IMR Press / RCM / Volume 20 / Issue 3 / DOI: 10.31083/j.rcm.2019.03.5201
Open Access Original Research
Optical Coherence Tomography Vulnerable Plaque Segmentation Based on Deep Residual U-Net
Show Less
1 College of Mechanical Engineering and Automation, Northeastern University, Shenyang, 110004, P. R. China
2 College of Information Science and Engineering, Northeastern University, Shenyang, 110004, P. R. China
*Correspondence: (Lincan Li)
Rev. Cardiovasc. Med. 2019, 20(3), 171–177;
Submitted: 20 May 2019 | Accepted: 14 August 2019 | Published: 30 September 2019
Copyright: © 2019 Li and Jia. Published by IMR press.
This is an open access article under the CC BY-NC 4.0 license

Automatic and accurate segmentation of intravascular optical coherence tomography imagery is of great importance in computer-aided diagnosis and in treatment of cardiovascular diseases. However, this task has not been well addressed for two reasons. First, because of the difficulty of acquisition, and the laborious labeling from personnel, optical coherence tomography image datasets are usually small. Second, optical coherence tomography images contain a variety of imaging artifacts, which hinder a clear observation of the vascular wall. In order to overcome these limitations, a new method of cardiovascular vulnerable plaque segmentation is proposed. This method constructs a novel Deep Residual U-Net to segment vulnerable plaque regions. Furthermore, in order to overcome the inaccuracy in object boundary segmentation which previous research has shown extensively, a loss function consisting of weighted cross-entropy loss and Dice coefficient is proposed to solve this problem. Thorough experiments and analysis have been carried out to verify the effectiveness and superior performance of the proposed method.

Intravascular optical coherence tomography
image semantic segmentation
encoder-decoder architecture
residual block
boundary segmentation
Figure 1.
Back to top