Negatives Make a Positive: An Embarrassingly Simple Approach to Semi-Supervised Few-Shot Learning.

Semi-Supervised Few-Shot Learning (SSFSL) aims to train a classifier that can adapt to new tasks using limited labeled data and a fixed amount of unlabeled data. Various sophisticated methods have been proposed to tackle the challenges associated with this problem. In this paper, we present a simple but quite effective approach to predict accurate negative pseudo-labels of unlabeled data from an indirect learning perspective. We leverage these pseudo-labels to augment the support set, which is typically limited in few-shot tasks, e.g., 1-shot classification. In such label-constrained scenarios, our approach can offer highly accurate negative pseudo-labels. By iteratively excluding negative pseudo-labels one by one, we ultimately derive a positive pseudo-label for each unlabeled sample in our approach. The integration of negative and positive pseudo-labels complements the limited support set, resulting in significant accuracy improvements for SSFSL. Our approach can be implemented in just few lines of code by only using off-the-shelf operations, yet it outperforms state-of-the-art methods on four benchmark datasets. Furthermore, our approach exhibits good adaptability and generalization capabilities when used as a plug-and-play counterpart alongside existing SSFSL methods and when extended to generalized linear models.

Ovarian conservation is associated with better survival in young patients with T1N0M0 cervical adenocarcinoma: a population-based study.

PURPOSE: Ovarian conservation is controversial in patients with cervical adenocarcinoma due to the risk of ovarian metastasis. The aim of this study is to evaluate the association of ovarian conservation with survival outcomes in young patients with T1N0M0 cervical adenocarcinoma. METHODS: Women who were 45 years of age or younger with T1N0M0 cervical adenocarcinoma from 1988 to 2013 recorded in the Surveillance, Epidemiology, and End Results (SEER) database were included. Propensity score weighting was used to balance the intragroup differences. Cause-specific survival (CSS) and overall survival (OS) were compared using Kaplan-Meier estimates. A multivariate Cox model was used to adjust for covariates including propensity score. A stratified analysis was then conducted. RESULTS: Totally, 1090 (79.7%) patients underwent oophorectomy and 278 (20.3%) patients whose ovaries were preserved were identified. Patients with preserved ovaries were younger, with a lower T classification and less likely to undergo pelvic lymphadenectomy (all p < 0.05). After propensity weighting, ovarian conservation group had better cause-specific survival (CSS) (5-year 98.8 versus 97.1%, 10-year 98.0 versus 95.2%, p = 0.0370) and overall survival (OS) (5-year 98.8 versus 97.1%, 10-year 96.5 versus 93.5%, p = 0.0025). After adjustment, the CSS benefit of ovarian conservation was marginally significant (p = 0.051) and OS benefit was still significant (p = 0.006). Stratified analysis showed that the CSS benefit was found in T1b classification (HR, 0.23; 95% CI 0.06-0.89, p = 0.033) and histological grade > 1 (HR 0.12; 95% CI 0.02-0.87; p = 0.035). CONCLUSION: Among young women with T1N0M0 cervical adenocarcinoma, ovarian conservation is associated with better survival.

Efficacy of caesarean scar defect repair in improving postmenstrual bleeding and factors associated with poor effect.

