Incorporating prior shape knowledge via data-driven loss model to improve 3D liver segmentation in deep CNNs.

Authors:
Saeed Mohaqeqi, PhD
Saeed Mohaqeqi, PhD
Shahed University
Deep Learning, Medical Image Processing, Python Programming
Karaj, Alborz | Iran (Islamic Republic of)

Int J Comput Assist Radiol Surg 2020 Feb 4;15(2):249-257. Epub 2019 Nov 4.

Department of Biomedical Engineering, Engineering Faculty, Shahed University, Tehran, Iran.

Purpose: Convolutional neural networks (CNNs) have obtained enormous success in liver segmentation. However, there are several challenges, including low-contrast images, and large variations in the shape, and appearance of the liver. Incorporating prior knowledge in deep CNN models improves their performance and generalization.

Methods: A convolutional denoising auto-encoder is utilized to learn global information about 3D liver shapes in a low-dimensional latent space. Then, the deep data-driven knowledge is used to define a loss function and combine it with the Dice loss in the main segmentation model. The resultant hybrid model would be forced to learn the global shape information as prior knowledge, while it tries to produce accurate results and increase the Dice score.

Results: The proposed training strategy improved the performance of the 3D U-Net model and reached the Dice score of 97.62% on the Sliver07-I liver dataset, which is competitive to the state-of-the-art automatic segmentation methods. The proposed algorithm enhanced the generalization and robustness of the hybrid model and outperformed the 3D U-Net model in the prediction of unseen images.

Conclusions: The results indicate that the incorporation of prior shape knowledge enhances liver segmentation tasks in deep CNN models. The proposed method improves the generalization and robustness of the hybrid model due to the abstract features provided by the data-driven loss model.

Download full-text PDF

Source
http://dx.doi.org/10.1007/s11548-019-02085-yDOI Listing
February 2020
1.659 Impact Factor

Publication Analysis

Top Keywords

liver segmentation
12
hybrid model
12
loss model
8
model
8
deep cnn
8
prior knowledge
8
incorporating prior
8
data-driven loss
8
cnn models
8
robustness hybrid
8
prior shape
8
shape knowledge
8
generalization robustness
8
learn global
8
u-net model
8
liver
6
segmentation
5
knowledge
5
accurate increase
4
improved performance
4

References

(Supplied by CrossRef)

P Campadelli et al.
Artif Intell Med 2009

T Heimann et al.
IEEE Trans Med Imaging 2009

F Lu et al.
Int J Comput Assist Radiol Surg 2017

P Hu et al.
Phys Med Biol 2016

AA Taha et al.
BMC Med Imaging 2015

K Kamnitsas et al.
Med Image Anal 2017

Olaf Ronneberger et al.
2015

Özgün Çiçek et al.
2016

Similar Publications