Semi-Supervised Learning with Adaptive Pseudo-Label Selection and Correction for Predicting Overall Survival Time of Esophageal Cancer

EEG power spectra parameterization and adaptive channel selection towards semi-supervised seizure prediction

New treatment options and centralization of surgery have improved survival for patients with non-metastatic esophageal or gastric cancer. It is unknown, however, which patients benefitted the most from treatment advances. The aim of this study was to identify best-case, typical and worst-case scenarios in terms of survival time, and to assess if survival associated with these scenarios changed over time. Patients with non-metastatic potentially resectable esophageal or gastric cancer diagnosed between 2006 and 2020 were selected from the Netherlands Cancer Registry. Best-case (20th percentile), upper-typical (40th percentile), typical (median), lower-typical (60th percentile) and worst-case (80th percentile) survival scenarios were defined, and regression analysis was used to investigate the change in survival time for each scenario across years. For patients with esophageal cancer (N=24 352) survival time improved on average 12.0 (until 2011), 1.5 (until 2018), 0.7, 0.4 and 0.2months per year for the best-case, upper-typical, median, lower-typical and worst-case scenario, respectively. For patients with gastric cancer (N=9993) survival time of the best-case scenario remained constant, whereas the upper-typical, median, lower-typical and worst-case scenario improved on average with 1.0 (until 2018), 0.5, 0.2 and 0.2months per year, respectively. Subgroup analyses showed that, survival scenarios improved for nearly all patients across treatment groups and for patients with squamous cell carcinomas or adenocarcinomas. Survival improved for almost all patients suggesting that in clinical practice the vast majority of patients benefitted from treatment advances. The clinically most meaningful survival advantage was observed for the best-case scenario of esophageal cancer.

BackgroundHealth-related quality of life and survival are two important outcome measures in cancer research and practice. The aim of this paper is to examine the relationship between quality of life data and survival time in cancer patients.MethodsA review was undertaken of all the full publications in the English language biomedical journals between 1982 and 2008. The search was limited to cancer, and included the combination of keywords 'quality of life', 'patient reported-outcomes' 'prognostic', 'predictor', 'predictive' and 'survival' that appeared in the titles of the publications. In addition, each study was examined to ensure that it used multivariate analysis. Purely psychological studies were excluded. A manual search was also performed to include additional papers of potential interest.ResultsA total of 451 citations were identified in this rapid and systematic review of the literature. Of these, 104 citations on the relationship between quality of life and survival were found to be relevant and were further examined. The findings are summarized under different headings: heterogeneous samples of cancer patients, lung cancer, breast cancer, gastro-oesophageal cancers, colorectal cancer, head and neck cancer, melanoma and other cancers. With few exceptions, the findings showed that quality of life data or some aspects of quality of life measures were significant independent predictors of survival duration. Global quality of life, functioning domains and symptom scores - such as appetite loss, fatigue and pain - were the most important indicators, individually or in combination, for predicting survival times in cancer patients after adjusting for one or more demographic and known clinical prognostic factors.ConclusionThis review provides evidence for a positive relationship between quality of life data or some quality of life measures and the survival duration of cancer patients. Pre-treatment (baseline) quality of life data appeared to provide the most reliable information for helping clinicians to establish prognostic criteria for treating their cancer patients. It is recommended that future studies should use valid instruments, apply sound methodological approaches and adequate multivariate statistical analyses adjusted for socio-demographic characteristics and known clinical prognostic factors with a satisfactory validation strategy. This strategy is likely to yield more accurate and specific quality of life-related prognostic variables for specific cancers.

To analyze the clinical and pathological informations of metastatic prostate cancer patients to find the predictive factors of the survival. To filter 364 cases of metastatic prostate cancer in the 940 cases of prostate cancer that were treated in Cancer Hospital Fudan University in Shanghai from March 1998 to June 2009, the cases had hormonal therapy and full clinical and pathological records. All the 364 cases were followed up and the clinical and pathological informations were analyzed, to find the predictive factors that related to the prognosis. Statistic software SPSS 15.0 was used for analysis. Cumulative survival was analyzed by the method of Kaplan-Meier. Cox regression was used for univariate and multivariate analysis. Log-rank method was used for the significance test. The last follow-up date was 30th June 2009 and the median follow-up time was 24 months. At the final follow-up, 240 cases were alive, 109 cases were dead and 15 cases were lost to follow up. The median survival time of metastatic prostate cancer was 64 months, and the one-year, two-year, three-year, four-year, five-year survival rate was 92%, 78%, 66%, 60%, 54%. The univariate analysis indicated that Gleason score (P = 0.033), clinical stage (P < 0.001), the effectiveness of hormonal therapy (P < 0.001), the prostate specific antigen (PSA) nadir during hormonal therapy (P < 0.001) and the time from the start of hormonal therapy to the PSA nadir (P = 0.002) were predictive factors for the survival time of metastatic prostate cancer. The multivariate analysis indicated that the PSA nadir during hormonal therapy (P < 0.001) and the time from the start of hormonal therapy to the PSA nadir (P < 0.001) were independent factors that predict the survival time of metastatic prostate cancer. The PSA nadir during hormonal therapy and the time from the start of hormonal therapy to the PSA nadir are independent factors that predict the survival time of metastatic prostate cancer.

