Brain MRI classifier



Training accuracy: {{train_acc}}

Test accuracy: {{test_acc}}


Detecting tumors in an MRI scan

This app trains a convolutional neural network on a dataset of MRI brain scans to detect tumors. The training data consists of 253 images of brains with tumors and 98 images of brains without tumors.

To train the network, click the TRAIN button. This button only works when running the app locally in development mode.

To test the trained classifier, click one of the images below. You'll see the output of each layer of the network as the image is processed. The final layer output is used to classify the image as having a tumor or not.

Network diagram

This diagram excludes pooling and normalization layers


Images with tumor

Images without tumor

Network propagation

Classification result:


Ground truth:

Classification error!