Fat Mass and Obesity-Associated Protein Regulates Granulosa Cell Aging by Targeting Matrix Metalloproteinase-2 Gene Via an N6-Methyladenosine-YT521-B Homology Domain Family Member 2-Dependent Pathway in Aged Mice.

In this study, we aimed to investigate the molecular mechanisms of RNA N6-methyladenosine (m6A) modification and how its associated proteins affect granulosa cell aging. A granulosa cell senescence model was constructed to detect the differences in total RNA m6A modification levels and the expression of related enzymes. Changes in downstream molecular expression and the effects on the cellular senescence phenotype were explored by repeatedly knocking down and overexpressing the key genes fat mass and obesity-associated protein (FTO), YT521-B homology domain family member 2 (YTHDF2), and matrix metalloproteinase-2 (MMP2). There was an increased total RNA m6A modification and decreased expression of the demethylase FTO and target gene MMP2 in senescent granulosa cells. FTO and MMP2 knockdown promoted granulosa cell senescence, whereas FTO and MMP2 overexpression retarded it. YTHDF2 and FTO can bind to the messenger RNA of MMP2. The extracellular signal-regulated kinase (ERK) pathway, which is downstream of MMP2, retarded the process of granulosa cell senescence through ERK activators. In granulosa cells, FTO can regulate the expression of MMP2 in an m6A-YTHDF2-dependent manner, influencing the activation status of the ERK pathway and contributing to the aging process of granulosa cells.

Global, regional and national burdens of reproduction-related congenital birth defects, 1990-2019.

Background: Reproduction-related congenital birth defects (RCBDs), including Klinefelter syndrome (KS), Turner syndrome (TS), and urogenital congenital anomalies (UCA), can lead to severe physical and psychosocial disorders. The global impact of RCBDs on children and adults is unknown, which limits high-quality development of populations and increases in life expectancy per capita. Methods: Annual incidence rates, prevalence rates (PR), and disability-adjusted life year (DALY) rates were collected for KS, TS, and UCA for 204 countries and territories, including at birth, for children younger than 1 year, and age-standardized (AS) for all ages. Linear regression was used to calculate their estimated annual percentage changes (EAPCs). Finally, the relationships between EAPCs of each indicator and sociodemographic index (SDI) was investigated using Pearson correlation analysis. Results: Globally, the age-standardized prevalence rate (ASPR) trend is decreasing in KS and TS and increasing in UCA. The DALY rates for children younger than 1 year were on a downward trend in KS and UCA, while they were still rising for TS. The AS-DALY rates were all on a downward trend in KS, TS, and UCA. The DALY rates of KS, TS and UCA were found higher in high-income countries in North America. In addition, the burdens of TS and UCA went down with increasing SDI, whereas the burden of KS increased with increasing SDI. Conclusion: The global burdens of RCBDs have decreased since 1990. This finding can help policymakers implement cost-effective interventions to reduce the burdens of RCBDs.

Learning to Double-Check Model Prediction From a Causal Perspective.

The present machine learning schema typically uses a one-pass model inference (e.g., forward propagation) to make predictions in the testing phase. It is inherently different from human students who double-check the answer during examinations especially when the confidence is low. To bridge this gap, we propose a learning to double-check (L2D) framework, which formulates double check as a learnable procedure with two core operations: recognizing unreliable predictions and revising predictions. To judge the correctness of a prediction, we resort to counterfactual faithfulness in causal theory and design a contrastive faithfulness measure. In particular, L2D generates counterfactual features by imagining: "what would the sample features be if its label was the predicted class" and judges the prediction by the faithfulness of the counterfactual features. Furthermore, we design a simple and effective revision module to revise the original model prediction according to the faithfulness. We apply the L2D framework to three classification models and conduct experiments on two public datasets for image classification, validating the effectiveness of L2D in prediction correctness judgment and revision.

CAPA-IVM improves the cytoplasmic quality of in vitro-matured oocytes from unstimulated mice.

