No abstract

Recently, remarkable progress has been made in weakly supervised object localization (WSOL) to promote object localization maps. The common practice of evaluating these maps applies an indirect and coarse way, i.e., obtaining tight bounding boxes which can cover high-activation regions and calculating intersection-over-union (IoU) scores between the predicted and ground-truth boxes. This measurement can evaluate the ability of localization maps to some extent, but we argue that the maps should be measured directly and delicately, i.e., comparing the maps with the ground-truth object masks pixel-wisely. To fulfill the direct evaluation, we annotate pixel-level object masks on the ILSVRC [1] validation set. We propose to use IoU-Threshold curves for evaluating the real quality of localization maps. Beyond the amended evaluation metric and annotated object masks, this work also introduces a novel self-enhancement method to harvest accurate object localization maps and object boundaries with only category labels as supervision. We propose a two-stage approach to generate the localization maps by simply comparing the similarity of point-wise features between the high-activation and the rest pixels. Based on the predicted localization maps, we explore to estimate object boundaries on a very large dataset. A hard-negative suppression loss is proposed for obtaining fine boundaries. We conduct extensive experiments on the ILSVRC and CUB [2] benchmarks. In particular, the proposed Self-Enhancement Maps achieve the state-of-the-art localization accuracy of 54.88% on ILSVRC. The code and the annotated masks are released at https://github.com/xiaomengyc/SEM.

ObjectiveBased on the theoretical basis of Gabor wavelet transformation, the application effects of feature extraction algorithm in Magnetic Resonance Imaging (MRI) and the role of feature extraction algorithm in the diagnosis of lumbar vertebra degenerative diseases were explored. Q, et al. (2020) The application of key feature extraction algorithm based on Gabor wavelet transformation in the diagnosis of lumbar intervertebral disc degenerative changes. PLoS ONE 15(2): e0227894. https://doi.org/10. ConclusionThe feature extraction algorithm based on Gabor wavelet transformation could easily and quickly realize the localization of the lumbar intervertebral disc, and the accuracy of the results was ensured. In addition, from the aspect of vertebral body tracking, the tracking effects based on the KLT algorithm were better and faster than those based on the maximum mutual information method.

No abstract

In order to explore the application of deep learning based intelligent imaging technology in the diagnosis of colorectal cancer, Tangdu Hospital patients are selected as the research object in this study. By scanning the cancer sites, then distinguishing and extracting the features of the tumors, the collected data are input into the designed in-depth learning intelligent assistant diagnosis system for comparison. The results show that in the analysis of image prediction accuracy, the best prediction accuracy of T1-weighted image method is matrix GLCM (gray level co-occurrence matrix) algorithm, the best prediction accuracy of adding T1-weighted image method is matrix MGLSZM (multi-gray area size matrix) algorithm, and the best prediction accuracy of T2-weighted image method is ALL combination of all texture features, and the best prediction accuracy of three imaging sequences is not more than 0.8. In the AUC analysis of the area under the curve of different texture features, it is found that T2-weighted imaging method has obvious advantages in differentiating colorectal cancer from other methods. Therefore, through this study, it is found that in the use of deep learning intelligent assistant diagnosis system for the diagnosis of colorectal cancer, it can provide useful information for the clinical diagnosis of colorectal cancer to a certain extent. Although there are some deficiencies in the research process, it still provides experimental basis for the diagnosis and treatment of colorectal cancer in later clinical stage.INDEX TERMS Deep learning, colorectal cancer, weighted images, accuracy, texture features.