Multilabel Convolutional Network With Feature Denoising and Details Supplement.

In multilabel images, the changeable size, posture, and position of objects in the image will increase the difficulty of classification. Moreover, a large amount of irrelevant information interferes with the recognition of objects. Therefore, how to remove irrelevant information from the image to improve the performance of label recognition is an important problem. In this article, we propose a convolutional network based on feature denoising and details supplement (FDDS) to address this issue. In FDDS, we first design a cascade convolution module (CCM) to collect spatial details of upper features, in order to enhance the information expression of features. Second, the feature denoising module (FDM) is further put forward to reallocate the weight of the feature semantic area, in order to enrich the effective semantic information of the current feature and perform denoising operations on object-irrelevant information. Experimental results show that the proposed FDDS outperforms the existing state-of-the-art models on several benchmark datasets, especially for complex scenes.

CapMatch: Semi-Supervised Contrastive Transformer Capsule With Feature-Based Knowledge Distillation for Human Activity Recognition.

This article proposes a semi-supervised contrastive capsule transformer method with feature-based knowledge distillation (KD) that simplifies the existing semisupervised learning (SSL) techniques for wearable human activity recognition (HAR), called CapMatch. CapMatch gracefully hybridizes supervised learning and unsupervised learning to extract rich representations from input data. In unsupervised learning, CapMatch leverages the pseudolabeling, contrastive learning (CL), and feature-based KD techniques to construct similarity learning on lower and higher level semantic information extracted from two augmentation versions of the data", weak" and "timecut", to recognize the relationships among the obtained features of classes in the unlabeled data. CapMatch combines the outputs of the weak-and timecut-augmented models to form pseudolabeling and thus CL. Meanwhile, CapMatch uses the feature-based KD to transfer knowledge from the intermediate layers of the weak-augmented model to those of the timecut-augmented model. To effectively capture both local and global patterns of HAR data, we design a capsule transformer network consisting of four capsule-based transformer blocks and one routing layer. Experimental results show that compared with a number of state-of-the-art semi-supervised and supervised algorithms, the proposed CapMatch achieves decent performance on three commonly used HAR datasets, namely, HAPT, WISDM, and UCI_HAR. With only 10% of data labeled, CapMatch achieves F1 values of higher than 85.00% on these datasets, outperforming 14 semi-supervised algorithms. When the proportion of labeled data reaches 30%, CapMatch obtains F1 values of no lower than 88.00% on the datasets above, which is better than several classical supervised algorithms, e.g., decision tree and k -nearest neighbor (KNN).

Globalized Multiple Balanced Subsets With Collaborative Learning for Imbalanced Data.

The skewed distribution of data brings difficulties to classify minority and majority samples in the imbalanced problem. The balanced bagging randomly undersampes majority samples several times and combines the selected majority samples with minority samples to form several balanced subsets, in which the numbers of minority and majority samples are roughly equal. However, the balanced bagging is the lack of a unified learning framework. Moreover, it fails to concern the connection of all subsets and the global information of the entire data distribution. To this end, this article puts several balanced subsets into an effective learning framework with a criterion function. In the learning framework, one regularization term called RS establishes the connection and realizes the collaborative learning of all subsets by requiring the consistent outputs of the minority samples in different subsets. Besides, another regularization term called RW provides the global information to each basic classifier by reducing the difference between the direction of the solution vector in each subset and that in the entire dataset. The proposed learning framework is called globalized multiple balanced subsets with collaborative learning (GMBSCL). The experimental results validate the effectiveness of the proposed GMBSCL.

Geometric Structural Ensemble Learning for Imbalanced Problems.

The classification on imbalanced data sets is a great challenge in machine learning. In this paper, a geometric structural ensemble (GSE) learning framework is proposed to address the issue. It is known that the traditional ensemble methods train and combine a series of basic classifiers according to various weights, which might lack the geometric meaning. Oppositely, the GSE partitions and eliminates redundant majority samples by generating hyper-sphere through the Euclidean metric and learns basic classifiers to enclose the minority samples, which achieves higher efficiency in the training process and seems easier to understand. In detail, the current weak classifier builds boundaries between the majority and the minority samples and removes the former. Then, the remaining samples are used to train the next. When the training process is done, all of the majority samples could be cleaned and the combination of all basic classifiers is obtained. To further improve the generalization, two relaxation techniques are proposed. Theoretically, the computational complexity of GSE could approach O(ndlog(nmin)log(n maj)) . The comprehensive experiments validate both the effectiveness and efficiency of GSE.