Ovarian tissue oocyte (OTO) in vitro maturation (IVM) is a strategy to improve fertility preservation efficiency. Here, the effects of capacitation IVM (CAPA-IVM) on OTO function were investigated. Immature cumulus-oocyte complexes (COCs) from unstimulated 28-day-old mouse ovaries (mimicking OTOs) underwent CAPA-IVM, standard IVM (S-IVM) or in vivo maturation following ovarian stimulation (OS; positive control), and oocyte meiotic maturation and cytoplasmic quality were assessed. CAPA-IVM resulted in improved oocyte meiotic maturation (P < 0.05) and cumulus expansion (P < 0.0001) compared to S-IVM, with expansion comparable to the OS group. MII OTO ROS was lower after CAPA-IVM than S-IVM (P < 0.0001) but not as low as in the OS group (P = 0.036). CAPA-IVM resulted in a better oocyte mitochondrial distribution than S-IVM (P < 0.05) and was similar to the OS group (P > 0.05). Mitochondrial membrane potential in MII OTOs was higher after CAPA-IVM than S-IVM and OS (P < 0.0001). Compared with S-IVM, CAPA-IVM resulted in lower rates of spindle/chromosome configuration and cortical granule distribution abnormalities (P < 0.05), which were similar to OS levels (P > 0.05). MII OTO intracellular Ca2+ levels were similar in the CAPA-IVM and OS groups (P > 0.05), while S-IVM decreased intracellular Ca2+ (P < 0.05). CAPA-IVM and S-IVM decreased mitochondrial Ca2+ levels (P < 0.05). CAPA-IVM increased expression of antioxidant genes (Sod2 and Sirt1) and Egfr (P < 0.05) but not apoptotic genes (Bcl2, Bax and Bcl2/Bax; P > 0.05). CAPA-IVM increased the OTO maturation rate and quality of oocytes from unstimulated mice to the extent that many features of oocyte cytoplasmic quality were comparable to superovulated in vivo matured oocytes.

Counterfactual Samples Synthesizing and Training for Robust Visual Question Answering.

Today's VQA models still tend to capture superficial linguistic correlations in the training set and fail to generalize to the test set with different QA distributions. To reduce these language biases, recent VQA works introduce an auxiliary question-only model to regularize the training of targeted VQA model, and achieve dominating performance on diagnostic benchmarks for out-of-distribution testing. However, due to the complex model design, ensemble-based methods are unable to equip themselves with two indispensable characteristics of an ideal VQA model: 1) Visual-explainable: The model should rely on the right visual regions when making decisions. 2) Question-sensitive: The model should be sensitive to the linguistic variations in questions. To this end, we propose a novel model-agnostic Counterfactual Samples Synthesizing and Training (CSST) strategy. After training with CSST, VQA models are forced to focus on all critical objects and words, which significantly improves both visual-explainable and question-sensitive abilities. Specifically, CSST is composed of two parts: Counterfactual Samples Synthesizing (CSS) and Counterfactual Samples Training (CST). CSS generates counterfactual samples by carefully masking critical objects in images or words in questions and assigning pseudo ground-truth answers. CST not only trains the VQA models with both complementary samples to predict respective ground-truth answers, but also urges the VQA models to further distinguish the original samples and superficially similar counterfactual ones. To facilitate the CST training, we propose two variants of supervised contrastive loss for VQA, and design an effective positive and negative sample selection mechanism based on CSS. Extensive experiments have shown the effectiveness of CSST. Particularly, by building on top of model LMH+SAR (Clark et al. 2019), (Si et al. 2021), we achieve record-breaking performance on all out-of-distribution benchmarks (e.g., VQA-CP v2, VQA-CP v1, and GQA-OOD).

Nomogram for predicting the risk of preterm delivery after IVF/ICSI treatment: an analysis of 11513 singleton births.

