We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

eess.IV

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Electrical Engineering and Systems Science > Image and Video Processing

Title: Combining CNNs With Transformer for Multimodal 3D MRI Brain Tumor Segmentation With Self-Supervised Pretraining

Abstract: We apply an ensemble of modified TransBTS, nnU-Net, and a combination of both for the segmentation task of the BraTS 2021 challenge. In fact, we change the original architecture of the TransBTS model by adding Squeeze-and-Excitation blocks, an increasing number of CNN layers, replacing positional encoding in Transformer block with a learnable Multilayer Perceptron (MLP) embeddings, which makes Transformer adjustable to any input size during inference. With these modifications, we are able to largely improve TransBTS performance. Inspired by a nnU-Net framework we decided to combine it with our modified TransBTS by changing the architecture inside nnU-Net to our custom model. On the Validation set of BraTS 2021, the ensemble of these approaches achieves 0.8496, 0.8698, 0.9256 Dice score and 15.72, 11.057, 3.374 HD95 for enhancing tumor, tumor core, and whole tumor, correspondingly. Our code is publicly available.
Subjects: Image and Video Processing (eess.IV); Computer Vision and Pattern Recognition (cs.CV)
Cite as: arXiv:2110.07919 [eess.IV]
  (or arXiv:2110.07919v1 [eess.IV] for this version)

Submission history

From: Mariia Dobko [view email]
[v1] Fri, 15 Oct 2021 08:03:27 GMT (840kb,D)

Link back to: arXiv, form interface, contact.