Convolutional neural networks have shown significant results in brain tumor segmentation. Automation of brain tumor segmentation can be very useful for 3D magnetic resonance images (MRIs) to assess the diagnostic and treatment of the disease. However, high memory consumption is still a problem in 3D-CNNs. The model studies 3D encoder-decoder architectures trained with patch based techniques to reduce memory consumption and lower the effect of unbalanced data. Different trained models are used to generate an ensemble that increase the model performance. Voxel-wise uncertainty information is also introduced, both epistemic and aleatoric with the help of test-time dropout (TTD) and data-augmentation (TTA) respectively. To increase the accuracy of segmentation a hybrid approach is proposed.
Input variables : 3D MRI Scans
Output Variables : Segmented image
Visit Model : github.com
Additional links : arxiv.org
| Model Category | : | Public |
| Date Published | : | September, 2020 |
| Healthcare Domain | : |
Medical Technology
Provider |
| Code | : | github.com |
| Health Risk Management |
| Disease Detection |