Background: There is a higher risk of preterm delivery (PTD) in singleton live births conceived after in vitro fertilization (IVF)/intracytoplasmic sperm injection (ICSI) compared with spontaneously conceived pregnancies. The objective of our study was to build a predictive nomogram model to suggest the possibility of PTD in singleton pregnancies after IVF/ICSI treatment. Method: 11513 IVF/ICSI cycles with singleton live births were enrolled retrospectively. These cycles were randomly allocated into a training group (80%) and a validation group (20%). We used the multivariate logistics regression analysis to determine prognostic factors for PTD in the training group. A nomogram based on the above factors was further established for predicting PTD. Receiver operating characteristic curves (ROC), areas under the ROC curves (AUC), concordance index (C-index), and calibration plots were analyzed for assessing the performance of this nomogram in the training and validation group. Results: There were fourteen risk factors significantly related to PTD in IVF/ICSI singleton live births, including maternal body mass index (BMI) > 24 kg/m2, smoking, uterine factors, cervical factors, ovulatory factors, double embryo transferred (DET), blastocyst transfer, FET, vanishing twin syndrome (VTS), obstetric complications (placenta previa, placenta abruption, hypertensive of pregnancies, and premature rupture of membrane), and a male fetus. These factors were further incorporated to construct a nomogram prediction model. The AUC, C-index, and calibration curves indicated that this nomogram exhibited fair performance and good calibration. Conclusions: We found that the occurrence of PTD increased when women with obesity, smoking, uterine factors, cervical factors, ovulatory factors, DET, VTS, and obstetric complications, and a male fetus. Furthermore, a nomogram was constructed based on the above factors and it might have great value for clinic use.

An autophagy-related diagnostic biomarker for uterine fibroids: FOS.

Uterine fibroids (UFs) are the most common benign gynecologic tumors in reproductive-aged women. The typical diagnostic strategies of UFs are transvaginal ultrasonography and pathological feature, while molecular biomarkers are considered conventional options in the assessment of the origin and development of UFs in recent years. Here, we extracted the differential expression genes (DEGs) and differential DNA methylation genes (DMGs) of UFs from the Gene Expression Omnibus (GEO) database, GSE64763, GSE120854, GSE45188, and GSE45187. 167 DEGs with aberrant DNA methylation were identified, and further Gene Ontology (GO) enrichment and Kyoto Encyclopedia of Genes and Genomes (KEGG) were performed by the relevant R package. We next discerned 2 hub genes (FOS, and TNFSF10) with autophagy involvement by overlapping 167 DEGs and 232 autophagic regulators from Human Autophagy Database. FOS was identified as the most crucial gene through the Protein-Protein Interactions (PPI) network with the correlation of the immune scores. Moreover, the down-regulated expression of FOS in UFs tissue at both mRNA and protein levels was validated by RT-qPCR and immunohistochemistry respectively. The area under the ROC curve (AUC) of FOS was 0.856, with a sensitivity of 86.2% and a specificity of 73.9%. Overall, we explored the possible biomarker of UFs undergoing DNA-methylated autophagy and provided clinicians with a comprehensive assessment of UFs.

Deconfounded Image Captioning: A Causal Retrospect.

Dataset bias in vision-language tasks is becoming one of the main problems which hinders the progress of our community. Existing solutions lack a principled analysis about why modern image captioners easily collapse into dataset bias. In this paper, we present a novel perspective: Deconfounded Image Captioning (DIC), to find out the answer of this question, then retrospect modern neural image captioners, and finally propose a DIC framework: DICv1.0 to alleviate the negative effects brought by dataset bias. DIC is based on causal inference, whose two principles: the backdoor and front-door adjustments, help us review previous studies and design new effective models. In particular, we showcase that DICv1.0 can strengthen two prevailing captioning models and can achieve a single-model 131.1 CIDEr-D and 128.4 c40 CIDEr-D on Karpathy split and online split of the challenging MS COCO dataset, respectively. Interestingly, DICv1.0 is a natural derivation from our causal retrospect, which opens promising directions for image captioning.

Inhibition of Oocyte Maturation by Follicular Extracellular Vesicles of Nonhyperandrogenic PCOS Patients Requiring IVF.

