Prognostic value of pretreatment systemic immune-inflammation index in patients with endometrial cancer: a meta-analysis.

Background: The present work focused on evaluating the systemic immune-inflammation index (SII) for its role in predicting endometrial cancer (EC) patient prognosis by meta-analysis. Methods: SII's role in predicting the prognosis of EC patients was analyzed by calculating combined hazard ratios (HRs) and 95% CIs. Results: As revealed by combined analysis, an increased SII predicted poor overall survival (HR = 2.01; 95% CI = 1.58-2.57; p < 0.001) as well as inferior progression-free survival (HR = 1.87; 95% CI = 1.36-2.58; p < 0.001) of EC. Conclusion: An increased SII score significantly predicted poor overall survival and progression-free survival in subjects with EC. The SII is suitable for predicting short- and long-term prognoses of patients with EC.

Catalytic activation of persulfate by nanoscale zero-valent iron-derived supported boron-doped porous carbon for bisphenol A degradation.

In this study, boron-doped porous carbon materials (BCs) with high surface areas were synthesized employing coffee grounds as carbon source and sodium bicarbonate and boric acid as precursors; afterward, nanoscale zero-valent iron (nZVI) and BCs composites (denoted as nZVI@BCs) were further prepared through reduction of FeSO4 by NaBH4 along with stirring. The performance of the nZVI@BCs for activating persulfate (PS) was evaluated for the degradation of bisphenol A (BPA). In comparison with nZVI@Cs/PS, nZVI@BCs/PS could greatly promote the degradation and mineralization of BPA via both radical and non-radical pathways. On the one hand, electron spin resonance and radical quenching studies represented that •OH, SO4•-, and O2•- were mainly produced in the nZVI@BCs/PS system for BPA degradation. On the other hand, the open circuit voltages of nZVI@BCs and nZVI@Cs in different systems indicated that non-radical pathway still existed in our system. PS could grab the unstable unpaired electron on nZVI@BCs to form a carbon material surface-confined complex ([nZVI@BCs]*) with a high redox potential, then accelerate BPA removal efficiency via direct electron transfer. Furthermore, the performances and mechanisms for BPA degradation were examined by PS activation with nZVI@BC composites at various conditions including dosages of nZVI@BCs, BPA and PS, initially pH value, temperature, common anions, and humid acid. Therefore, this study provides a novel insight for development of high-performance carbon catalysts toward environmental remediation.

Expression profiles of host miRNAs and circRNAs and ceRNA network during Toxoplasma gondii lytic cycle.

Toxoplasma gondii is an opportunistic protozoan parasite that is highly prevalent in the human population and can lead to adverse health consequences in immunocompromised patients and pregnant women. Noncoding RNAs, such as microRNAs (miRNAs) and circular RNAs (circRNAs), play important regulatory roles in the pathogenesis of many infections. However, the differentially expressed (DE) miRNAs and circRNAs implicated in the host cell response during the lytic cycle of T. gondii are unknown. In this study, we profiled the expression of miRNAs and circRNAs in human foreskin fibroblasts (HFFs) at different time points after T. gondii infection using RNA sequencing (RNA-seq). We identified a total of 7, 7, 27, 45, 70, 148, 203, and 217 DEmiRNAs and 276, 355, 782, 1863, 1738, 6336, 1229, and 1680 DEcircRNAs at 1.5, 3, 6, 9, 12, 24, 36, and 48 h post infection (hpi), respectively. Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway enrichment analyses revealed that the DE transcripts were enriched in immune response, apoptosis, signal transduction, and metabolism-related pathways. These findings provide new insight into the involvement of miRNAs and circRNAs in the host response to T. gondii infection.

Clinical value of FAT1 mutations to indicate the immune response in colorectal cancer patients.

