Clinical significance of 18F-FDG-PET/CT for detection of incidental pre-malignant and malignant colonic lesions: correlation with colonoscopic and histopathological results.

BACKGROUND: Incidental colorectal fluorodeoxyglucose (FDG) uptake, observed during positron emission tomography/computed tomography (PET/CT) scans, attracts particular attention due to its potential to represent both benign and pre-malignant/malignant lesions. Early detection and excision of these lesions are crucial for preventing cancer development and reducing mortality. This research aims to evaluate the correlation between incidental colorectal FDG uptake on PET/CT with colonoscopic and histopathological results. METHODS: Retrospective analysis was performed on data from all patients who underwent PET/CT between December 2019 and December 2023 in our hospital. The study included 79 patients with incidental colonic FDG uptake who underwent endoscopy. Patient characteristics, imaging parameters, and the corresponding colonoscopy and histopathological results were studied. A comparative analysis was performed among the findings from each of these modalities. The optimal cut-off value of SUVmax for 18F-FDG PET/CT diagnosis of premalignant and malignant lesions was determined by receiver operating characteristic (ROC) curves. The area under the curve (AUC) of SUVmax and the combined parameters of SUVmax and colonic wall thickening (CWT) were analyzed. RESULTS: Among the 79 patients with incidental colorectal FDG uptake, histopathology revealed malignancy in 22 (27.9%) patients and premalignant polyps in 22 (27.9%) patients. Compared to patients with benign lesions, patients with premalignant and malignant lesions were more likely to undergo a PET/CT scan for primary evaluation (p = 0.013), and more likely to have focal GIT uptake (p = 0.001) and CWT (p = 0.001). A ROC curve analysis was made and assesed a cut-off value of 7.66 SUVmax (sensitivity: 64.9% and specificity: 82.4%) to distinguish premalignant and malignant lesions from benign lesions. The AUCs of the SUVmax and the combined parameters of SUVmax and CWT were 0.758 and 0.832 respectively. CONCLUSION: For patients undergo PET/CT for primary evaluation, imaging features of colorectal focal FDG uptake and CWT were more closely associated with premalignant and malignant lesions. The SUVmax helps determine benign and premalignant/malignant lesions of the colorectum. Moreover, the combination of SUVmax and CWT parameters have higher accuracy in estimating premalignant and malignant lesions than SUVmax.

Sublayer-Sulfur-Vacancy-Induced Charge Redistribution of FeCuS Nanoflower Awakening Alkaline Hydrogen Evolution.

Advancing the progress of sustainable and green energy technologies requires the improvement of valid electrocatalysts for the hydrogen evolution reaction (HER). Reconfiguring charge distribution through heteroatom doping-induced vacancy serves as an effective approach to implement high performance for HER catalysts. Here, we successfully fabricated Fe-doped CuS (FeCuS) with the sublayer sulfur vacancy to judge its HER performance and dissect the activity origins. Density functional theory calculation further elucidates that the primary factor contributing to the heightened HER activity is that the sublayer sulfur vacancies awaken the charge redistribution. In addition to effectively decreasing the energy barrier associated with the Volmer step, it modulates the adsorption/desorption capacity of H*. As a result, its intrinsic activity for the HER has significantly increased. Concretely, the obtained FeCuS displays an excellent catalytic performance, whose Tafel slope is only 59 mV dec-1 and the overpotential (at 10 mA cm-2) is as low as 71 mV in an alkaline environment, surpassing the majority of previously documented catalysts in scientific literature. This work shows that the construction of sublayer sulfur vacancies by Fe doping can achieve the charge redistribution and precise tuning of electronic structure; thereby, the inert CuS can be transformed into highly efficient electrocatalysts.

CT radiomics based on the peritumoral adipose region of gastric adenocarcinoma for preoperative prediction of lymph node metastasis.

