A Novel Focal Tversky Loss Function With Improved Attention U-Net for Lesion Segmentation

By Nabila Abraham, Naimul Mefraz Khan. IEEE International Symposium on Biomedical Imaging, 2018

945 citations recorded by the source corpus.

We propose a generalized focal loss function based on the Tversky index to address the issue of data imbalance in medical image segmentation. Compared to the commonly used Dice loss, our loss function achieves a better trade off between precision and recall when training on small structures such as lesions. To evaluate our loss function, we improve the attention U-Net model by incorporating an image pyramid to preserve contextual features. We experiment on the BUS 2017 dataset and ISIC 2018 dataset where lesions occupy 4.84% and 21.4% of the images area and improve segmentation accuracy when compared to the standard U-Net by 25.7% and 3.6%, respectively.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging

View Paper: https://www.semanticscholar.org/paper/fa56122bf5e9c66cd4418513d0b8d914024e4e9b

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

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

A survey of loss functions for semantic segmentation (2020, background): “Focal Tversky loss [12] also attempts to learn hard-examples such as with small ROIs(region of interest) with the help of γ coefficient as shown below:”
CHAOS Challenge - Combined (CT-MR) Healthy Abdominal Organ Segmentation (2020, methodology): “Nürnberger) • Modified Attention U-Net [26], employing soft attention gates and multiscaled input image pyramid for better feature representation is used.”
U-Net and Its Variants for Medical Image Segmentation: A Review of Theory and Applications (2020, methodology): “Medical ultrasound use cases also include analysis of soft tissue such as nerve bundles [39], [50], [156], [305], [306].”
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.”
Sharp U-Net: Depthwise Convolutional Network for Biomedical Image Segmentation (2021, background): “It is worth mentioning that other loss functions such as the focal Tversky loss [25] and distance-based losses [26,27] can also be used for training.”
Unified Focal loss: Generalising Dice and cross entropy-based losses to handle class imbalanced medical image segmentation (2021, methodology): “Current state-of-the-art models for the BUS2017 dataset incorporate attention gates, which may provide benefits in class imbalanced situations by using contextual information from the gating signal to refine skip connections, highlighting the regions of interest (Abraham and Khan, 2019).”
AAU-Net: An Adaptive Attention U-Net for Breast Lesions Segmentation in Ultrasound Images (2022, methodology): “Our comparative methods include U-net [10], SegNet [9], Att U-net [25], U-net++ [7], U-net3+ [8], Abraham and Khan [24], SKNet [31], AE U-Net [33] and”
Towards a guideline for evaluation metrics in medical image segmentation (2022, neutral): “The analysis utilized our medical image segmentation framework MIScnn [18] and was performed with the following parameters: Sampling in 64% training, 16% validation, and 20% testing sets; resizing into 512 × 512 pixel images; value intensity normalization via Z-score; extensive online image augmenta”
Medical Image Segmentation Review: The Success of U-Net (2022, background): “Abraham et al. [14] propose the Focal Tversky Attention U-Net with a generalized focal loss function that modulates the Tversky index [130] to address the issue of data imbalance and improve precision and recall balance in medical image segmentation.”