We evaluated the accuracy of oncologists' estimates of expected survival time in recurrent ovarian cancer. Oncologists estimated expected survival time at baseline for each patient, who were then followed up for survival time. We hypothesized that oncologists' estimates of expected survival time would be independently significant predictors of survival, unbiased (approximately equal proportions [50%] living longer versus shorter than their expected survival time), or imprecise (<30% within 0.75-1.33 times their observed survival time). We also hypothesized that simple multiples (0.25, 0.5, 2, and 3) of each expected survival time would define ranges that accurately described 3 scenarios for survival time: worst-case (10% of participants with the shortest survival), typical (middle 50%), and best-case (10% with the longest survival) scenarios. There were 898 participants; the median (interquartile range) for expected survival time was 12 months (range; 8-14) and the median for observed survival time was 13 months (range; 12-14). Oncologists' estimates of expected survival time were independently significant predictors of observed survival time (HR 0.96 per month, 95% CI 0.94-0.98, p < .0001). As hypothesized, 55% lived longer than their expected survival time, 45% shorter than their expected survival time, and 23% of estimates of expected survival time were within 0.75 to 1.33 times their observed survival time. Simple multiples of the expected survival time provided ranges that accurately described 3 scenarios for survival time: 7% of patients died within 0.25 times their expected survival time (worst-case), 53% lived between 0.5 and 2 times their expected survival time (typical), and 13% lived longer than 3 times their expected survival time (best case). Oncologists' estimates of expected survival time were independently significant predictors of survival time. Simple multiples of the expected survival time provided accurate ranges for scenarios for survival that are useful for explaining prognosis.

Semi-supervised learning is a good way to train a classification model using a small number of labeled and large number of unlabeled data. We applied semi-supervised learning to a synthetic aperture radar(SAR) image classification model with a limited number of datasets that are difficult to create. To address the previous difficulties, semi-supervised learning uses a model trained with a small amount of labeled data to generate and learn pseudo labels. Besides, a lot of number of papers use a single fixed threshold to create pseudo labels. In this paper, we present a semi-supervised synthetic aperture radar(SAR) image classification method that applies different thresholds for each class instead of all classes sharing a fixed threshold to improve SAR classification performance with a small number of labeled datasets.

Worst-case, typical, and best-case scenarios for survival, based on simple multiples of an individual's expected survival time (EST), estimated by their oncologist, are a useful way of formulating and explaining prognosis. We aimed to determine the accuracy and prognostic significance of oncologists' estimates of EST, and the accuracy of the resulting scenarios for survival time, in advanced gastric cancer. Sixty-six oncologists estimated the EST at baseline for each of the 152 participants they enrolled in the INTEGRATE trial. We hypothesized that oncologists' estimates of EST would be unbiased (∼50% would be longer or shorter than the observed survival time [OST]); imprecise (<33% within 0.67-1.33 times the OST); independently predictive of overall survival (OS); and accurate at deriving scenarios for survival time with approximately 10% of patients dying within a quarter of their EST (worst-case scenario), 50% living within half to double their EST (typical scenario), and 10% living three or more times their EST (best-case scenario). Oncologists' estimates of EST were unbiased (45% were shorter than the OST, 55% were longer); imprecise (29% were within 0.67-1.33 times observed); moderately discriminative (Harrell's C-statistic 0.62, p = .001); and an independently significant predictor of OS (hazard ratio, 0.89; 95% confidence interval, 0.83-0.95; p = .001) in a Cox model including performance status, number of metastatic sites, neutrophil-to-lymphocyte ratio ≥3, treatment group, age, and health-related quality of life (EORTC-QLQC30 physical function score). Scenarios for survival time derived from oncologists' estimates were remarkably accurate: 9% of patients died within a quarter of their EST, 57% lived within half to double their EST, and 12% lived three times their EST or longer. Oncologists' estimates of EST were unbiased, imprecise, moderately discriminative, and independently significant predictors of OS. Simple multiples of the EST accurately estimated worst-case, typical, and best-case scenarios for survival time in advanced gastric cancer. Results of this study demonstrate that oncologists' estimates of expected survival time for their patients with advanced gastric cancer were unbiased, imprecise, moderately discriminative, and independently significant predictors of overall survival. Simple multiples of the expected survival time accurately estimated worst-case, typical, and best-case scenarios for survival time in advanced gastric cancer.

