NOT FOR MEDICAL USE. This is a web based (but locally run) prototype system for diagnosing chest X-ray images.
The patient data remains on your computer and all computation occurs in your browser.
The goals of this system are:
Let people play with deep learning tools to know how they work and their limitations.
Show the potential of open data (needed to build a public system like this).
Create a tool to help teach radiology.
Demonstrate a model delivery system that can scale to provide free medical tools to the world.
The system contains three main components: Out Of Distribution error:
This is a heatmap showing how the image differs from our training data.
If the heatmap is too bright then the image is very different from our training data and the model will likely not work.
We will prevent an image from being processed if it is not similar enough to our training data in order to prevent errors in predictions.
Predictive image regions:
The brighter each pixel is in the heatmap the more influence it can have on the predictions.
If the color is bright it means that a change in these pixels will change the prediction.
A probability indicating how likely the image contains the disease. 50% means the network is not sure.
NOT FOR MEDICAL USE. This is a prototype system for diagnosing
chest x-rays using neural networks. All processing is done on your
device and images are not sent to the server. If you continue you
assume all liability when using the system. A neural network model
(~150mb) will be downloaded to your browser.
By Joseph Paul Cohen, Paul Bertin, and Vincent Frappier 2019
New: Download the offline version here for Mac OSX