Immunotherapy is currently approved for CRC whose tumors have high MSI-H. To find additional biomarkers for immunotherapy in CRC, targeted sequencing was performed on tumor tissues from a discovery cohort of 161 CRC patients. Validation cohorts from the cBioPortal were also used for survival and tumor cell infiltration analyses. The FAT1-mutated CRC group often co-occurred with MSI events and displayed a higher tumor mutational burden compared to the FAT1 wild-type CRC. Overall survival was higher in patients with FAT1 mutations than in patients with wild type FAT1. The altered PI3K-AKT pathway and immune pathways were enriched in the FAT1-mutated CRC. A higher infiltration rate of immune cells including CD4+ T cells, CD8+ T cells, macrophages M1 and regulatory T cells were also observed in the colorectal tumors with FAT1 mutation compared to tumors with wild type FAT1. The results showed that CRC patients with FAT1 mutations exhibited an immunotherapy-favorable profile.

Characterization, antioxidant and antitumor activities of phenolic compounds from Amomum villosum Lour.

Amomum villosum Lour. (A. villosum), known as Sharen in China, is widely used for culinary and medicinal purposes due to containing a diverse set of bioactive compounds. In this study, the optimum ethanol extraction process was optimized and the composition and biological activities (antioxidant and antitumor) of five different fractions (dichloromethane, petroleum ether, ethyl acetate, n-butanol and H2O) extracted from the ethanol extract of A. villosum were investigated. The results showed that the optimal extraction conditions were extraction temperature 80°C, extraction time 120 min, ethanol concentration 40% and solid-liquid ratio 1:25 g/mL. Moreover, 35 bioactive compounds were successfully identified by UPLC-ESI-QTOF-MS/MS from five factions for the first time, including 12 phenolic acids and derivatives, 2 organic acids, 12 flavonoids and derivatives, 2 oxylipins and 7 proanthocyanidins. Among them, ethyl acetate fraction (Fr-EtOAc) exhibited the highest content of total phenolic (374.01 mg GAE/g DW) and flavonoid (93.11 mg RE/g DW), where vanillic acid, catechin, epicatechin and protocatechuic acid were the predominant phenolic compounds that accounting for 81.65% of the quantified bioactive compounds. In addition, Fr-EtOAc demonstrated excellent total antioxidant activity (IC50 of DPPH and ABTS assays were 0.23, 0.08 mg/mL, respectively, and FRAP assay was 322.91 mg VCE/100 g DW) and antitumor activity (1,000 µg/mL, 79.04% inhibition rate). The results could provide guidance for the industrial production and application of A. villosum.

Efficacy and Safety of Nerve Block for Postoperative Analgesia in Patients Undergoing Breast Cosmetic Surgery: A Systematic Review and Meta-analysis of Randomized Controlled Trials.

BACKGROUND: To evaluate the postoperative analgesic efficacy and safety of nerve block (NB) in patients undergoing breast surgery for cosmetic purposes. METHODS: PubMed, Web of Science, Embase and Cochrane Libraries were searched from inception to September 2022, to identify all eligible randomized controlled trials (RCTs). Continuous data are presented as mean difference (MD) with 95% confidence intervals (CI), whereas dichotomous data are provided as odds ratios (OR) with 95% CI. This meta-analysis was performed in RevMan 5.4. RESULTS: A total of 10 RCTs with 565 patients were meta-analyzed. Compared to the control group, the pain score of the NB group was significantly lower at postoperative 2, 3-4, 6-8, 12-16 and 24 h. Opioid consumption in the first postoperative 24 h was significantly lower in the NB group (MD = - 9.02, 95% CI - 14.29 to - 3.75, P < 0.05), I2 = 95%). In addition, the NB group showed a prolonged time to first postoperative analgesic requirement (MD = 43.15, 95% CI 4.74-81.56, P < 0.05, I2 = 96%), decreased incidence of additional postoperative analgesia (OR 0.14, 95% CI 0.07-0.28, P < 0.05, I2 = 0%) and reduced incidence of postoperative nausea or vomiting (OR 0.33; 95% CI 0.22-0.48; P < 0.05; I2 = 0%). There was no significant difference in operation duration between the two groups. CONCLUSIONS: Nerve block is an effective and safe option for postoperative analgesia after breast cosmetic surgery. LEVEL OF EVIDENCE II: This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266 .

