Cascade Faster R-CNN Detection for Vulnerable Plaques in OCT Images

Abstract

Through the application of optical coherence tomography (OCT), computerized medical image analysis can provide intelligent, auxiliary diagnosis for patients with atherosclerosis. In order to provide higher-resolution imaging for medical interventions, a novel deformable cascade faster region-based convolutional neural networks (Faster R-CNN) is proposed to realize vascular plaque recognition in OCT images. The cascade convolutional network is designed for lumen feature detection, and the deformable convolution and region of interest pooling are adapted to anchor scaling and offset changes, thereby improving classification precision. To deal with the problem of limited samples, an innovative, multiple task loss function is established for online instant training of unlabeled samples. Experimental results prove that the proposed deformable cascade Faster R-CNN outperforms the original Faster R-CNN method in terms of recognition accuracy. Thus, the proposed deformable cascade Faster R-CNN method is effective at recognizing vulnerable plaques.

Keywords

• Cardiovascular disease
• optical coherence tomography
• vulnerable plaque detection
• faster R-CNN