Caesarean scar defect (CSD) can cause postmenstrual bleeding. Defect repair is an effective technique to improve this symptom, but there are still a few patients getting little improvement. This retrospective study evaluates the efficacy of scar repair and explores the factors associated with poor effect. In total, 123 patients were involved in the final analysis. All of them complained about menstruation period >7 days due to postmenstrual bleeding. Before surgery, 87.8% of patients had a menstruation period more than 10 days and 20.3% had a period more than 15 days. After surgery, a normal menstruation period (< =7 days) was achieved in 46.3% (95%CI 37.3%-55.6%) of patients and a menstruation period lasting no more than 10 days was achieved in 74.8% (95%CI 66.2%-82.2%). Through multivariate logistic analysis, four factors were found dependently associated with poor effect (defined as menstruation period >10 days after surgery): repeated caesarean section (OR 9.75, 95%CI 2.30-41.36, 0.002) was a risk factor, while defect volume >600 mm3 (OR 0.14, 95%CI 0.03-0.56, 0.006), interval from caesarean section to symptom emerging >3 months (OR 0.25, 95%CI 0.07-0.94, 0.041) and straight or retroflexed uterus (OR 0.19, 95%CI 0.05-0.79, 0.022) were protective factors. Impact statement What is already known on this subject? Caesarean scar defect can cause postmenstrual bleeding. Defect repair can improve this symptom, but there are still a few patients getting little improvement after surgery. What do the results of this study add? Defect volume >600 mm3, interval from caesarean section to symptom emerging >3 months and straight or retroflexed uterus are protective factors of poor effect (defined as menstruation period >10 days after surgery), and repeated caesarean section is a risk factor. What are the implications of these findings for clinical practice and/or further research? These findings may help in counselling the patients and in medical decision. Further researches are needed to explore other factors associated with surgical effect and build prediction models.

Neuroendocrine-like cells -derived CXCL10 and CXCL11 induce the infiltration of tumor-associated macrophage leading to the poor prognosis of colorectal cancer.

Our previous study revealed that neuroendocrine differentiation in colorectal cancer is one of the important factors leading to worse prognosis. In this study, we apply immunohistochemical staining, Western-blot, RT-PCR and ELISA to investigate the underlying mechanism that how the neuroendocrine differentiation to affect the prognosis of colorectal cancer. The interaction of colorectal cancer cells, neuroendocrine-like cells and tumor-associated macrophages in colorectal cancer progress is also investigated. By analyzing 82 cases of colorectal cancer patients treated in our institution, we found that colorectal adenocarcinoma with neuroendocrine differentiation had increasing number of tumor-associated macrophages and worse prognosis. Further evaluation of cytology showed that neuroendocrine cells have the ability to recruit tumor-associated macrophages to infiltrate the tumor tissue, and the tumor-associated macrophages enhance the proliferation and invasion abilities of the colon cancer cells. Moreover, we confirmed that CXCL10 and CXCL11 are the key chemokines in neuroendocrine-like cells and they promote the chemotaxis activity of tumor-associated macrophages. The secretion of CXCL10 and CXCL11 by neuroendocrine-like cells can recruit tumor-associated macrophages to infiltrate in tumor tissues. The latter enhances the proliferation and invasion of colorectal cancer cell and lead to poor prognosis.

Clinicopathological features and prognosis of gastroenteropancreatic neuroendocrine tumors: analysis from a single-institution.

BACKGROUND: The gastroenteropancreatic neuroendocrine neoplasm (GEP-NEN) is the most common type of neuroendocrine neoplasm. We summarized data in our centre to investigate the clinicopathological features, diagnostic methods, therapeutic approaches and prognosis for this neoplasm to increase knowledge of this disease in Asian populations. METHOD: A total of 122 patients treated at Sun Yet-san Memorial Hospital of Sun Yat-sen University between January 2000 and December 2011 were analyzed retrospectively. RESULTS: Pancreas was the most common site of involvement (65/122, 53.3%); this disease has no special symptoms; positive rates of chromogranin A (CgA) and synaptophysin (Syn) were 81.1% and 87.7%, respectively. The positive rate of Syn had statistical difference among the three grades, but not CgA. Some 68 patients had G1 tumors, 32 G2 tumors and 22 G3 tumors, and Chi-square test showed that higher grading was correlated with worse prognosis (χ2=32.825, P=0.0001). A total of 32 patients presented with distant metastasis, and 8 cases emerged during following up. Cox proportional hazards regression modeling showed that the tumor grade (P=0.01), lymphatic metastasis (P=0.025) and distant metastasis (P=0.031) were predictors of unfavorable prognosis. The overall 5-year survival rate was 39.6%, the 5-year survival rate of G1 was 55.7%, and the G2 and G3 were 34.2% and 0%, respectively. CONCLUSIONS: The incidence of gastroenteropancreatic neuroendocrine tumors has risen over the last 12 years. All grades of these diseases metastasize readily, and further research regarding the treatment of patients after radical surgery is needed to prolong disease-free survival.