Population Genetic Structure of the Rubber Tree Powdery Mildew Pathogen (Erysiphe quercicola) from China.

In order to manage agricultural pathogens, it is crucial to understand the population structure underlying epidemics. Rubber tree powdery mildew, caused by Erysiphe quercicola, is a serious threat to rubber plantations worldwide, especially in subtropical environments including all rubber tree-growing regions in China. However, the population structure of the pathogen is uncertain. In this study, 16 polymorphic microsatellite markers were used to genotype powdery mildew samples from the main rubber tree-growing regions including Yunnan (YN), Hainan (HN), western Guangdong (WG), and eastern Guangdong (EG). YN had higher genotypic diversity (Simpson's indices), genotypic evenness, Nei's gene diversity, allelic richness, and private allelic richness than the other regions. Cluster analysis, discriminant analysis of principal components, pairwise divergence, and shared multilocus genotype analyses all showed that YN differed significantly from the other regions. The genetic differentiation was small among the other three regions (HN, WG, and EG). Analysis of molecular variance indicated that the variability among regions accounted for 22.37% of the total variability. Genetic differentiation was significantly positively correlated (Rxy = 0.772, P = 0.001) with geographic distance. Linkage equilibrium analysis suggested possible occurrence of sexual recombination although asexual reproduction predominates in E. quercicola. The results suggested that although significant genetic differentiation of E. quercicola occurred between YN and the other regions, pathogen populations from the other three regions lacked genetic differentiation.

Integrative Analysis of Metabolome and Transcriptome Provides Insights into the Mechanism of Flower Induction in Pineapple (Ananas comosus (L.) Merr.) by Ethephon.

Exogenous ethylene is commonly utilized to initiate flower induction in pineapple (Ananas comosus (L.) Merr.). However, the molecular mechanisms and metabolic changes involved are not well understood. In this study, we explored the genetic network and metabolic shifts in the 'Comte de Paris' pineapple variety during ethylene-induced flowering. This was achieved through an integrative analysis of metabolome and transcriptome profiles at vegetative shoot apexes (0 d after ethephon treatment named BL_0d), the stage of bract primordia (8 d after ethephon treatment named BL_8d), stage of flower primordia (18 d after ethephon treatment named BL_18d), and the stage of stopped floret differentiation (34 d after ethephon treatment named BL_34d). We isolated and identified 804 metabolites in the pineapple shoot apex and inflorescence, categorized into 24 classes. Notably, 29, 31, and 46 metabolites showed significant changes from BL_0d to BL_8d, BL_8d to BL_18d, and BL_18d to BL_34d, respectively. A marked decrease in indole was observed, suggesting its role as a characteristic metabolite during flower induction. Transcriptomic analysis revealed 956, 1768, and 4483 differentially expressed genes (DEGs) for BL_0d vs. BL_8d, BL_8d vs. BL_18d, and BL_18d vs. BL_34d, respectively. These DEGs were significantly enriched in carbohydrate metabolism and hormone signaling pathways, indicating their potential involvement in flower induction. Integrating metabolomic and transcriptomic data, we identified several candidate genes, such as Agamous-Like9 (AGL9), Ethylene Insensitive 3-like (ETIL3), Apetala2 (AP2), AP2-like ethylene-responsive transcription factor ANT (ANT), and Sucrose synthase 2 (SS2), that play potentially crucial roles in ethylene-induced flower induction in pineapple. We also established a regulatory network for pineapple flower induction, correlating metabolites and DEGs, based on the Arabidopsis thaliana pathway as a reference. Overall, our findings offer a deeper understanding of the metabolomic and molecular mechanisms driving pineapple flowering.

iTRAQ-Based Phosphoproteomic Analysis Exposes Molecular Changes in the Small Intestinal Epithelia of Cats after Toxoplasma gondii Infection.