PURPOSE: To construct and validate CT radiomics model based on the peritumoral adipose region of gastric adenocarcinoma to preoperatively predict lymph node metastasis (LNM). METHODS AND METHODS: 293 consecutive gastric adenocarcinoma patients receiving radical gastrectomy with lymph node dissection in two medical institutions were stratified into a development set (from Institution A, n = 237), and an external validation set (from Institution B, n = 56). Volume of interest of peritumoral adipose region was segmented on preoperative portal-phase CT images. The least absolute shrinkage and selection operator method and stepwise logistic regression were used to select features and build radiomics models. Manual classification was performed according to routine CT characteristics. A classifier incorporating the radiomics score and CT characteristics was developed for predicting LNM. Area under the receiver operating characteristic curve (AUC) was used to show discrimination between tumors with and without LNM, and the calibration curves and Brier score were used to evaluate the predictive accuracy. Violin plots were used to show the distribution of radiomics score. RESULTS: AUC values of radiomics model to predict LNM were 0.938, 0.905, and 0.872 in the training, internal test, and external validation sets, respectively, higher than that of manual classification (0.674, all P values < 0.01). The radiomics score of the positive LNM group were higher than that of the negative group in all sets (both P-values < 0.001). The classifier showed no improved predictive power compared with the radiomics signature alone with AUC values of 0.916 and 0.872 in the development and external validation sets, respectively. Multivariate analysis showed that radiomics score was an independent predictor. CONCLUSIONS: Radiomics model based on peritumoral adipose region could be a useful approach for preoperative LNM prediction in gastric adenocarcinoma.

Gastrodin alleviates the deterioration of depressive-like behavior and glucolipid metabolism promoted by chronic stress in type 2 diabetic mice.

The growing burden of psychological stress among diabetes patients has contributed to a rising incidence of depression within this population. It is of significant importance to conduct research on the impact of stress on diabetes patients and to explore potential pharmacological interventions to counteract the stress-induced exacerbation of their condition. Gastrodin is a low molecular weight bioactive compound extracted from the rhizome of Gastrodiae elata Blume, and it may be a preventive strategy for diabetes and a novel treatment for depression symptoms. However, its relevant pharmacological mechanisms for protecting against the impacts of psychological stress in diabetic patients are unclear. In this study, we performed 5 weeks CUMS intervention and simultaneously administered gastrodin (140 mg/kg, once daily) on T2DM mice, to investigate the potential protective effects of gastrodin. The protective effect of gastrodin was evaluated by behavioral tests, biochemical analysis, histopathological examination, RT-qPCR and gut microbiota analysis. We found that the depressive-like behavior and glucolipid metabolism could be deteriorated by chronic stress in type 2 diabetic mice, while gastrodin showed a protective effect against these exacerbations by regulating HPA hormones, activating FXR and Cyp7a1, reducing inflammatory and oxidative stress responses, and regulating ileal gut microbiota abundance. Gastrodin might be a potential therapeutic agent for mitigating the deterioration of diabetes conditions due to chronic stress.

A Systematic Review and Meta-Analysis of MRI Radiomics for Predicting Microvascular Invasion in Patients with Hepatocellular Carcinoma.

BACKGROUND: The prediction power of MRI radiomics for microvascular invasion (MVI) in patients with hepatocellular carcinoma (HCC) remains uncertain. OBJECTIVE: To investigate the prediction performance of MRI radiomics for MVI in HCC. METHODS: Original studies focusing on preoperative prediction performance of MRI radiomics for MVI in HCC, were systematically searched from databases of PubMed, Embase, Web of Science and Cochrane Library. Radiomics quality score (RQS) and risk of bias of involved studies were evaluated. Meta-analysis was carried out to demonstrate the value of MRI radiomics for MVI prediction in HCC. Influencing factors of the prediction performance of MRI radiomics were identified by subgroup analyses. RESULTS: 13 studies classified as type 2a or above according to the Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis statement were eligible for this systematic review and meta-analysis. The studies achieved an average RQS of 14 (ranging from 11 to 17), accounting for 38.9% of the total points. MRI radiomics achieved a pooled sensitivity of 0.82 (95%CI: 0.78 - 0.86), specificity of 0.79 (95%CI: 0.76 - 0.83) and area under the summary receiver operator characteristic curve (AUC) of 0.88 (95%CI: 0.84 - 0.91) to predict MVI in HCC. Radiomics models combined with clinical features achieved superior performances compared to models without the combination (AUC: 0.90 vs 0.85, P < 0.05). CONCLUSION: MRI radiomics has the potential for preoperative prediction of MVI in HCC. Further studies with high methodological quality should be designed to improve the reliability and reproducibility of the radiomics models for clinical application. The systematic review and meta-analysis was registered prospectively in the International Prospective Register of Systematic Reviews (No. CRD42022333822).

Preoperative CT radiomics of esophageal squamous cell carcinoma and lymph node to predict nodal disease with a high diagnostic capability.

