Deep Attentive Features for Prostate Segmentation in 3D Transrectal Ultrasound.

This repository is the implementation of the DAF3DNet for Prostate Segmentation by Haoran Dou in MUSIC Lab at Shenzhen University.

Deep Attentive Features for Prostate Segmentation in 3D Transrectal Ultrasound.
Yi Wang, Haoran Dou, Xiaowei Hu, Lei Zhu, Xin Yang, Ming Xu, Jing Qin, Pheng-Ann Heng, Tianfu Wang, and Dong Ni.
IEEE Transactions on Medical Imaging, 2019. [Paper] [arXiv]

Automatic prostate segmentation in transrectal ultrasound (TRUS) images is of essential importance for image-guided prostate interventions and treatment planning. However, developing such automatic solutions remains very challenging due to the missing/ambiguous boundary and inhomogeneous intensity distribution of the prostate in TRUS, as well as the large variability in prostate shapes. This paper develops a novel 3D deep neural network equipped with attention modules for better prostate segmentation in TRUS by fully exploiting the complementary information encoded in different layers of the convolutional neural network (CNN). Our attention module utilizes the attention mechanism to selectively leverage the multilevel features integrated from different layers to refine the features at each individual layer, suppressing the non-prostate noise at shallow layers of the CNN and increasing more prostate details into features at deep layers. Experimental results on challenging 3D TRUS volumes show that our method attains satisfactory segmentation performance. The proposed attention mechanism is a general strategy to aggregate multi-level deep features and has the potential to be used for other medical image segmentation tasks.

Usage

Dependencies

This work depends on the following libraries:
Pytorch == 0.4.0
Python == 3.6

Train and Validate

Run

python Train.py

You can rewrite the DataOprate.py to train your own data.

Result

One case to illustrate the effectiveness of the proposed attention module for the feature refinement.

metric results

Metric	3D FCN	3D U-Net	Ours
Dice	0.8210	0.8453	0.9004
Jaccard	0.6985	0.7340	0.8200
CC	0.5579	0.6293	0.7762
ADB	9.5801	8.2715	3.3198
95HD	25.113	20.390	8.3684
Precision	0.8105	0.8283	0.8995
Recall	0.8486	0.8764	0.9055

Citation

If this work is helpful for you, please cite our paper as follows:

@ARTICLE{8698868,
  author={Y. {Wang} and H. {Dou} and X. {Hu} and L. {Zhu} and X. {Yang} and M. {Xu} and J. {Qin} and P. -A. {Heng} and T. {Wang} and D. {Ni}},
  journal={IEEE Transactions on Medical Imaging}, 
  title={Deep Attentive Features for Prostate Segmentation in 3D Transrectal Ultrasound}, 
  year={2019},
  volume={38},
  number={12},
  pages={2768-2778},
  doi={10.1109/TMI.2019.2913184}}