Segmentation of Coronary Calcified Plaque in Intravascular OCT Images Using a Two-Step Deep Learning Approach

Abstract

We developed a fully automated, two-step deep learning approach for characterizing coronary calcified plaque in intravascular optical coherence tomography (IVOCT) images. First, major calcification lesions were detected from an entire pullback using a 3D convolutional neural network (CNN). Second, a SegNet deep learning model with the Tversky loss function was used to segment calcified plaques in the major calcification lesions. The fully connected conditional random field and the frame interpolation of the missing calcification frames were used to reduce classification errors. We trained/tested the networks on a large dataset comprising 8,231 clinical images from 68 patients with 68 vessels and 4,320 ex vivo cadaveric images from 4 hearts with 4 vessels. The 3D CNN model detected major calcifications with high sensitivity (97.7%), specificity (87.7%), and F1 score (0.922). Compared to the standard one-step approach, our two-step deep learning approach significantly improved sensitivity (from 77.5% to 86.2%), precision (from 73.5% to 75.8%), and F1 score (from 0.749 to 0.781). We investigated segmentation performance for varying numbers of training samples; at least 3,900 images were required to obtain stable segmentation results. We also found very small differences in calcification attributes (e.g., angle, thickness, and depth) and identical calcium scores on repetitive pullbacks, indicating excellent reproducibility. Applied to new clinical pullbacks, our method has implications for real-time treatment planning and imaging research.

Automatic Assessment of Calcified Plaque and Nodule by Optical Coherence Tomography Adopting Deep Learning Mode

“…Lee et al developed a two-step deep learning method combined with 3D convolutional neural network and SegNet. The sensitivity, precision and F1 score of the model were 86.2%, 75.8% and 78.1% [14]. Our study proposed a Multi-Scale and Multi-Task u-net network (MS-MT unet) based on the u-net model.…”

“…The deep learning is a necessary way to achieve arti cial intelligence. The common deep learning algorithm methods include U-net, Deeplab v3+ and SegNet [14]. Ughi et al rstly used the automatic algorithm to identify the characteristics of atherosclerotic plaques by textural features and optical attenuation coe cient [10].…”

“…In addition to high accuracy of qualitative and quantitative evaluation, the calculation time duration is a key factor in clinical application of the deep learning algorithm. Lee et al reported that their deep learning model analysis took 0.3 sec per frame [14]. Assuming there are 400 frames in a single pullback, the automatic analysis would take around 120 sec compared with 0.5-2 hours by cardiologist' experience.…”

Recent Trends in Artificial Intelligence-Assisted Coronary Atherosclerotic Plaque Characterization

“…Thereafter, the generated features were selected using the analysis of variance (ANOVA) statistic. The Wilcoxon signed-rank test can also be used for feature organization [ 101 ].…”

“…Similarly, CNN architectures ResNet-50, ResNet-101 and Inception-v3 using SVM and the discriminant analysis (DA) classifier could characterize plaques more efficiently. In [ 101 ], the 3D CNN model, along with a SegNet architecture, was utilized for calcified plaque segmentation. In addition to plaque detection, the 3D U–Net CNN was used in challenging tasks such as coronary artery lumen delineation and stenosis grading [ 79 ].…”