CONTEXT: Polycystic ovary syndrome (PCOS) is one of the most common diseases that contribute to subfertility. Recent evidence showed that oocytes of women with PCOS matured in vitro away from the follicular fluid presented better potentials, whereas the reason remained unclear. OBJECTIVE: This work aimed to investigate whether follicular extracellular vesicles (EVs) of PCOS patients interfere with the quality of oocytes. METHODS: Follicular EVs of women with PCOS (PCOS-EVs) and control women (CTRL-EVs) were isolated and determined using Western blotting, nanoparticle tracking analysis, and transmission electron microscopy. The 2 types of EVs were co-cultured with murine germinal vesicle oocytes, respectively. Fluorescence-labeled EVs were used to visualize internalization by oocytes. After co-culture, oocyte maturation rates were calculated. Mitochondria distribution and reactive oxygen species (ROS) level were detected in the different groups. Spindle morphology was evaluated using immunofluorescence. Moreover, the expression of catalase (CAT), glutathione synthetase (GSS), and superoxide dismutase (SOD) was determined in the oocytes. RESULTS: Both PCOS-EVs and CTRL-EVs are bilayered vesicles, approximately 100 to 150 nm in size, and enriched in EV-associating protein markers. EVs were internalized by oocytes within 1 hour. Oocyte maturation rate decreased significantly in the PCOS-EV group compared with the CTRL-EV group, whereas the abnormal mitochondria distribution rate and abnormal spindle rate were significantly increased in the PCOS-EV group. Moreover, PCOS-EVs increased the ROS level and the expression of CAT, GSS, and SOD in the oocytes. CONCLUSION: PCOS-EVs interfered with oocyte mitochondria and spindles and inhibited oocyte maturation. Moreover, oxidative stress induced by PCOS-EVs might be a potential cause.

Parental chest computerized tomography examination before IVF/ICSI has no impact on pregnancy and neonatal outcomes: a cohort study of 2680 fresh transfer cycles.

BACKGROUND: Some concern has been expressed regarding the negative effects of low-level ionizing radiation exposure in the context of radiological evaluation prior to IVF/ICSI treatment, but the available evidence is limited and conflicting. The aim of this study is to evaluate pregnancy and neonatal outcomes of couples who did chest computed tomography (CT) prior to IVF/ICSI. METHODS: This was a retrospective cohort study of 2680 IVF/ICSI fresh embryo transfer cycles conducted from January 2019 - August 2020. Fertility outcomes were compared between couples that had or had not undergone CT examination within 3 months prior to the date of oocyte retrieval and sperm collection. Miscarriage was the primary study outcome, while secondary outcomes included the number of oocytes collected, oocyte maturation, normal fertilization, number of good quality cleavage stage embryos, blastocyst formation, implantation, clinical pregnancy, ectopic pregnancy, live birth, multiple birth, Cesarean section rates, gestational weeks, maternal obstetric complications, birth weight, newborn sex ratio, and birth defect incidence. Propensity score matching was used to control for potential confounding variables. RESULTS: Of the 2680 cycles included in this study, couples underwent CT examination in 731 cycles. After 1:1 propensity score matching, 670 cycles were included in each group. When comparing demographic and fertility-related variables between groups that had and had not undergone CT examination after propensity score matching, we detected no significant differences in miscarriage rates (16.99% vs. 15.77%, OR = 1.10, 95CI% = 0.74 to 1.68). Similarly, both groups exhibited comparable oocyte and embryonic development, implantation rates (41.99% vs. 40.42%, OR = 1.07, 95%CI = 0.87 to 1.31), clinical pregnancy rates (45.67% vs. 44.48%, OR = 1.05, 95%CI = 0.85 to 1.30), ectopic pregnancy rates (2.94% vs. 1.68%, OR = 1.78, 95%CI = 0.59 to 5.36), live birth rates (36.57% vs. 35.67%, OR = 1.04, 95%CI = 0.83 to 1.30), multiple birth rates, Cesarean section rates, gestational weeks, maternal obstetric complication rates, and neonatal outcomes. CONCLUSIONS: Chest CT examination before IVF/ICSI has no impact on pregnancy and neonatal outcomes associated with fresh embryo transfer. TRIAL REGISTRATION: Not applicable.

