Tversky Loss Function for Image Segmentation Using 3D Fully Convolutional Deep Networks

By Seyed Sadegh Mohseni Salehi, Deniz Erdoğmuş, A. Gholipour. MLMI@MICCAI, 2017

1,143 citations recorded by the source corpus.

Fully convolutional deep neural networks carry out excellent potential for fast and accurate image segmentation. One of the main challenges in training these networks is data imbalance, which is particularly problematic in medical imaging applications such as lesion segmentation where the number of lesion voxels is often much lower than the number of non-lesion voxels. Training with unbalanced data can lead to predictions that are severely biased towards high precision but low recall (sensitivity), which is undesired especially in medical applications where false negatives are much less tolerable than false positives. Several methods have been proposed to deal with this problem including balanced sampling, two step training, sample re-weighting, and similarity loss functions. In this paper, we propose a generalized loss function based on the Tversky index to address the issue of data imbalance and achieve much better trade-off between precision and recall in training 3D fully convolutional deep neural networks. Experimental results in multiple sclerosis lesion segmentation on magnetic resonance images show improved F2 score, Dice coefficient, and the area under the precision-recall curve in test data. Based on these results we suggest Tversky loss function as a generalized framework to effectively train deep neural networks.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/6bf187cf239e66767688ed7dd88f6a408bf465f0

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Optimization for Medical Image Segmentation: Theory and Practice When Evaluating With Dice Score or Jaccard Index (2020, background): “Indeed, in the risk minimization framework for medical image segmentation, there are numerous examples where each of these measures are optimized [9], [15], [16].”
Boundary loss for highly unbalanced segmentation (2018, methodology): “The recent research in (Salehi et al., 2017; Abraham and Khan, 2018) investigated losses based on the Tversky similarity index in order to provide a better trade-off between precision and recall.”
U-Net and Its Variants for Medical Image Segmentation: A Review of Theory and Applications (2020, background): “Tversky loss is a modification of the Dice loss that gives different weights to false positive and false negative results [364].”
Medical Image Segmentation Using Deep Learning: A Survey (2020, background): “[88] proposed the Tversky Loss (TL) that is a regularized version of Dice loss to control”
Unified Focal loss: Generalising Dice and cross entropy-based losses to handle class imbalanced medical image segmentation (2021, methodology): “The Tversky index (Salehi et al., 2017) is closely related to the DSC, but enables optimisation for output imbalance by assigning weights α and β to false positives and false negatives, respectively: where p 0 i is the probability of pixel i belonging to the foreground class and p 1 i is the…”
A survey of loss functions for semantic segmentation (2020, background): “Tversky index (TI) [11] can also be seen as an generalization of Dice’s coefficient.”
Optimizing the Dice Score and Jaccard Index for Medical Image Segmentation: Theory and Practice (2019, background): “y) 2 D(y;y~) and D(y;y~) = 2J(y;y~) 1 + J(y;y~) : (4) Indeed, in the risk minimization framework for medical image segmentation, there are numerous examples where each of these measures are optimized [8,15,19]. In risk minimization, we replace a similarity S : YY! [0;1] with its corresponding loss 1”
Unsupervised domain adaptation for medical imaging segmentation with self-ensembling (2018, background): “The Tversky Loss (Salehi et al., 2017), for instance, is quite similar to the Dice loss but with two additional hyperparameters ( α and β ).”
AnatomyNet: Deep learning for fast and fully automated whole‐volume segmentation of head and neck anatomy (2018, methodology): “The dice loss, the minus of dice coefficient (DSC), can be employed to partly address the problem by turning pixel-wise labeling problem into minimizing class-level distribution distance [31].”
Reducing the Hausdorff Distance in Medical Image Segmentation With Convolutional Neural Networks (2019, positive): “Tversky Index as the objective function [32].”