Toxoplasma gondii, an obligate intracellular parasite, has the ability to invade and proliferate within most nucleated cells. The invasion and destruction of host cells by T. gondii lead to significant changes in the cellular signal transduction network. One important post-translational modification (PTM) of proteins is phosphorylation/dephosphorylation, which plays a crucial role in cell signal transmission. In this study, we aimed to investigate how T. gondii regulates signal transduction in definitive host cells. We employed titanium dioxide (TiO2) affinity chromatography to enrich phosphopeptides in the small intestinal epithelia of cats at 10 days post-infection with the T. gondii Prugniuad (Pru) strain and quantified them using iTRAQ technology. A total of 4998 phosphopeptides, 3497 phosphorylation sites, and 1805 phosphoproteins were identified. Among the 705 differentially expressed phosphoproteins (DEPs), 68 were down-regulated and 637 were up-regulated. The bioinformatics analysis revealed that the DE phosphoproteins were involved in various cellular processes, including actin cytoskeleton reorganization, cell necroptosis, and MHC immune processes. Our findings confirm that T. gondii infection leads to extensive changes in the phosphorylation of proteins in the cat intestinal epithelial cells. The results of this study provide a theoretical foundation for understanding the interaction between T. gondii and its definitive host.

A Real-world Dataset and Benchmark For Foundation Model Adaptation in Medical Image Classification.

Foundation models, often pre-trained with large-scale data, have achieved paramount success in jump-starting various vision and language applications. Recent advances further enable adapting foundation models in downstream tasks efficiently using only a few training samples, e.g., in-context learning. Yet, the application of such learning paradigms in medical image analysis remains scarce due to the shortage of publicly accessible data and benchmarks. In this paper, we aim at approaches adapting the foundation models for medical image classification and present a novel dataset and benchmark for the evaluation, i.e., examining the overall performance of accommodating the large-scale foundation models downstream on a set of diverse real-world clinical tasks. We collect five sets of medical imaging data from multiple institutes targeting a variety of real-world clinical tasks (22,349 images in total), i.e., thoracic diseases screening in X-rays, pathological lesion tissue screening, lesion detection in endoscopy images, neonatal jaundice evaluation, and diabetic retinopathy grading. Results of multiple baseline methods are demonstrated using the proposed dataset from both accuracy and cost-effective perspectives.

Transcriptomic, 16S ribosomal ribonucleic acid and network pharmacology analyses shed light on the anticoccidial mechanism of green tea polyphenols against Eimeria tenella infection in Wuliangshan black-boned chickens.

BACKGROUND: Eimeria tenella is an obligate intracellular parasitic protozoan that invades the chicken cecum and causes coccidiosis, which induces acute lesions and weight loss. Elucidating the anticoccidial mechanism of action of green tea polyphenols could aid the development of anticoccidial drugs and resolve the problem of drug resistance in E. tenella. METHODS: We constructed a model of E. tenella infection in Wuliangshan black-boned chickens, an indigenous breed of Yunnan Province, China, to study the efficacy of green tea polyphenols against the infection. Alterations in gene expression and in the microbial flora in the cecum were analyzed by ribonucleic acid (RNA) sequencing and 16S ribosomal RNA (rRNA) sequencing. Quantitative real-time polymerase chain reaction was used to verify the host gene expression data obtained by RNA sequencing. Network pharmacology and molecular docking were used to clarify the interactions between the component green tea polyphenols and the targeted proteins; potential anticoccidial herbs were also analyzed. RESULTS: Treatment with the green tea polyphenols led to a reduction in the lesion score and weight loss of the chickens induced by E. tenella infection. The expression of matrix metalloproteinase 7 (MMP7), MMP1, nitric oxide synthase 2 and ephrin type-A receptor 2 was significantly altered in the E. tenella infection plus green tea polyphenol-treated group and in the E. tenella infection group compared with the control group; these genes were also predicted targets of tea polyphenols. Furthermore, the tea polyphenol (-)-epigallocatechin gallate acted on most of the targets, and the molecular docking analysis showed that it has good affinity with interferon induced with helicase C domain 1 protein. 16S ribosomal RNA sequencing showed that the green tea polyphenols had a regulatory effect on changes in the fecal microbiota induced by E. tenella infection. In total, 171 herbs were predicted to act on two or three targets in MMP7, MMP1, nitric oxide synthase 2 and ephrin type-A receptor 2. CONCLUSIONS: Green tea polyphenols can directly or indirectly regulate host gene expression and alter the growth of microbiota. The results presented here shed light on the mechanism of action of green tea polyphenols against E. tenella infection in chickens, and have implications for the development of novel anticoccidial products.