Impact of elevated body mass index on cumulative live birth rate and obstetric safety in women undergoing assisted reproductive technology.

This study evaluated the impact of elevated body mass index (BMI) on short- and long-term outcomes of in-vitro fertilization (IVF)/intracytoplasmic sperm injection (ICSI) treatments. A total of 7229 patients undergoing IVF/ICSI fresh cycles and subsequent frozen embryo transfer cycles from 2014 to 2020 were divided into normal (18.5-24.9 kg/m2) and high BMI (≥ 25 kg/m2) groups. Ovarian response, pregnancy outcomes, and safety of both mother and fetus were the main outcome measures. Furthermore, multivariate analysis was used to determine whether BMI was associated with cumulative live birth rate (CLBR). Results showed that for younger women (< 38 year), CLBR was significantly reduced in the high BMI group compared with the normal BMI control and was accompanied by fewer retrieved oocytes and available embryos. Additionally, the incidence of hypertensive disorders of pregnancy, fetal macrosomia, and cleft lip and palate birth defects resulting from cumulative live births was significantly higher compared with the normal BMI group. No differences were observed among older women (≥ 38 year). Multivariate analysis revealed that high BMI was a risk factor for CLBR. Our study suggested that elevated BMI has a greater adverse impact on younger women.

PLP1 may serve as a potential diagnostic biomarker of uterine fibroids.

Objective: We aim to identify the crucial genes or potential biomarkers associated with uterine fibroids (UFs), which may provide clinicians with evidence about the diagnostic biomarker of UFs and reveal the mechanism of its progression. Methods: The gene expression and genome-wide DNA methylation profiles were obtained from Gene Expression Omnibus database (GEO). GSE45189, GSE31699, and GSE593 datasets were included. GEO2R and Venn diagrams were used to analyze the differentially expressed genes (DEGs) and extract the hub genes. Gene Ontology (GO) analysis was performed by the online tool Database for Annotation, Visualization, and Integrated Discovery (DAVID). The mRNA and protein expression of hub genes were validated by RT-qPCR, western blot, and immunohistochemistry. The receiver operating characteristic (ROC) curve was used to evaluate the diagnostic value. Results: We detected 22 DEGs between UFs and normal myometrium, which were enriched in cell maturation, apoptotic process, hypoxia, protein binding, and cytoplasm for cell composition. By finding the intersection of the data between differentially expressed mRNA and DNA methylation profiles, 3 hub genes were identified, including transmembrane 4 L six family member 1 (TM4SF1), TNF superfamily member 10 (TNFSF10), and proteolipid protein 1 (PLP1). PLP1 was validated to be up-regulated significantly in UFs both at mRNA and protein levels. The area under the ROC curve (AUC) of PLP1 was 0.956, with a sensitivity of 79.2% and a specificity of 100%. Conclusion: Overall, our results indicate that PLP1 may be a potential diagnostic biomarker for uterine fibroids.

Association between fertility treatments and breast cancer risk in women with a family history or BRCA mutations: a systematic review and meta-analysis.