PURPOSE: To develop CT radiomics models of resectable esophageal squamous cell carcinoma (ESCC) and lymph node (LN) to preoperatively identify LN+. MATERIALS AND METHODS: 299 consecutive patients with ESCC were enrolled in the study, 140 of whom were LN+ and 159 were LN-. Of the 299 patients, 249 (from the same hospital) were randomly divided into a training cohort (n = 174) and a test cohort (n = 75). The remaining 50 patients, from a second hospital, were assigned to an external validation cohort. In the training cohort, preoperative contrast-enhanced CT radiomics features of ESCC and LN were extracted, then integrated with clinical features to develop three models: ESCC, LN and combined. The performance of these models was assessed using area under receiver operating characteristic curve (AUC), and F-1 score, which were validated in both the test cohort and external validation cohort. RESULTS: An ESCC model was developed for the training cohort utilizing the 8 tumor radiomics features, and an LN model was constructed using 9 nodal radiomics features. A combined model was constructed using both ESCC and LN extracted features, in addition to cT stage and LN+ distribution. This combined model had the highest predictive ability among the three models in the training cohort (AUC = 0.948, F1-score = 0.878). The predictive ability was validated in both the test and external validation cohorts (AUC = 0.885 and 0.867, F1-score = 0.816 and 0.773, respectively). CONCLUSION: To preoperatively determine LN+, the combined model is superior to models of ESCC and LN alone.

Profile of the bile acid FXR-FGF15 pathway in the glucolipid metabolism disorder of diabetic mice suffering from chronic stress.

Background: Imbalances in bile acid (BA) synthesis and metabolism are involved in the onset of diabetes and depression in humans and rodents. However, the role of BAs and the farnesoid X receptor (FXR)/fibroblast growth factor (FGF) 15 signaling pathway in the development of diabetes and depression is still largely unknown. Therefore, we investigated the potential molecular mechanisms of BAs that may be associated with glucolipid metabolism disorders in diabetic mice subjected to chronic stress. Methods: The type 2 diabetes mellitus (T2DM) mouse model was induced by feeding mice a high-fat diet and administering an intraperitoneal injection of streptozotocin (STZ). The chronic unpredictable mild stress (CUMS) procedure was performed by introducing a series of mild stressors. Forty mice were randomly divided into the regular chow feeding group and the high-fat diet feeding group. After two weeks of feeding, the mice were randomly divided into four groups: the Control group, CUMS group, T2DM group, and T2DM+CUMS group. The T2DM group and T2DM+CUMS group received an intraperitoneal injection of STZ to induce the T2DM model. The CUMS and T2DM+CUMS groups were exposed to CUMS to induce depressive-like phenotypes. Blood and tissue samples were obtained for pertinent analysis and detection. Results: Compared with the T2DM mice, T2DM+CUMS mice had higher blood glucose and lipid levels, insulin resistance, inflammation of the liver and pancreas, impaired liver function, and increased total bile acids. These changes were accompanied by attenuated FXR signaling. Chronic stress was found to attenuate FXR expression and its downstream target, FGF15, in the ileum when compared with the T2DM group. Conclusion: FXR may play a role in the diabetic disorder of glucolipid metabolism when aggravated by chronic stress. FXR and its downstream target, FGF15, may be therapeutic targets for treating comorbid T2DM and depression.

Corrigendum: Could peripheral 5-HT level be used as a biomarker for depression diagnosis and treatment? A narrative minireview.

[This corrects the article DOI: 10.3389/fphar.2023.1149511.].

Multi-cohort validation study of a four-gene signature for risk stratification and treatment response prediction in hepatocellular carcinoma.

BACKGROUND: The intricate molecular landscape of hepatocellular carcinoma (HCC) presents a significant challenge to achieving precise risk stratification through clinical genetic testing. At present, there is a paucity of robust gene signatures that could assist clinicians in making clinical decisions for patients with HCC. METHODS: We obtained gene expression profiles of patients with HCC from 20 independent cohorts available in public databases. A gene signature was developed by employing two machine learning algorithms. In addition to validating the signature with high-throughput data in public cohorts, we external validated the signature in 64 HCC cases by RT-PCR method. We compared genomic, transcriptomic and proteomic features between different subgroups. We also compared our signature to 130 gene signatures that have already been published. RESULTS: We developed a novel four-gene signature, designated as HCC4, that demonstrates significant potential for the prediction of survival outcomes in more than 1300 patients with HCC. The HCC4 also has potential for predicting recurrence and tumor volume doubling time, assessing transcatheter arterial chemoembolization and immunotherapy responses, and non-invasive detection of HCC. The high HCC4 score group shows a higher frequency of mutations in genes TP53, RB1 and TSC1/2, as well as increased activity of cell-cycle, glycolysis and hypoxia signaling pathways, higher cancer stemness score, and lower lipid metabolism activity. In seven HCC cohorts, HCC4 exhibited a higher average C-index in predicting overall survival compared to the 130 signatures previously published. Drug screening indicated that patients with high HCC4 scores were more sensitive to agents targeting AURKA, TUBB, JMJD6 and KIFC1. CONCLUSIONS: Our findings demonstrated that HCC4 is a powerful tool for improving risk stratification and for identifying HCC patients who are most likely to benefit from TACE treatment, immunotherapy, and other experimental therapies.