To evaluate the claim that oncologists overestimate expected survival time (EST) in advanced cancer. We pooled 7 prospective studies in which observed survival time (OST) was compared with EST (median survival in a group of similar patients estimated at baseline by the treating oncologist). We hypothesized that EST would be well calibrated (approximately 50% of EST longer than OST) and imprecise (<30% of EST within 0.67 to 1.33 of OST), and that multiples of EST would provide well-calibrated scenarios for survival time: worst-case (approximately 10% of OST <1/4 of EST), typical (approximately 50% of OST within half to double EST), and best-case (approximately 10% of OST >3 times EST). Associations between baseline characteristics and calibration of EST were assessed. Characteristics of 1,211 patients: median age 66 years, male 61%, primary site lung (40%) and upper gastrointestinal (16%). The median OST was 8 months, and EST was 9 months. Oncologists' estimates of EST were well calibrated (50% longer than OST) and imprecise (28% within 0.67 to 1.33 of OST). Scenarios for survival time based on simple multiples of EST were well calibrated: 8% of patients had an OST less than 1/4 their EST (worst-case), 56% had an OST within half to double their EST (typical), and 11% had an OST greater than 3 times their EST (best-case). Calibration was independent of age, sex, and cancer type. Oncologists were no more likely to overestimate survival time than to underestimate it. Simple multiples of EST provide well-calibrated estimates of worst-case, typical, and best-case scenarios for survival.

A deep learning-based cancer survival time classifier for small datasets

Adaptive multi-view selection for semi-supervised emotion recognition of posts in online student community

DNA microarrays allow the simultaneous measurement of thousands of gene expression levels in any given patient sample. Gene expression data have been shown to correlate with survival in several cancers, however, analysis of the data is difficult, since typically at most a few hundred patients are available, resulting in severely underdetermined regression or classification models. Several approaches exist to classify patients in different risk classes, however, relatively little has been done with respect to the prediction of actual survival times. We introduce CASPAR, a novel method to predict true survival times for the individual patient based on microarray measurements. CASPAR is based on a multivariate Cox regression model that is embedded in a Bayesian framework. A hierarchical prior distribution on the regression parameters is specifically designed to deal with high dimensionality (large number of genes) and low sample size settings, that are typical for microarray measurements. This enables CASPAR to automatically select small, most informative subsets of genes for prediction. Validity of the method is demonstrated on two publicly available datasets on diffuse large B-cell lymphoma (DLBCL) and on adenocarcinoma of the lung. The method successfully identifies long and short survivors, with high sensitivity and specificity. We compare our method with two alternative methods from the literature, demonstrating superior results of our approach. In addition, we show that CASPAR can further refine predictions made using clinical scoring systems such as the International Prognostic Index (IPI) for DLBCL and clinical staging for lung cancer, thus providing an additional tool for the clinician. An analysis of the genes identified confirms previously published results, and furthermore, new candidate genes correlated with survival are identified.

Cancer is one of the major causes of death. For cancer, the general conventional treatment and standard of care for clinical oncology remains surgery followed by radiation and/or systemic chemotherapy as deemed appropriate based on the clinical findings. Chemoimmunotherapy is an approach to treat cancer where chemotherapy is given along with immunotherapy. Chemoimmunotherapy may be useful to enhance survival time in cancer by improve immunity of the patients. This approach may enhance the therapeutic efficacy. A comparative study was done to assess the therapeutic efficacy of the whole cell vaccine and the tumor extract with or without combination chemotherapy with the synthesized glutamine and glutamic acid derivatives and analogs as well as the standard drug etoposide against Ehrlich Ascites Carcinoma (EAC) cells in Swiss Albino mice. The study showed promising results with the compound 5-N-n-hexyl-2-(4-iso-butylbenzenesulphonyl)glutamine. The compound when combined with the whole cell vaccine as well as the tumor extract increases the survival time and the therapeutic efficacy which is comparable with that of standard drug etoposide.