Women with hereditary breast cancer factors are more likely to be infertile and tend to receive fertility treatments. The safety of fertility treatments that contain hormone-related medications for ovarian stimulation has gained wide attention; however, evidence of the safety of fertility treatments is limited. This study aims to assess the association between fertility treatments and the incidence rate of breast cancer in women with a family history of breast cancer or BRCA mutations. A literature search was conducted in PubMed, Cochrane Library, and Embase. Studies concerning the effect of fertility treatments on breast cancer risk in genetically susceptible women were included. The fixed and random effects models were used to estimate the summary effects. Risk Of Bias In Non-randomized Studies - of Interventions instrument was used to assess the risk of bias in the included studies. A total of 5,282 studies were screened. Five cohort studies and three case-control studies were included. Breast cancer risk was not significantly increased by fertility treatments in general genetically susceptible women [pooled odds ratio (OR) 1.18, 95% confidence interval (CI) 0.96-1.45], women with a family history of breast cancer (pooled OR 1.35, 95% CI 0.97-1.89), or women with BRCA mutations (pooled OR 1.02, 95% CI 0.74-1.4). In subgroup analyses, there was no significant increase in breast cancer risk whether in BRCA1 mutation carriers (pooled OR 1.18, 95% CI 0.81-1.72), BRCA2 mutation carriers (pooled OR 0.54, 95% CI 0.09-3.34), or in the women treated with in vitro fertilization (pooled OR 0.75, 95% CI 0.51-1.1), clomiphene citrate (pooled OR 1.07, 95% CI 0.78-1.45) or gonadotropins (pooled OR 1.32, 95% CI 0.8-2.18). This is the first meta-analysis concerning the impact of fertility treatments on breast cancer risk in genetically susceptible women. Despite the finding that fertility treatment did not significantly increase breast cancer risk in genetically susceptible women, large prospective cohorts with more detailed information are required. Further investigations are needed to explore subtypes of breast cancer, genetic background of hormone-related breast cancer, and the association between BRCA mutations and the incidence of hormone receptor-positive breast cancer. Registration number: PROSPERO(CRD42021281336).

Transcriptome-wide N6-methyladenine methylation in granulosa cells of women with decreased ovarian reserve.

BACKGROUND: The emerging epitranscriptome plays an essential role in female fertility. As the most prevalent internal mRNA modification, N6-methyladenine (m6A) methylation regulate mRNA fate and translational efficiency. However, whether m6A methylation was involved in the aging-related ovarian reserve decline has not been investigated. Herein, we performed m6A transcriptome-wide profiling in the ovarian granulosa cells of younger women (younger group) and older women (older group). RESULTS: m6A methylation distribution was highly conserved and enriched in the CDS and 3'UTR region. Besides, an increased number of m6A methylated genes were identified in the older group. Bioinformatics analysis indicated that m6A methylated genes were enriched in the FoxO signaling pathway, adherens junction, and regulation of actin cytoskeleton. A total of 435 genes were differently expressed in the older group, moreover, 58 of them were modified by m6A. Several specific genes, including BUB1B, PHC2, TOP2A, DDR2, KLF13, and RYR2 which were differently expressed and modified by m6A, were validated using qRT-PCR and might be involved in the decreased ovarian functions in the aging ovary. CONCLUSIONS: Hence, our finding revealed the transcriptional significance of m6A modifications and provide potential therapeutic targets to promote fertility reservation for aging women.

Effects of the Zishen Yutai Pill Compared With Placebo on Live Births Among Women in a Fresh Embryo Transfer Cycle: A Randomized Controlled Trial.

OBJECTIVE: To assess the efficacy of the Zishen Yutai Pill compared with placebo on live birth rates among women after fresh embryo transfer cycles. METHODS: We conducted a double-blind, multicenter, placebo-controlled, randomized trial to investigate whether administration of the Zishen Yutai Pill would improve pregnancy outcomes among women undergoing fresh embryo transfer after in vitro fertilization or intracytoplasmic sperm injection. The primary outcome was live birth rate. Secondary outcomes were rates of implantation, biochemical pregnancy, clinical pregnancy, pregnancy loss, cycle cancellation, and maternal, fetal, and neonatal complications. A total sample size of 2,265 women (1:1 in two groups) was used to detect a live birth rate difference between the Zishen Yutai Pill and placebo. Participants were enrolled and randomized to receive 5 g of the Zishen Yutai Pill or placebo orally, three times per day during the study. RESULTS: Recruitment was completed between April 2014 and June 2017, with 2,580 patients screened. Two thousand two hundred sixty-five patients were randomized: 1,131 to the Zishen Yutai Pill and 1,134 to placebo. Characteristics were similar between groups. In intention-to-treat analysis, the rates of live birth in the Zishen Yutai Pill (ZYP) group and placebo group were 26.8% and 23.0% (rate ratio [RR], 1.16; 95% CI 1.01-1.34; P=.038), respectively. The implantation rates were 36.8% and 32.6% in the ZYP and placebo groups, respectively (RR 1.13; 95% CI 1.01-1.25; P=.027). The biochemical pregnancy rate for the ZYP group was 35.5% compared with 31.1% in the placebo group (RR 1.14; 95% CI 1.02-1.28; P=.026). The rates of clinical pregnancy in the ZYP and placebo groups were 31.2% compared with 27.3%, respectively (RR 1.14; 95% CI 1.00-1.30; P=.043). There were no significant between-group differences in the rates of pregnancy loss, maternal, or neonatal complications (all P>.05). CONCLUSION: The Zishen Yutai Pill increased the rate of live birth after fresh embryo transfer compared with placebo. CLINICAL TRIAL REGISTRATION: Chictr.org.cn, Chictr-TRC-14004494.