Corrigendum: Lactobacillus reuteri strain 8008 attenuated the aggravation of depressive-like behavior induced by CUMS in high-fat diet-fed mice through regulating the gut microbiota.

[This corrects the article DOI: 10.3389/fphar.2023.1149185.].

CT radiomics for prediction of microvascular invasion in hepatocellular carcinoma: A systematic review and meta-analysis.

The power of computed tomography (CT) radiomics for preoperative prediction of microvascular invasion (MVI) in hepatocellular carcinoma (HCC) demonstrated in current research is variable. This systematic review and meta-analysis aim to evaluate the value of CT radiomics for MVI prediction in HCC, and to investigate the methodologic quality in the workflow of radiomics research. Databases of PubMed, Embase, Web of Science, and Cochrane Library were systematically searched. The methodologic quality of included studies was assessed. Validation data from studies with Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement type 2a or above were extracted for meta-analysis. Eleven studies were included, among which nine were eligible for meta-analysis. Radiomics quality scores of the enrolled eleven studies varied from 6 to 17, accounting for 16.7%-47.2% of the total points, with an average score of 14. Pooled sensitivity, specificity, and Area Under the summary receiver operator Characteristic Curve (AUC) were 0.82 (95% CI 0.77-0.86), 0.79 (95% CI 0.75-0.83), and 0.87 (95% CI 0.84-0.91) for the predictive performance of CT radiomics, respectively. Meta-regression and subgroup analyses showed radiomics model based on 3D tumor segmentation, and deep learning model achieved superior performances compared to 2D segmentation and non-deep learning model, respectively (AUC: 0.93 vs. 0.83, and 0.97 vs. 0.83, respectively). This study proves that CT radiomics could predict MVI in HCC. The heterogeneity of the included studies precludes a definition of the role of CT radiomics in predicting MVI, but methodology warrants uniformization in the radiology community regarding radiomics in HCC.

Predicting skip metastasis in lateral lymph nodes of papillary thyroid carcinoma based on clinical and ultrasound features.

Background: Skip metastasis in papillary thyroid cancer (PTC), defined as lateral lymph node metastasis (LLNM) without the involvement of central lymph node metastasis (CLNM), is generally unpredictable. Our study aimed to develop a model to predict skip metastasis by using clinicopathological and ultrasound factors of PTC. Methods: We retrospectively reviewed the medical records of patients who underwent total thyroidectomy and central lymph node dissection (CLND) plus lateral lymph node dissection (LLND) between January 2019 and December 2021 at the First Affiliated Hospital of Soochow University. Furthermore, univariate and multivariate analyses assessed the clinical and ultrasound risk factors. Receiver operating characteristic (ROC) curves were used to find the optimal cut-off values for age and dominant nodule diameter. Multivariate logistic regression analysis results were used to construct a nomogram and were validated internally. Results: In all patients, the skip metastasis rate was 15.4% (41/267). Skip metastasis was more frequently found in patients with a tumour size ≤10 mm (OR 0.439; P = 0.033), upper tumour location (OR 3.050; P=0.006) and fewer CLNDs (OR 0.870; P = 0.005). After analysing the clinical and ultrasound characteristics of the tumour, five factors were ultimately associated with lateral lymph node skip metastasis and were used to construct the model. These factors were an age >40 years, tumour diameter <9.1 mm, upper tumour location, non-smooth margin and extrathyroidal extension. The internally evaluated calibration curves indicated an excellent correlation between the projected and actual skip metastasis probability. The nomogram performed well in discrimination, with a concordance index of 0.797 (95% CI, 0.726 to 0.867). Conclusions: This study screened for predictors of skip metastasis in PTC and established a nomogram that effectively predicted the risk of potential skip metastasis in patients preoperatively. The method can predict and distinguish skip metastases in PTC in a simple and inexpensive manner, and it may have future therapeutic utility.