Dynamic RNA profiles in the small intestinal epithelia of cats after Toxoplasma gondii infection.

BACKGROUND: Felids are the only definitive hosts of Toxoplasma gondii. However, the biological features of the feline small intestine following T. gondii infection are poorly understood. We investigated the changes in the expression of RNAs (including mRNAs, long non-coding RNAs and circular RNAs) in the small intestinal epithelia of cats following T. gondii infection to improve our understanding of the life cycle of T. gondii and cat responses to T. gondii infection. METHODS: Fifteen cats were randomly assigned to five groups, and the infection groups were inoculated with 600 tissue cysts of the T. gondii Pru strain by gavage. The small intestinal epithelia of cats were collected at 6, 10, 14, and 30 days post infection (DPI). Using high-throughput RNA sequencing (RNA-seq), we investigated the changes in RNA expression. The expression levels of differentially expressed (DE) genes and non-coding RNAs (ncRNAs) identified by RNA-seq were validated by quantitative reverse transcription PCR (qRT-PCR). Differential expression was determined using the DESeq R package. RESULTS: In total, 207 annotated lncRNAs, 20,552 novel lncRNAs, 3342 novel circRNAs and 19,409 mRNAs were identified. Among these, 70 to 344 DE mRNAs, lncRNAs and circRNAs were detected, and the post-cleavage binding sites between 725 ncRNAs and 2082 miRNAs were predicted. Using the co-location method, we predicted that a total of 235 lncRNAs target 1044 protein-coding genes, while the results of co-expression analysis revealed that 174 lncRNAs target 2097 mRNAs. Pathway enrichment analyses of the genes targeted by ncRNAs suggested that most ncRNAs were significantly enriched in immune or diseases-related pathways. NcRNA regulatory networks revealed that a single ncRNA could be directly or indirectly regulated by multiple genes or ncRNAs that could influence the immune response of cats. Co-expression analysis showed that 242 circRNAs, mainly involved in immune responses, were significantly associated with T. gondii infection. In contrast, 1352 protein coding RNAs, mainly involved in nucleic acid process/repair pathways or oocyte development pathways, were negatively associated with T. gondii infection. CONCLUSIONS: This study is the first to reveal the expression profiles of circRNAs, lncRNAs and mRNAs in the cat small intestine following T. gondii infection and will facilitate the elucidation of the role of ncRNAs in the pathogenesis of T. gondii infection in its definitive host, thereby facilitating the development of novel intervention strategies against T. gondii infection in humans and animals.

Prognostic Role of Geriatric Nutritional Risk Index in Patients with Pancreatic Cancer: A Meta-Analysis.

