No abstract

In view of the low detection efficiency of traditional aerial image-based self-explosive insulators and the need to manually extract the features of self-explosive insulators, a self-explosive insulator detection algorithm based on the improved YOLO v4 is proposed: We use the hybrid data augmentation method to increase the number of defective samples of self-explosive insulators and increase the diversity of self-explosive insulators to improve the detection accuracy of the algorithm. In addition, the channel attention mechanism is combined with the YOLO v4 algorithm to improve the feature extraction capability of algorithms to self-explosion insulators. The proposed algorithm is verified using the expanded self-explosive insulator dataset and the traffic signs dataset TT100K. The experimental results show that the proposed algorithm has a mAP of 92.0 % on the self-explosive insulator dataset and has an F1-score of 78.70% on TT100K dataset, which has positive significance for the research on self-explosive insulator detection.

<div>Accurate segmentation of rectal cancer and rectal wall based on high-resolution T2-weighted magnetic resonance imaging (MRI-HRT2) is the basis of rectal cancer staging. However, complex imaging background, highly characteristics variation and poor contrast hindered the research progress of the automatic rectal cancer segmentation. In this study, a multi-task learning network, namely mask segmentation with boundary constraints (MSBC-Net), is proposed to overcome these limitations and to obtain accurate segmentation results by locating and segmenting rectal cancer and rectal wall automatically. Specifically, at first, a region of interest (RoI)-based segmentation strategy is designed to enable end-to-end multi-task training, where a sparse object detection module is used to automatically localize and classify rectal cancer and rectal wall to mitigate the problem of background interference, and a mask and boundary segmentation block is used to finely segment the RoIs; second, a modulated deformable backbone is introduced to handle the variable features of rectal cancer, which effectively improves the detection performance of small objects and adaptability of the proposed model. Moreover, the boundary head is fused into the mask head to segment the ambiguous boundary of the target and constrain the mask head to obtain more refined segmentation results. In total, 592 annotated rectal cancer patients in MRI-HRT2 are enrolled, and the comprehensive results show that the proposed MSBC-Net outperforms state-of-the-art methods with a dice similarity coefficient (DSC) of 0.801 (95\% CI, 0.791-0.811), which can be well extended to other medical image segmentation tasks with high potential clinical applicability.</div>

No abstract

No abstract