OCT image segmentation by use of CNN

Optical coherence tomography (OCT) can achieve high-resolution and cross-sectional imaging of the internal microstructure in materials and biologic systems by measuring backscattered and back-reflected light. Commonly, for characterizing superficial plaques in inter-coronary arteries, the acquired OCT images are manually differentiated into four types: lipid tissue (LT), fibrous tissue (FT), mixed tissue (MT) and calcified tissue (CA). The accuracy heavily relies on the experience of human observers. In this study, we proposed a convolutional neural network (CNN) based method to automatically characterize plaques in OCT images. Unlike traditional methods, our method uses the image as a direct input and performs classification as a single-step process. Figure 1 shows three segmentation examples with our CNN-based method, where each row shows an original OCT image, ground truth characterization image, and the CNN-based characterization result.

Methodology:

Our CNN-based automatic plaque characterization method includes two steps: tissue area detection and CNN-based pixel classification. First, we used Otsu’s automatic thresholding based method to detect the tissue area in an OCT image. Second, we used a CNN-based classifier to classify each pixel in the tissue area into five different tissue categories: LT, FT, MT, CA and background (BK). The BK pixel was defined as the pixel that did not belong to any of the other four tissue types.

Overview of our proposed method.

 

The architecture of our proposed CNN-base classifier.

Results:

The tissue classification on two example OCT images. FT: dark green; LT: red; MT: light green; CA: white. Each column shows respectively the original OCT images, ground truth characterization images and the CNN-based characterization results in our experiments, respectively.

Related publications:

  • Shenghua He, Mark Anastasio, Hua Li*, “CNN-based automatic plaque characterization for intracoronary optical coherence tomography images”, SPIE Medical Imaging Conference Proceedings, 2018.