SCMR 22nd Annual Scientific Sessions
Atherosclerosis is caused by plaques, which are made up of cholesterol, fatty substances, cellular waste products, calcium and fibrin, clogging the arteries. Plaques reduce the blood supply and are likely to cause strokes once rupturing . Clinically, plaque compositions are important indicators of their risks . High-resolution multi-contrast MRI has emerged as a promising tool to visualize atherosclerosis plaques. However, the detection and segmentation of plaque composition by radiologists requires intensive training and is time consuming. In this work, we demonstrate a convolutional neural network (CNN) providing a much more efficient and accurate way for automated segmentation and analysis, compared with traditional Bayesian methods.
The dataset in this study comes from Chinese Atherosclerosis Risk Evaluation study (CARE II) , which recruits over 1000 patients from 13 medical centers and hospitals in China, and is collected by Center for Biomedical Imaging Research (CBIR) of Tsinghua University. All MR imaging is performed on 3T MR scanners with the following imaging sequences: TOF, T1W, T2W and MP-RAGE. 16 slice locations per sequence per subject are kept with 4 images representing 4 different contrast weightings. Thus, we have 1098 subjects and 70272 images in total. Each 4 images of the same slice location are co-registered and shared with one pixel-level label.
U-net, a CNN for biomedical image segmentation, is employed as the base model. The input of U-net is modified to adapt 4-channel images as a whole, and the output gives a labeled plaque composition map. Focal loss is introduced to tackle with data imbalance between easy and hard tissues, and it is combined with dice loss to optimize apparent image similarity. We replace some of the convolution units with deformable ones, adding learnable offsets to regular grid sampling locations to capture internal geometric transformation. A two-stage training scheme, with a lighter-load stage 1 (only images with lesions) and a full-data stage 2 (at a smaller learning rate), can significantly reduce training time and improve segmentation accuracy. All the training and evaluation are performed with Keras (TensorFlow backend) as the deep learning framework, using an Ubuntu server with one Titan X GPU.
As can be seen from the evaluation index of image segmentation, U-net achieves a significant improvement over MEPPS (Morphology-Enhanced Probabilistic Plaque Segmentation) , and is able to complete the segmentation of fibrous tissue, calcification, necrotic/lipid core and hemorrhage.
Conclusion: In this study, we apply a CNN (U-net) to automatically identify carotid plaque components. U-net outperforms MEPPS in almost all tissue classes and metrics. The improvement mainly comes from the elimination of manually crafted features, as well as the large amount of training data with consistent labels. It has great potential to become a reliable clinic tool in the future.