Multimodal Brain Tumor Segmentation using BraTS 2018 Dataset.
The data was collected from Multimodal Brain Tumor Segmentation Challenge 2018 (BraTS) Data. My google drive directory link (view only):- https://drive.google.com/drive/folders/1RSjZ6ASBMSPgUtFQAzvpBx1aW5VXPtAM?usp=sharing
Copy of the model weights:- https://drive.google.com/file/d/11WJbOZ9KdNMwNAGX8ZYyyjD1b3nMH4uG/view?usp=sharing
The data were distributed after their pre-processing, i.e. co-registered to the same anatomical template, interpolated to the same resolution (1 mm^3) and skull-stripped.
The data consist of two folders LGG(Lower Grade Glioma) and HGG(High Grade Glioma). Each consists mri scan of a patient, each folder itself having four modaltlies and the segmentated results.
As all the scans were formated in NIfTI format (i.e. .nii.gz), so we have used SimpleITK library for converting .nii.gz format to 3D numpy array.
Provided data was already skull-stripped.
Segmentation of gliomas in pre-operative MRI scans. Use the provided clinically-acquired training data to produce segmentation labels.
Here we have proposed U-Net for our semnatic segmentation problem:-
Wikipedia:- Sørensen's original formula was intended to be applied to discrete data. Given two sets, X and Y, it is defined as:-
Here |X| and |Y| are the cardinalities of the two sets (i.e. the number of elements in each set). The Sørensen index equals twice the number of elements common to both sets divided by the sum of the number of elements in each set.
For our metric, small modification was made in denominator side i.e. instead of sum of absolute of X and Y sum of square of X and Y was taken.Here, X = y_true, Y = y_pred.
def dice_coef(y_true, y_pred, epsilon=1e-6):
intersection = K.sum(K.abs(y_true * y_pred), axis=-1)
return (2. * intersection) / (K.sum(K.square(y_true),axis=-1) + K.sum(K.square(y_pred),axis=-1) + epsilon)
def dice_coef_loss(y_true, y_pred):
return 1-dice_coef(y_true, y_pred)
HGG Result Samples
LGG Result Samples
Test Data | Dice Coefficient |
---|---|
HGG Set-1 | 0.9795 |
HGG Set-2 | 0.9855 |
HGG Set-3 | 0.9793 |
LGG | 0.9950 |
MIT © Aryaman Sinha