Discriminative Style Learning for Cross-Domain Image Captioning.

The cross-domain image captioning, which is trained on a source domain and generalized to other domains, usually faces the large domain shift problem. Although prior work has attempted to leverage both paired source and unpaired target data to minimize this shift, the performance is still unsatisfactory. One main reason lies in the large discrepancy in language expression between two domains, where diverse language styles are adopted to describe an image from different views, resulting in different semantic descriptions for an image. To tackle this problem, this paper proposes a Style-based Cross-domain Image Captioner (SCIC) which incorporates the discriminative style information into the encoder-decoder framework, and interprets an image as a special sentence according to external style instructions. Technically, we design a novel "Instruction-based LSTM", which adds the instruct gate to collect a style instruction, and then outputs a specified format according to that instruction. Two objectives are designed to train I-LSTM: 1) generating correct image descriptions and 2) generating correct styles, thus the model is expected to accurately capture the semantic meanings of an image by the special caption as well as understand the syntactic structure of the caption. We use MS-COCO as the source domain, and Oxford-102, CUB-200, Flickr30k as the target domains. Experimental results demonstrate that our model consistently outperforms the previous methods, and the style information incorporating with I-LSTM significantly improves the performance, with 5% CIDEr improvements at least on all datasets.

Auto-Encoding and Distilling Scene Graphs for Image Captioning.

We propose scene graph auto-encoder (SGAE) that incorporates the language inductive bias into the encoder-decoder image captioning framework for more human-like captions. Intuitively, we humans use the inductive bias to compose collocations and contextual inferences in discourse. For example, when we see the relation "a person on a bike", it is natural to replace "on" with "ride" and infer "a person riding a bike on a road" even when the "road" is not evident. Therefore, exploiting such bias as a language prior is expected to help the conventional encoder-decoder models reason as we humans and generate more descriptive captions. Specifically, we use the scene graph-a directed graph ( G) where an object node is connected by adjective nodes and relationship nodes-to represent the complex structural layout of both image ( I) and sentence ( S). In the language domain, we use SGAE to learn a dictionary set ( D) that helps reconstruct sentences in the S→ GS → D → S auto-encoding pipeline, where D encodes the desired language prior and the decoder learns to caption from such a prior; in the vision-language domain, we share D in the I→ GI → D → S pipeline and distill the knowledge of the language decoder of the auto-encoder to that of the encoder-decoder based image captioner to transfer the language inductive bias. In this way, the shared D provides hidden embeddings about descriptive collocations to the encoder-decoder and the distillation strategy teaches the encoder-decoder to transform these embeddings to human-like captions as the auto-encoder. Thanks to the scene graph representation, the shared dictionary set, and the Knowledge Distillation strategy, the inductive bias is transferred across domains in principle. We validate the effectiveness of SGAE on the challenging MS-COCO image captioning benchmark, where our SGAE-based single-model achieves a new state-of-the-art 129.6 CIDEr-D on the Karpathy split, and a competitive 126.6 CIDEr-D (c40) on the official server, which is even comparable to other ensemble models. Furthermore, we validate the transferability of SGAE on two more challenging settings: transferring inductive bias from other language corpora and unpaired image captioning. Once again, the results of both settings confirm the superiority of SGAE. The code is released in https://github.com/yangxuntu/SGAE.

Context-Aware Visual Policy Network for Fine-Grained Image Captioning.