Background: Anaplastic carcinoma of the thyroid (ATC) is the most aggressive thyroid gland malignancy. Therapies modalities were controversial due to its poor survival prognosis. The survival data after various therapeutic modalities for ATC is still limited. Methods: This study is a retrospective cohort study by taking data from the Cancer Registry in Bali, Denpasar, Indonesia. All patients with anaplastic thyroid cancer who visited Sanglah Hospital in the range 2018 to 2020 were included in this study. Data regarding duration of survival, received treatment modalities, tumor size from radiological data, and laboratories were analyzed with SPSS 25.0. Results: This study included 10 subjects with ATC, mostly male with a mean age of 60.7 years old. The mean survival was 74.8 days with a median survival of 74.5 days. Of the ten patients, 2 patients died without getting any therapy, both surgical and medical therapy. In this study, there was a significant difference in duration of survival between therapeutic modalities. There was no influence of tumor volume factors, gap from initial presentation to therapeutic modality, nodes involved, blood gas analysis at the initial presentation, NLR, PLR, TSH, and FT4 for the survival time of anaplastic thyroid cancer. Conclusion: There is a significant difference in duration of survival between therapeutic modalities for anaplastic ca thyroid. Even so, isthmectomy, tracheostomy, and chemotherapy therapy, only provide an extension of survival time for an average of 40 days.

BackgroundTo determine the expression of bone morphogenetic protein-2 (BMP-2) and its receptors BMPRIA, BMPRIB, and BMPRII in epithelial ovarian cancer (EOC) and to analyze their influence on the prognosis of ovarian cancer patients.MethodsSemi-quantitative RT-PCR and western blot were applied to detect the expression of BMP-2 and its receptors BMPRIA, BMPRIB, and BMPRII in EOC, benign ovarian tumors, and normal ovarian tissue at the mRNA and protein levels. Immunohistochemistry was used to determine the expression of BMP-2 and its receptors in 100 patients with EOC to analyze their influence on the five-year survival rate and survival time of ovarian cancer patients.Results(1) The mRNA and protein expression levels of BMP-2, BMPRIB, and BMPRII in ovarian cancer tissue were remarkably lower than those in benign ovarian tumors and normal ovarian tissue, while no significant differences in BMPRIA expression level was found among the three kinds of tissues. (2) The five-year survival rate and the average survival time after surgery of EOC patients with positive expression of BMP-2, BMPRIB, and BMPRII were remarkably higher than those of patients with negative expression of BMP-2, BMPRIB, and BMPRII. BMPRIA expression was not associated with the five-year survival rate or with the average survival time of ovarian cancer patients.ConclusionsBMP-2, BMPRIB, and BMPRII exhibited low expression in EOC tissue, and variation or loss of expression may indicate poor prognosis for ovarian cancer patients.

BACKGROUND: Small-cell lung cancer (SCLC) is a highly aggressive neoplasm, characterized by early development of metastasis and very poor clinical outcome. The therapeutic possibilities are limited and poorly changed over the past three decades. Enhancer of zeste homolog 2 (EZH2) is a member of the polycomb repressive complex 2. It is involved in epigenetic gene silencing through the trimethylation of H3K27 and it has been identified as a biomarker of aggressive and highly proliferating tumors, emerging as a potential target for cancer therapy. OBJECTIVE: We aimed to correlate the expression profile of the EZH2 and H3K27me3 proteins in tumor tissues of non-treated SCLC patients with patient's outcome. METHODS: EZH2 and H3K27me3 expression were analyzed by immunohistochemistry in 48 SCLC tumor biopsies. The expression of the markers was quantified in a semi quantitative way and correlated with clinic pathological factors and patients overall survival. To mimic an EZH2 loss, H209 SCLC cells were treated with an EZH2 inhibitor for 48h and the in vivo growth pattern was investigated in the chorioallantoic membrane (CAM) assay. RESULTS: All patients died within one year after diagnosis. There was a heterogeneous expression pattern for EZH2 and H3K27me3. Strong expression of EZH2 and H3K27me3 was observed in appr. 50% of the patients. The Kaplan Meier curve showed a significant association of high expression pattern of EZH2 and shorter overall patient survival time after 12 and 25 weeks of follow up, respectively (p=0.030; p=0.014) Different markers of tumor aggressiveness after EZH2 inhibition were investigated in the CAM model. CONCLUSION: Strong EZH2 and H3K27me3 expression in non-treated SCLC patients was related to worse prognosis. EZH2 inhibitors should be considered in future trials for SCLC. Citation Format: Ana Paula Fernandes, Rita deCassia S Alves, Guilherme Watte, Philipp Kunze, Adriana Vial Roehe, Regine Schneider-Stock. Strong expression of EZH2 and H3K27me3 is associated with poor survival time in small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 3713.