Background: The Geriatric Nutritional Risk Index (GNRI) is used to assess the prognosis of patients with pancreatic cancer. Therefore, this meta-analysis was conducted to evaluate the association between the GNRI and prognosis in pancreatic cancer.Methods: We performed a pooled analysis of the hazard ratios (HRs) and 95% confidence intervals (CIs) of the GNRI for survival in pancreatic cancer. Using pooled odds ratios (ORs) and 95% CIs, we investigated the association between the GNRI and clinicopathological characteristics of pancreatic cancer.Results: Six studies were included in this meta-analysis, totaling 1,513 patients. A low GNRI was significantly associated with a poorer overall survival (OS) in the pooled results (HR, 1.95; 95% CI, 1.29-2.94; p = 0.002) in pancreatic cancer. However, GNRI was not significantly associated with progression-free survival (PFS) in pancreatic cancer (HR, 1.57; 95% CI, 0.90-2.73; p = 0.114). The pooled results indicated that a low GNRI was significantly associated with tumor location of pancreas head (OR, 2.18; 95% CI, 1.45-3.29; p < 0.001).Conclusions: This meta-analysis showed that low GNRI was significantly associated with poor OS but not with poor PFS in patients with pancreatic cancer. The GNRI is a novel and effective risk factor and a potential biomarker for the prognosis of pancreatic cancer.

Molecular Epidemiological Investigation of Cyclospora spp. in Holstein Cattle in Partial Areas of the Yunnan Province, China.

Cyclospora spp. is a food-borne intestinal protozoan, which is widely distributed in the world and poses the risk of zoonosis. In order to reveal the prevalence of Cyclospora spp. in Holstein cattle in partial areas of the Yunnan Province, 524 fresh fecal samples of Holstein cattle were collected from Dali, Kunming, Chuxiong, and Qujing in Yunnan Province. A nested PCR amplification of the small subunit (SSU) rRNA gene of Cyclospora spp. was carried out, and the products of the nested PCR were further analyzed by restriction fragment length polymorphism (RFLP) using Bsp E Ⅰ. The results of the present study showed that 13 samples were positive for Cyclospora spp., and the total infection rate of Cyclospora sp. was 2.48%. The infection of Cyclospora spp. was detected in Dali, Qujing, and Chuxiong. Chuxiong showed the highest infection rate (5.71%), and infection rate in Dali and Qujing was 2.19% and 3.16%, respectively. Interestingly, the infection of Cyclospora spp. was not detected in Kunming. The infection of Cyclospora spp. showed no significant differences among different regions (p > 0.05). Cyclospora sp. infection was detected in all ages and sexes, but the differences were not significant (p > 0.05). Sequencing and phylogenetic analysis showed that five Cyclospora spp. samples were closely related to the Cyclospora spp. of humans, and the others were closely related to the Cyclospora spp. of bovines. The results of the present study suggested that there was an infection of Cyclospora spp. in Holstein cattle in the Yunnan Province, and the Cyclospora spp. showed a risk of zoonosis. Thus, the prevention and control of Cyclospora spp. should be strengthened in the Yunnan Province, China. The results of this investigation provide data references for the further research of Cyclosporiasis in Holstein cattle in the Yunnan Province.

Nonsurgical Chin Augmentation Using Hyaluronic Acid: A Systematic Review of Technique, Satisfaction, and Complications.

BACKGROUND: Nonsurgical chin augmentation using hyaluronic acid (HA) has become a common procedure in cosmetic practices. This is offered to patients that prefer a nonsurgical, temporary method to correcting underdeveloped or retruded chin and restoring the volume loss. This systematic review highlights the main HA injection technique and associated patient satisfaction and complications of chin augmentation to further guide practitioners. METHODS: A systematic review was performed according to PRISMA guidelines. PubMed, Embase, and Web of Science were searched using the appropriate keywords. Data collected from each study included patient satisfaction and complications, in addition to injection protocol and technique. RESULTS: A total of 1305 studies were found based on search criteria. After full-text screening for inclusion and exclusion criteria, 8 studies were included. A total of 917 patients underwent HA chin augmentation, with different injection protocols. Most patients were satisfied with the results, and there were only 2 relatively major complications reported. The most common adverse events were local responses at the injection sites (swelling, bruising, pain, redness, and itching). There were no reports of vascular complications. CONCLUSIONS: HA filler is an effective temporary method to correct chin retraction and absorption for chin augmentation, with a high degree of patient satisfaction and a low risk of severe complications. LEVEL OF EVIDENCE III: This journal requires that authors assign a level of evidence to each article. For a full description of these Evidence-Based Medicine ratings, please refer to the Table of Contents or the online Instructions to Authors www.springer.com/00266 .