Multiple Partial Empirical Kernel Learning with Instance Weighting and Boundary Fitting.

By dividing the original data set into several sub-sets, Multiple Partial Empirical Kernel Learning (MPEKL) constructs multiple kernel matrixes corresponding to the sub-sets, and these kernel matrixes are decomposed to provide the explicit kernel functions. Then, the instances in the original data set are mapped into multiple kernel spaces, which provide better performance than single kernel space. It is known that the instances in different locations and distributions behave differently. Therefore, this paper defines the weight of instance in accordance with the location and distribution of the instances. According to the location, the instances can be categorized into intrinsic instances, boundary instances and noise instances. Generally, the boundary instances, as well as the minority instances in the imbalanced data set, are assigned high weight. Meanwhile, a regularization term, which regulates the classification hyperplane to fit the distribution trend of the class boundary, is constructed by the boundary instances. Then, the weight of instance and the regularization term are introduced into MPEKL to form an algorithm named Multiple Partial Empirical Kernel Learning with Instance Weighting and Boundary Fitting (IBMPEKL). Experiments demonstrate the good performance of IBMPEKL and validate the effectiveness of the instance weighting and boundary fitting.

Expression of serum exosomal miR-23b-3p in non-small cell lung cancer and its diagnostic efficacy.

The present study aimed to investigate the expression of serum exosomal miR-23b-3p in non-small cell lung cancer (NSCLC) and to determine its diagnostic efficacy for NSCLC. From October, 2017 to October, 2019, 80 patients with NSCLC, 60 patients with pneumonia and 30 healthy subjects undergoing physical examination were enrolled at the People's Hospital of Yangzhong City. Serum samples were collected from the 3 groups of patients. The expression of miR-23b-3p in exosomes was detected by RT-qPCR. The Chi-squared test was used to analyze the expression level of miR-23b-3p in exosomes, and the patients with NSCLC were divided into 2 groups according to the expression level. The association between the patient clinicopathological parameters and receiver operating characteristic (ROC) curves was used to evaluate the diagnostic efficacy of serum exosomal miR-23b-3p in NSCLC. The expression level of serum exosomal miR-23b-3p in the patients with NSCLC was significantly higher than that in patients with pneumonia (t=10.332, P<0.001) and healthy subjects (t=12.810, P<0.001); serum exosomal miR-23b-3p was significantly associated with tumor size, depth of invasion, liver metastasis and TNM stage (P<0.05). The area under the curve (AUC) for miR-23b-3p was 0.915 (95% CI, 0.84-0.92), the optimal relative expression of miR-23b-3p was 3.46, the sensitivity of diagnosis was 87.4%, and the specificity was 93.8%, all higher than that of carcinoembryonic antigen (CEA). The ROCAUC of NSCLC was 0.645 (95% CI, 0.641-0.772) and for Cyfra21-1 it was 0.745 (95% CI, 0.701-0.812). Compared with the patients with pneumonia and the healthy subjects, the patients with NSCLC exhibited a higher level of serum exosomal miR-23b-3p. On the whole, these findings indicate that miR-23b-3p has a higher clinical diagnostic efficacy and may thus be a potential biomarker for the early diagnosis of NSCLC.

[Analysis of the therapeutic effect of esophagectomy with extended 2-field lymph node dissection for esophageal carcinoma].

