A web application which detects brain tumours and projects the model of the brain in AR. Reports3D employs the U-Net CNN architecture which detects and segments brain tumours based on MRI scans and renders the volume of location and size of the tumour and brain. This volumetric rendering can e viewed in an Augmented Reality enviroment.
Model View in Website and Augmented Reality
The following image depicts the architecture of Reports3D:
The segmentation results of our CNN are depicted in the image below:
The table below delineates the DSC scores for different regions of the tumour
Mean Dice Similarity Coefficient | Median Dice Similarity Coefficient | |
---|---|---|
Full Tumour | 0.87 | 0.90 |
Tumour Core | 0.76 | 0.84 |
Enhancing Tumour | 0.71 | 0.80 |
3-D Model Results
3-D Model on web page | 3-D Model in Augmented Reality |
---|---|