Modulation of long noncoding RNA (lncRNA) and messenger RNA (mRNA) expression in the liver of Beagle dogs by Toxocara canis infection.

BACKGROUND: Long non-coding RNAs (lncRNAs) and messenger RNAs (mRNAs) play crucial roles in regulating various physiological and pathological processes. However, the role of lncRNAs and mRNAs in mediating the liver response during Toxocara canis infection remains incompletely understood. METHODS: In the present study, the expression profile of lncRNAs and mRNAs was investigated in the liver of Beagle dogs infected by T. canis using high-throughput RNA sequencing. RESULTS: Compared with the control groups, 876 differentially expressed (DE) lncRNAs and 288 DEmRNAs were identified at 12 h post-infection (hpi), 906 DElncRNAs and 261 DEmRNAs were identified at 24 hpi, and 876 DElncRNAs and 302 DEmRNAs were identified at 36 days post-infection (dpi). A total of 16 DEmRNAs (e.g. dpp4, crp and gnas) were commonly identified at the three infection stages. Enrichment and co-localization analyses identified several pathways involved in immune and inflammatory responses during T. canis infection. Some novel DElncRNAs, such as LNC_015756, LNC_011050 and LNC_011052, were also associated with immune and inflammatory responses. Also, LNC_005105 and LNC_005401 were associated with the secretion of anti-inflammatory cytokines, which may play a role in the healing of liver pathology at the late stage of infection. CONCLUSIONS: Our data provided new insight into the regulatory roles of lncRNAs and mRNAs in the pathogenesis of T. canis and improved our understanding of the contribution of lncRNAs and mRNAs to the immune and inflammatory response of the liver during T. canis infection.

A Holistically-Guided Decoder for Deep Representation Learning With Applications to Semantic Segmentation and Object Detection.

Both high-level and high-resolution feature representations are of great importance in various visual understanding tasks. To acquire high-resolution feature maps with high-level semantic information, one common strategy is to adopt dilated convolutions in the backbone networks to extract high-resolution feature maps, such as the dilatedFCN-based methods for semantic segmentation. However, due to many convolution operations are conducted on the high-resolution feature maps, such methods have large computational complexity and memory consumption. To balance the performance and efficiency, there also exist encoder-decoder structures that gradually recover the spatial information by combining multi-level feature maps from a feature encoder, such as the FPN architecture for object detection and the U-Net for semantic segmentation. Although being more efficient, the performances of existing encoder-decoder methods for semantic segmentation are far from comparable with the dilatedFCN-based methods. In this paper, we propose one novel holistically-guided decoder which is introduced to obtain the high-resolution semantic-rich feature maps via the multi-scale features from the encoder. The decoding is achieved via novel holistic codeword generation and codeword assembly operations, which take advantages of both the high-level and low-level features from the encoder features. With the proposed holistically-guided decoder, we implement the EfficientFCN architecture for semantic segmentation and HGD-FPN for object detection and instance segmentation. The EfficientFCN achieves comparable or even better performance than state-of-the-art methods with only 1/3 of their computational costs for semantic segmentation on PASCAL Context, PASCAL VOC, ADE20K datasets. Meanwhile, the proposed HGD-FPN achieves higher mean Average Precision (mAP) when integrated into several object detection frameworks with ResNet-50 encoding backbones.

Token Sparsification for Faster Medical Image Segmentation.