OBJECTIVE: To summarize the surgical effect and clinical application value of esophagectomy with extended 2-field lymph node dissection for patients with esophageal carcinoma. METHODS: From June 1987 to December 2008, 1690 patients with esophageal cancer underwent esophagectomy with extended 2-field (thoracic and abdominal) dissection of lymph nodes. Patients with the middle and lower thoracic esophageal cancer underwent Ivor-Lewis esophagectomy, and patients with upper thoracic esophageal cancer underwent Akiyama esophagectomy. 2-field (thoracic and abdominal) lymph node metastases information and the 1, 3, 5, 10-year survival rates were analyzed retrospectively. RESULTS: Lymph node metastases were found in 713 patients. The lymph node metastases rate was 42.2% (713/1690).Thoracic lymph node metastasis rate was 39.3% (665/1690), among which in the right pleural apical para-tracheal triangle was 20.7% (349/1690), in the posterior upper mediastinum was 26.3% (444/1690), in the lower mediastinum was 18.2% (307/1690). Abdominal lymph node metastasis rate was 20.1% (339/1690). THE Postoperative complication rate was 16.4% (278/1690), among which the pulmonary complication rate ranking the first, was 43.6% (136/312). The operative mortality rate was 0.2%. The 1-year, 3-year, 5-year and 10-year survival rates were 88.2% (1388/1574), 63.5% (868/1367), 54.8% (705/1287) and 30.8% (232/754), respectively. The 5-year survival rate in patients without lymph node metastasis was 76.2% (448/588), but that in patients with lymph node metastases was 36.8% (257/669). CONCLUSION: The results of this study demonstrated that Ivor-Lewis and Akiyama esophagectomy with two-field lymph node dissection exposes the operation fields clearly and make radical lymphadenectomy thoroughly, especially the lymph nodes in the posterior upper mediastinum around the recurrent laryngeal nerve and in the right pleural apical para-tracheal triangle. It is essential that patients with esophageal carcinoma with lymph node metastases should undergo esophagectomy with extended 2-field dissection of lymph nodes. This can elevate the postoperative 5-year survival rate remarkably.

[Right para-tracheal triangle lymphadenectomy for esophageal carcinoma].

OBJECTIVE: To investigate the incidence of lymph node metastasis (LNM) in the right para-tracheal triangle (RPT) of esophageal carcinoma patients and the technique of dissection. METHODS: On the top of double mediastinal and abdominal lymphadenectomy, 333 esophageal carcinoma patients received RPT lymphadenectomy through the right pleural apical approach from 1990 to 2001. RESULTS: In these 333 patients, the lymph node metastasis (LNM) rate in the RPT was 36.40%. A total of 457 nodes among 2 159 nodes removed gave a metastasis degree of 24.96%. The LNM rates in RPT for cervical, upper third, middle third, and lower third segments of esophagus were 66.67%, 45.45%, 34.19% and 15.79% (P < 0.05), while their respective metastasis degrees were 44.44%, 27.04%, 24.32% and 18.92% (P > 0.05). The frequency of positive nodes in the RPT for PTI, PT1, PT2, PT3 and PT4 was 0, 17.24%, 28.7%, 45.16% and 53.57%, while those of metastasis degree were 0, 8.77%, 17.62%, 33% and 41.17% (P < 0.01). The frequency of LNM in the RPT in papillary, erosive, patch-like and covert type of early tumor was 40%, 3.85%, 0 and 0 (P < 0.05), while those of the metastasis degree were 29.41%, 1.82%, 0 and 0 (P < 0.01). Higher rate of LNM in progressive stenotic esophageal carcinoma was observed compared with those of the other gross types (56.52%, P < 0.05), so was the degree (P < 0.01). The frequency of LNM in the RPT for mono-focal and multi-focal tumor was 34.98% and 70% without significant difference (P > 0.05), while the degree was 24.29% and 53.33% (P < 0.05). Postoperative complications were: leak (0.6%), and recurrent laryngeal nerve injury (1.2%). No injury of vein or infra-clavicular artery, tracheal damage or mortality occurred. CONCLUSION: 1. The lymph node metastasis from esophageal carcinoma has a tendency of wide spread and right para-tracheal triangle is an important region to be doomed. 2. With location, depth of tumor invasion and differentiation of tumor as major factors affecting LNM of esophageal carcinoma, dissection of this region should be paid more emphasis. 3. In early lesions, higher frequency of LNM in the RPT is found in papillary and erosive lesions than in the other macroscopic types. 4. Exposing the RPT, lymph node by dissection through a right pleural apical approach is very important and significant.

[Correlation of aberrant methylation of MGMT gene to MTHFR C677T genetic polymorphisms in esophageal squamous cell carcinoma].

