task since it requires expert radiologists for making the decision. Also, statistics show that Pneumonia is most prevalent in South Asia and sub-Saharan Africa where there are only a few experts to classify chest X-ray for abnormalities.
Hence there is a lot of research in this area for building models that can automatically detect pneumonia from chest X-rays. State of the art models like CheXNet developed by Stanford ML group is making detections at a level exceeding practicing radiologists. One significant problem with the current solution is that it doesn’t leverage the vast amount of labeled data which is generated every day. Using this data could make models more robust and effective for pneumonia detection.
Current work explores ARTML methodology for building models that are scalable and giving the power of continual learning.