In order to ensure the safety of transmission lines, the use of unmanned aerial vehicle (UAV) images for automatic object detection has important application prospects, such as the detection of birds’ nests. The traditional bird’s nest detection methods mainly include the study of morphological characteristics of the bird’s nest. These methods have poor applicability and low accuracy. In this work, we propose a deep learning-based birds’ nests automatic detection framework—region of interest (ROI) mining faster region-based convolutional neural networks (RCNN). First, the prior dimensions of anchors are obtained by using k-means clustering to improve the accuracy of coordinate boxes generation. Second, in order to balance the number of foreground and background samples in the training process, the focal loss function is introduced in the region proposal network (RPN) classification stage. Finally, the ROI mining module is added to solve the class imbalance problem in the classification stage, combined with the characteristics of difficult-to-classify bird’s nest samples in the UAV images. After parameter optimization and experimental verification, the deep learning-based bird’s nest automatic detection framework proposed in this work achieves high detection accuracy. In addition, the mean average precision (mAP) and formula 1 (F1) score of the proposed method are higher than the original faster RCNN and cascade RCNN. Our comparative analysis verifies the effectiveness of the proposed method.
Background: Lymph node (LN) metastasis is the main prognostic factor for local recurrence and overall survival of patients with rectal cancer. The accurate evaluation of LN status in rectal cancer patients is associated with improved treatment and prognosis. This study aimed to apply deep transfer learning to classify LN status in patients with rectal cancer to improve N staging accuracy.
Methods:The study included 129 patients with 325 rectal cancer screenshots of LN T2-weighted (T2W) images from April 2018 to March 2019. Deep learning was applied through a pre-trained model, Inception-v3, for recognition and detection of LN status. The results were compared to manual identification by experienced radiologists. Two radiologists reviewed images and independently identified their status using various criteria with or without short axial (SA) diameter measurements. The accuracy, positive predictive value (PPV), negative predictive value (NPV), sensitivity, specificity, and the area under the receiver operating characteristic (ROC) curve (AUC) were calculated.Results: When the same radiologist performed the analysis, the AUC was not significantly different in the presence or absence of LN diameter measurements (P>0.05). In the deep transfer learning method, the PPV, NPV, sensitivity, and specificity were 95.2%, 95.3%, 95.3%, and 95.2%, respectively, and the AUC and accuracy were 0.994 and 95.7%, respectively. These results were all higher than that achieved with manual diagnosis by the radiologists.
Conclusions:The internal details of LNs should be used as the main criteria for positive diagnosis when using MRI. Deep transfer learning can improve the MRI diagnosis of positive LN metastasis in patients with rectal cancer.
Because of the poor radio frequency coil uniformity and gradient-driven eddy currents, there is much noise and intensity inhomogeneity (bias) in brain magnetic resonance (MR) image, and it severely affects the segmentation accuracy. Better segmentation results are difficult to achieve by traditional methods; therefore, in this paper, a modified brain MR image segmentation and bias field estimation model based on local and global information is proposed. We first construct local constraints including image neighborhood information in Gaussian kernel mapping space, and then the complete regularization is established by introducing nonlocal spatial information of MR image. The weighting between local and global information is automatically adjusted according to image local information. At the same time, bias field information is coupled with the model, and it makes the model reduce noise interference but also can effectively estimate the bias field information. Experimental results demonstrate that the proposed algorithm has strong robustness to noise and bias field is well corrected.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.