Lactobacillus reuteri strain 8008 attenuated the aggravation of depressive-like behavior induced by CUMS in high-fat diet-fed mice through regulating the gut microbiota.

Objective: Gut microbiota play a key role in the pathogenesis of obesity and depression. Probiotics are a preventive strategy for obesity and a novel treatment for depression symptoms. However, the ameliorative or therapeutic effect of potential probiotic candidate Lactobacillus reuteri (L. reuteri) on obesity and depression comorbidity still remains unclear. We investigated the effects of chronic unpredictable mild stress (CUMS) in high-fat diet-fed mice and the effects of Lactobacillus reuteri strain 8008 on various disease indicators of obesity and depression comorbidity disease. Methods: Forty male C57BL/6 mice were randomized into 2 groups: the normal control (NC) group (n = 10) and the high-fat diet (HFD) group (n = 30), being fed with normal diet (ND) or high-fat diet (HFD) for 8 weeks, respectively. Then the obese mice fed with HFD were randomly allocated into 3 sub-groups: the HFD group (n = 10); the HFD + CUMS group (n = 10); the HFD + CUMS + L.r group (n = 10). The latter 2 subgroups underwent CUMS for 4 weeks to build the obesity and depression comorbidity mice model. During the duration of treatment, mice were gavaged with 0.5 mL PBS solution or L. reuteri (2 × 109 CFU/mL) once a day, respectively. The body weight, food intake, organ weight, behavioral indicators, histology, blood lipids, levels of inflammatory cytokines and tight junction proteins and abundance of colonic contents bacteria were measured. Results: The obesity and depression comorbidity mice model was successfully established after HFD feeding and chronic stress. The comorbid mice demonstrated inflammatory responses increase in liver and adipose tissues, worsened damage to the intestinal barrier as well as gut microbiota disorder. Gavaged with L. reuteri attenuated depressive-like behavior, improved blood lipids and insulin resistance, reduced inflammation in liver and adipose tissues, improved intestinal tight junctions as well as the microbiome dysbiosis in obesity and depression comorbidity mice. Conclusion: Lactobacillus reuteri strain 8008 could alleviate depressive-like behaviors and related indicators of obesity disorders by regulating the gut microbiota in obesity and depression comorbid mice.

Could peripheral 5-HT level be used as a biomarker for depression diagnosis and treatment? A narrative minireview.

The serotonin hypothesis of depression is still influential, but the relationship between peripheral 5-HT levels and depression is still unknown. This review aimed to verify whether peripheral 5-HT levels could be used as a biomarker for depression diagnosis and treatment. PubMed and EMBASE were searched using terms appropriate to the area of research. Articles from 1957 to 2022 in the following terms were identified: depression, 5-HT, serotonin and peripheral (serum, plasma, blood platelets). 33 studies were included: seven clinical trials about periphery 5-HT levels in depressive patients compared to normal subjects, 15 clinical trials about changes of peripheral 5-HT levels in patients with depression after drug treatment and 11 animal experiments about peripheral 5-HT levels in animal models of depression. Peripheral 5-HT levels presented three different outcomes before and after antidepressant treatments: increased, decreased and no significant change. In conclusion, changes in peripheral 5-HT levels did not show consistent results among these studies. Peripheral 5-HT level could not be used as a biomarker both for depression diagnosis and for antidepressant efficacy evaluation.

CT radiomics for prediction of microvascular invasion in hepatocellular carcinoma: A systematic review and meta-analysis

Abstract The power of computed tomography (CT) radiomics for preoperative prediction of microvascular invasion (MVI) in hepatocellular carcinoma (HCC) demonstrated in current research is variable. This systematic review and meta-analysis aim to evaluate the value of CT radiomics for MVI prediction in HCC, and to investigate the methodologic quality in the workflow of radiomics research. Databases of PubMed, Embase, Web of Science, and Cochrane Library were systematically searched. The methodologic quality of included studies was assessed. Validation data from studies with Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) statement type 2a or above were extracted for meta-analysis. Eleven studies were included, among which nine were eligible for meta-analysis. Radiomics quality scores of the enrolled eleven studies varied from 6 to 17, accounting for 16.7%-47.2% of the total points, with an average score of 14. Pooled sensitivity, specificity, and Area Under the summary receiver operator Characteristic Curve (AUC) were 0.82 (95% CI 0.77-0.86), 0.79 (95% CI 0.75-0.83), and 0.87 (95% CI 0.84-0.91) for the predictive performance of CT radiomics, respectively. Meta-regression and subgroup analyses showed radiomics model based on 3D tumor segmentation, and deep learning model achieved superior performances compared to 2D segmentation and non-deep learning model, respectively (AUC: 0.93 vs. 0.83, and 0.97 vs. 0.83, respectively). This study proves that CT radiomics could predict MVI in HCC. The heterogeneity of the included studies precludes a definition of the role of CT radiomics in predicting MVI, but methodology warrants uniformization in the radiology community regarding radiomics in HCC.

