Search

The detection of ballastless track surface (BTS) defects is a prerequisite for ensuring the safe operation of high-speed railways. Traditional convolutional neural networks fail to fully exploit contextual information and lack global pixel representations. The extensive stacking of convolutions leads deep learning models to play a black-box detection role, lacking interpretability. Due to the current lack of sufficient high-quality surface data for ballastless tracks, it is a severe constraint on the accurate identification of the substructure state in high-speed railways. This paper proposes an intelligent detection method for BTS defects named TrackNet based on self-attention and transfer learning. The method enhances the fusion ability of global features of BTS defects using multihead self-attention. The model's dependence on extensive defect data is reduced by transferring knowledge from large-scale publicly available datasets. Experimental results demonstrate that compared to advanced Swin Transformer model results, the TrackNet model achieves improvements in average accuracy and F1-score by 5.15% and 5.16%, respectively, on limited test data. The TrackNet model visualizes the decision regions of the model in identifying BTS defects, revealing the black-box recognition mechanism of deep learning models. This research performs engineering applications and provides valuable insights for the multiclass recognition of BTS defects in high-speed railways. [ABSTRACT FROM AUTHOR]

Background: To compare the right and left transthoracic approach on the post-operative survival of patients with lymph node-negative esophageal squamous cell carcinoma.Methods: Six hundred and ninety-five ESCC patients who underwent esophagectomy between 1990 and 2005 were retrospectively enrolled in the present study and were confirmed by histology to be of no lymph node metastasis. Those who had received neoadjuvant chemotherapy or radiotherapy were excluded from the study. Patients were divided into two groups, the left (n=545) and right (n=150) transthoracic groups. The follow-up duration ranged from 1 to 20 years with a mean of 7 years. Kaplan-Meier and univariate and multivariate Cox proportional hazards were used for analysis.Results: 3- and 5-year CSS rates were 62.0 % and 44.0 % in the left group, while the corresponding figures in the right group were 56.0 % and 40.0 %(P<0.05). The overall survival for the two groups was significantly different (P=0.045). Survival analyses were stratified by stages, which found that the favorable survival advantage was not present. When the survival curves were stratified by tumor locations, a significant difference was not revealed. Surgical approaches were regarded as one of the prognostic factors in the univariate analysis (P=0.019). However, this significance could not be confirmed in multivariate Cox regression analysis (P=0.193).Conclusions: The left transthoracic approach is superior in some aspects to the right transthoracic approach regarding surgical and oncological outcomes in the treatment of lymph node negative ESCC. [ABSTRACT FROM AUTHOR]

The study on the effects of different cutting intensities on the eco-physiological characteristics and growth status of Stylosanthes guianensis showed that moderate cutting could promote the regrowth capacity of the grass. The total content of leaf chlorophyll had no significant difference between zero-, light-and heavy cutting, but the ratio of leaf chlorophyll a/b increased to 1.59: 1 and 1.52: 1 under light-and heavy cutting, respectively, compared with 1.22:1 under zero-cutting, which could be very helpful for the plant to over-compensate itself. Cutting also affected the growth status and activities of plant roots. At the earlier stage of heavy cutting, the total length, total surface area and mean diameter of plant roots decreased by 54.9%, 66.5% and 27.2%, respectively, and root activities also decreased greatly, in comparing with zero-cutting. But, the above-mentioned indices could be recovered to a higher level under different cutting treatments than under zero-cutting at the later growth period of the grass. As for the total annual yield of the grass, it was 3 179.8 g x m(-2) under light-cutting, 3006.1 g x m(-2) under heavy- cutting, and 2 936.98 g x m(-2) under zero-cutting, indicating that rational cutting could improve grass productivity.