BACKGROUND & OBJECTIVE: As one of the principal causes of gene inactivation, aberrant hypermethylation in the promoter of cancer-related genes has attracted more and more attention. However, such studies on esophageal cancer are still limited. This study was to investigate the association between aberrant hypermethylation of MGMT gene and clinical characteristics as well as MTHFR C677T genetic polymorphisms in esophageal squamous cell carcinoma in a Chinese population. METHODS: A molecular epidemiologic study was conducted at Yangzhong County, Jiangsu Province of China, on histologically confirmed esophageal squamous cell carcinoma patients who were operated in the People's Hospital of Yangzhong County between January 2005 and March 2006. Peripheral blood samples, esophageal cancer tissues and paracancerous normal tissues were collected. Methylation-specific polymerase chain reaction(MSP) was used to detect the CpG island methylation status of MGMT gene. Restrictive fragment length polymorphism (RFLP) technique was used to test polymorphisms of folate metabolism enzyme gene MTHFR. The association between methylation status of MGMT gene and clinical characteristics as well as MTHFR C677T polymorphisms were analyzed. RESULTS: Among 125 esophageal squamous cell carcinoma patients, the aberrant hypermethylation rate of MGMT gene was 27.2% in cancer tissues and 11.2% in paracancerous normal tissues. No hypermethylation was found in normal esophageal tissues from 10 healthy adult subjects. Methylation rate of MGMT gene in cancer tissues was significantly higher in the patients with lymph node metastasis than in those without lymph node metastasis (37.3% vs. 18.2%, P=0.017). No association was found between aberrant DNA methylation and selected factors including sex, age, tobacco smoking, alcohol drinking and green tea drinking. After adjusting by potential confounders, variant allele of MTHFR C677T was found to be associated with hypermethylation of MGMT gene. Compared with wild type CC, the odds ratio was 3.34 (95% CI: 1.07-10.39) for CT and 3.83 (95% CI: 1.13-12.94) for TT. CONCLUSION: Aberrant CpG island hypermethylation of MGMT gene is closely related with the genesis and progression of esophageal squamous cell carcinoma.

[Regulations and lymphadenectomy strategy of mediastinal and upper abdominal lymph node metastasis in thoracic esophageal carcinoma].

BACKGROUND & OBJECTIVE: Regional lymph node metastasis plays an important role in the prognosis of esophageal carcinoma. However, the range of lymph node dissection is still controversial. This study was to investigate the regulations of lymph node metastasis of thoracic esophageal carcinoma in the mediastinum and upper abdomen, and explore the rational lymphadenectomy with Ivor-Lewis procedure. METHODS: A total of 1 412 thoracic esophageal carcinoma patients underwent radical esophagectomy and mediastinal and abdominal lymphadenectomy by Ivor-Lewis procedure from 1990 to 2005 at Yangzhong People's Hospital; 517 of them underwent right para-trachea triangle field lymphadenectomy through the right pleural apical approach. The regulations of regional lymph node metastasis were analyzed. RESULTS: Of the 1,412 patients, 323 (22.88%) had postoperative complications, 2 (0.14%) died during hospitalization, and 547 (38.74%) had lymph node metastasis. The lymph node metastasis rates were 32.30% in the right para-trachea triangle, 18.43% in the upper mediastinum, 5.31% in the lower mediastinum, and 17.28% in the upper abdomen(P<0.001). Of the 13 916 resected lymph nodes, 2 662 (19.13%) were positive; the metastasis degree (positive lymph nodes/resected lymph modes) were 23.83% in the right para-trachea triangle, 18.92% in the upper mediastinum, 21.07% in the lower mediastinum, and 17.20% in the upper abdomen. For those patients with the cancer focuses in the upper, middle and lower segments of the esophagus, the lymph node metastasis rates were 40.59%, 36.97% and 44.35% (P=0.093), respectively, while the lymph node metastasis degree in these 3 fields were 19.60%, 18.35%, and 21.82%, respectively. Both the lymph node metastasis rate and degree were significantly higher in the patients at advanced stage than in the patients at early stage (46.56% vs. 7.75%, 21.82% vs. 4.01%, P<0.001). CONCLUSIONS: Regional lymph node metastasis, especially in the right para-trachea triangle and upper mediastinum, is a key factor for thoracic esophageal carcinoma. Ivor-Lewis esophagectomy with two-field lymph node dissection is a safe operation for thoracic esophageal carcinoma, and may increase the chances of complete resection.