Identification of Genetic Variations and Candidate Genes Responsible for Stalk Sugar Content and Agronomic Traits in Fresh Corn via GWAS across Multiple Environments.

The stem and leaves of fresh corn plants can be used as green silage or can be converted to biofuels, and the stalk sugar content and yield directly determine the application value of fresh corn. To identify the genetic variations and candidate genes responsible for the related traits in fresh corn, the genome-wide scan and genome-wide association analysis (GWAS) were performed. A total of 32 selective regions containing 172 genes were detected between sweet and waxy corns. Using the stalk sugar content and seven other agronomic traits measured in four seasons over two years, the GWAS identified ninety-two significant single nucleotide polymorphisms (SNPs). Most importantly, seven SNPs associated with the stalk sugar content were detected across multiple environments, which could explain 13.68-17.82% of the phenotypic variation. Accessions differing in genotype for certain significant SNPs showed significant variation in the stalk sugar content and other agronomic traits, and the expression levels of six important candidate genes were significantly different between two materials with different stalk sugar content. The genetic variations and candidate genes provide valuable resources for future studies of the molecular mechanism of the stalk sugar content and establish the foundation for molecular marker-assisted breeding of fresh corn.

Exploring the Antibacterial and Biosensing Applications of Peroxidase-Mimetic Ni0.1Cu0.9S Nanoflower.

Nanozymes, as artificial enzymes with the biological action of natural enzymes, have enormous potential in the fields of disease diagnosis, bacteriostasis, biosensing, etc. In this work, the Ni0.1Cu0.9S nanoflower was successfully synthesized through a one-step hydrothermal method. A combined strategy of Ni doping and morphology design was employed to adjust its electronic structure and active sites, endowing the Ni0.1Cu0.9S nanoflower with excellent peroxidase-like activity. Therefore, it can catalyze the decomposition of H2O2 to generate •OH with higher antibacterial activity, establishing a broad-spectrum antibacterial system based on the Ni0.1Cu0.9S nanoflower against E. coli and S. aureus, which avoids the harm of a high concentration of H2O2. Additionally, the colorless substrate TMB can be catalytically oxidized into blue ox-TMB via •OH. As a result, a colorimetric technique with rapid and accurate detection of ascorbic acid (AA) by the unaided eye was designed, in view of the specific inhibition effect towards the oxidation of TMB. This detection platform has a wide linear range (10~800 µM) with a low limit of detection (0.84 µM) and exhibits a satisfactory selectivity toward the detection of AA. This study sheds new light on the application of copper-containing nanozymes in the fields of biomedicine and bioassay.

DO-Conv: Depthwise Over-Parameterized Convolutional Layer.

Convolutional layers are the core building blocks of Convolutional Neural Networks (CNNs). In this paper, we propose to augment a convolutional layer with an additional depthwise convolution, where each input channel is convolved with a different 2D kernel. The composition of the two convolutions constitutes an over-parameterization, since it adds learnable parameters, while the resulting linear operation can be expressed by a single convolution layer. We refer to this depthwise over-parameterized convolutional layer as DO-Conv, which is a novel way of over-parameterization. We show with extensive experiments that the mere replacement of conventional convolutional layers with DO-Conv layers boosts the performance of CNNs on many classical vision tasks, such as image classification, detection, and segmentation. Moreover, in the inference phase, the depthwise convolution is folded into the conventional convolution, reducing the computation to be exactly equivalent to that of a convolutional layer without over-parameterization. As DO-Conv introduces performance gains without incurring any computational complexity increase for inference, we advocate it as an alternative to the conventional convolutional layer. We open sourced an implementation of DO-Conv in Tensorflow, PyTorch and GluonCV at https://github.com/yangyanli/DO-Conv.