Can we use sparse tokens for dense prediction, e.g., segmentation? Although token sparsification has been applied to Vision Transformers (ViT) to accelerate classification, it is still unknown how to perform segmentation from sparse tokens. To this end, we reformulate segmentation as a sparse encoding → token completion → dense decoding (SCD) pipeline. We first empirically show that naïvely applying existing approaches from classification token pruning and masked image modeling (MIM) leads to failure and inefficient training caused by inappropriate sampling algorithms and the low quality of the restored dense features. In this paper, we propose Soft-topK Token Pruning (STP) and Multi-layer Token Assembly (MTA) to address these problems. In sparse encoding, STP predicts token importance scores with a lightweight sub-network and samples the topK tokens. The intractable topK gradients are approximated through a continuous perturbed score distribution. In token completion, MTA restores a full token sequence by assembling both sparse output tokens and pruned multi-layer intermediate ones. The last dense decoding stage is compatible with existing segmentation decoders, e.g., UNETR. Experiments show SCD pipelines equipped with STP and MTA are much faster than baselines without token pruning in both training (up to 120% higher throughput) and inference (up to 60.6% higher throughput) while maintaining segmentation quality. Code is available here: https://github.com/cvlab-stonybrook/TokenSparse-for-MedSeg.

Asynchronous feature regularization and cross-modal distillation for OCT based glaucoma diagnosis.

Glaucoma has become a major cause of vision loss. Early-stage diagnosis of glaucoma is critical for treatment planning to avoid irreversible vision damage. Meanwhile, interpreting the rapidly accumulated medical data from ophthalmic exams is cumbersome and resource-intensive. Therefore, automated methods are highly desired to assist ophthalmologists in achieving fast and accurate glaucoma diagnosis. Deep learning has achieved great successes in diagnosing glaucoma by analyzing data from different kinds of tests, such as peripapillary optical coherence tomography (OCT) and visual field (VF) testing. Nevertheless, applying these developed models to clinical practice is still challenging because of various limiting factors. OCT models present worse glaucoma diagnosis performances compared to those achieved by OCT&VF based models, whereas VF is time-consuming and highly variable, which can restrict the wide employment of OCT&VF models. To this end, we develop a novel deep learning framework that leverages the OCT&VF model to enhance the performance of the OCT model. To transfer the complementary knowledge from the structural and functional assessments to the OCT model, a cross-modal knowledge transfer method is designed by integrating a designed distillation loss and a proposed asynchronous feature regularization (AFR) module. We demonstrate the effectiveness of the proposed method for glaucoma diagnosis by utilizing a public OCT&VF dataset and evaluating it on an external OCT dataset. Our final model with only OCT inputs achieves the accuracy of 87.4% (3.1% absolute improvement) and AUC of 92.3%, which are on par with the OCT&VF joint model. Moreover, results on the external dataset sufficiently indicate the effectiveness and generalization capability of our model.

Clinical characteristics and prognosis of patient with leptospirosis: A multicenter retrospective analysis in south of China.

Purpose: Leptospirosis is a zoonotic disease caused by pathogenic spirochetes of the genus Leptospira. However, there is currently no consensual definition or diagnostic criteria for severe and different forms of leptospirosis. Therefore, more insight on clinical manifestations, risk factors, and outcomes of leptospirosis is warranted. The identification of leptospirosis with distinct clinical manifestations and prognosis in our population. Methods: Multiple correspondence analysis and hierarchical classification on principal components were presented to identify different clinical types of leptospirosis. The outcomes were clinical phenotypes, laboratory and imaging findings, and prognosis. Results: The 95 enrolled patients had median values of 54.0 years (39.0-65.0) for age, 9.0 (7.0-14.0) for total hospital stay lengths, of whom 86.3% was male and 40.0% was transferred to ICU. Three clinical types were distinguished: mild leptospirosis (n=43, 45.3%) with less organ dysfunction and shorter hospital stays; respiratory leptospirosis (n=28, 29.5%) with hemoptysis, and respiratory and circulatory failure; and hepato-renal leptospirosis (n=24, 25.3%) with worst liver and kidney dysfunction. Total hospital mortality was 15.8% and was associated with dyspnea and high levels of neutrophil counts. Conclusions: The identification of leptospirosis with distinct clinical manifestations and prognosis in our population may assist clinicians to distinguish leptospirosis-like disease. Moreover, dyspnea and neutrophil count were found to be independent risk factors for severe leptospirosis progression.