Uncertainty Quantification using Variational Inference for Biomedical Image Segmentation

Anonymous

Keywords: Deep learning

TL;DR: Uncertainty Quantification using Variational Inference for Biomedical Image Segmentation

Abstract:

Deep learning motivated by convolutional neural networks
has been highly successful in a range of medical imaging problems like
image classification, image segmentation, image synthesis etc. However
for validation and interpretability, not only do we need the predictions
made by the model but also how confident it is while making those
predictions. This is important in safety critical applications for the people
to accept it. In this work, we used an encoder decoder architecture based
on variational inference techniques for segmenting brain tumour images.
We compare different backbones architectures like U-Net, V-Net and
FCN as sampling data from the conditional distribution for the encoder.
We evaluate our work on the publicly available BRATS dataset using
Dice Similarity Coefficient (DSC) and Intersection Over Union (IOU)
as the evaluation metrics. Our model outperforms previous state of the
art results while making use of uncertainty quantification in a principled
bayesian manner.

Community Implementations: [ 1 code implementation](https://www.catalyzex.com/paper/arxiv:2008.07588/code)