With the maturity of visual detection techniques, we are more ambitious in describing visual content with open-vocabulary, fine-grained and free-form language, i.e., the task of image captioning. In particular, we are interested in generating longer, richer and more fine-grained sentences and paragraphs as image descriptions. Image captioning can be translated to the task of sequential language prediction given visual content, where the output sequence forms natural language description with plausible grammar. However, existing image captioning methods focus only on language policy while not visual policy, and thus fail to capture visual context that are crucial for compositional reasoning such as object relationships (e.g., "man riding horse") and visual comparisons (e.g., "small(er) cat"). This issue is especially severe when generating longer sequences such as a paragraph. To fill the gap, we propose a Context-Aware Visual Policy network (CAVP) for fine-grained image-to-language generation: image sentence captioning and image paragraph captioning. During captioning, CAVP explicitly considers the previous visual attentions as context, and decides whether the context is used for the current word/sentence generation given the current visual attention. Compared against traditional visual attention mechanism that only fixes a single visual region at each step, CAVP can attend to complex visual compositions over time. The whole image captioning model-CAVP and its subsequent language policy network-can be efficiently optimized end-to-end by using an actor-critic policy gradient method. We have demonstrated the effectiveness of CAVP by state-of-the-art performances on MS-COCO and Stanford captioning datasets, using various metrics and sensible visualizations of qualitative visual context.

Learning to Compose and Reason with Language Tree Structures for Visual Grounding.

Grounding natural language in images, such as localizing "the black dog on the left of the tree", is one of the core problems in artificial intelligence, as it needs to comprehend the fine-grained language compositions. However, existing solutions merely rely on the association between the holistic language features and visual features, while neglect the nature of composite reasoning implied in the language. In this paper, we propose a natural language grounding model that can automatically compose a binary tree structure for parsing the language and then perform visual reasoning along the tree in a bottom-up fashion. We call our model RvG-Tree: Recursive Grounding Tree, which is inspired by the intuition that any language expression can be recursively decomposed into two constituent parts, and the grounding confidence score can be recursively accumulated by calculating their grounding scores returned by the two sub-trees.RvG-Tree can be trained end-to-end by using the Straight-Through Gumbel-Softmax estimator that allows the gradients from the continuous score functions passing through the discrete tree construction. Experiments on several benchmarks show that our model achieves the state-of-the-art performance with more explainable reasoning.

Effect of electro-acupuncture on ovarian function of women with diminished ovarian reserve: study protocol for a randomized controlled trial.

BACKGROUND: Diminished ovarian reserve (DOR) is the precursor state of ovarian failure and can cause the decline of women's reproductive function. DOR also leads to poor outcome of in vitro fertilization and embryo transfer (IVF-ET) by affecting the oocytes, high qualified embryo rate, pregnancy rate, etc. Some studies have demonstrated that acupuncture can improve ovarian function. But to date, there is limited evidence indicating that acupuncture or electro-acupuncture is efficient to DOR. This trial aims to evaluate the efficiency and safety of electro-acupuncture for the ovarian function and the following outcome of IVF-ET in DOR patients. METHODS: This will be a multicenter randomized controlled clinical trial. A total of more than 338 women with DOR will be randomly allocated to treatment and control groups in 1:1 ratio receiving acupuncture before undergoing IVF-ET. The primary outcome will be the clinical pregnancy rate per cycle of IVF-ET after acupuncture. The secondary outcomes will be ovarian reserve function, outcomes of IVT-ET, blood biochemical index, Massachusetts General Hospital Acupuncture Sensation Scale (MASS), scores from the self-rating anxiety and depression scale, quality of life, and pregnancy outcomes. The safety of acupuncture will also be assessed. DISCUSSION: The results of this trial may provide high-quality evidence regarding the effectiveness of electro-acupuncture in the treatment of DOR and following outcomes of IVF-ET. This will also help patients with DOR and their physicians by offering a new treatment option. TRIAL REGISTRATION: Chinese Clinical Trial Registry ChiCTR1900024626 . Registered on 19 July 2019.