Welcome to Brain Tumor Radiogenomic Classification project!
Train model or Use pretrained models
The aim of this project is to determine the presence of MGMT factor, which causes the formation of glioblastoma, by MRI scan. This can help in the early stages of a medical examination.
The results are evaluated on the area under the ROC curve between the predicted probability and the observed target
An ROC curve (receiver operating characteristic curve) is a graph showing the performance of a classification model at all classification thresholds
This curve plots two parameters:
- True Positive Rate
- False Positive Rate
True Positive Rate (TPR) is a synonym for recall and is therefore defined as follows (proportion of samples that were correctly classified)
The 3D CNN preserve temporal information and propagate it through the layers of the network. The tensor xi in this case is 4D and has size Ni × D × H × W, where Ni is the number of filters used in the i-th block. Each filter is 4-dimensional and it has size Ni−1 × t × d × d where t denotes the temporal extent of the filter. The filters are convolved in 3D, over space dimension
Where in the 2D case this informations is lost
The compared results for the [Baseline] Regression Classifier and 3D ResNet are prominent. The 3D ResNet's performance is much better for evry single scan type
3D CNN showed better results compared to 2D classifier
- big dataset
- filters are convolved in 3D over space dimension
Improved score with “vanilla” residual blocks