DEKR-SPrior: An Efficient Bottom-Up Keypoint Detection Model for Accurate Pod Phenotyping in Soybean.

The pod and seed counts are important yield-related traits in soybean. High-precision soybean breeders face the major challenge of accurately phenotyping the number of pods and seeds in a high-throughput manner. Recent advances in artificial intelligence, especially deep learning (DL) models, have provided new avenues for high-throughput phenotyping of crop traits with increased precision. However, the available DL models are less effective for phenotyping pods that are densely packed and overlap in in situ soybean plants; thus, accurate phenotyping of the number of pods and seeds in soybean plant is an important challenge. To address this challenge, the present study proposed a bottom-up model, DEKR-SPrior (disentangled keypoint regression with structural prior), for in situ soybean pod phenotyping, which considers soybean pods and seeds analogous to human people and joints, respectively. In particular, we designed a novel structural prior (SPrior) module that utilizes cosine similarity to improve feature discrimination, which is important for differentiating closely located seeds from highly similar seeds. To further enhance the accuracy of pod location, we cropped full-sized images into smaller and high-resolution subimages for analysis. The results on our image datasets revealed that DEKR-SPrior outperformed multiple bottom-up models, viz., Lightweight-OpenPose, OpenPose, HigherHRNet, and DEKR, reducing the mean absolute error from 25.81 (in the original DEKR) to 21.11 (in the DEKR-SPrior) in pod phenotyping. This paper demonstrated the great potential of DEKR-SPrior for plant phenotyping, and we hope that DEKR-SPrior will help future plant phenotyping.

DG2GAN: improving defect recognition performance with generated defect image sample.

This article aims to improve the deep-learning-based surface defect recognition. In actual manufacturing processes, there are issues such as data imbalance, insufficient diversity, and poor quality of augmented data in the collected image data for product defect recognition. A novel defect generation method with multiple loss functions, DG2GAN is presented in this paper. This method employs cycle consistency loss to generate defect images from a large number of defect-free images, overcoming the issue of imbalanced original training data. DJS optimized discriminator loss is introduced in the added discriminator to encourage the generation of diverse defect images. Furthermore, to maintain diversity in generated images while improving image quality, a new DG2 adversarial loss is proposed with the aim of generating high-quality and diverse images. The experiments demonstrated that DG2GAN produces defect images of higher quality and greater diversity compared with other advanced generation methods. Using the DG2GAN method to augment defect data in the CrackForest and MVTec datasets, the defect recognition accuracy increased from 86.9 to 94.6%, and the precision improved from 59.8 to 80.2%. The experimental results show that using the proposed defect generation method can obtain sample images with high quality and diversity and employ this method for data augmentation significantly enhances surface defect recognition technology.

HIF-1 induced tiRNA-Lys-CTT-003 is protective against cisplatin induced ferroptosis of renal tubular cells in mouse AKI model.

HIF-1 activation is protective in acute kidney injury (AKI), but its underlying mechanism is not fully understood. Stress-induced tRNA derived small RNAs play an emerging role in cellular processes. This study investigated the role of HIF-1 associated tiRNA-Lys-CTT-003 (tiR-Lys) in an AKI mouse model. Our sequencing results showed that ischemia can promote the production of renal tiR-Lys by activating HIF-1α. FG-4592, a HIF-1 inducer, can also upregulate the expression of tiR-Lys in renal tubular cells. Both overexpression of tiR-Lys and FG-4592 pre-treatment could improve mitochondrial damage and lipid peroxidation with alleviated renal function and morphological damage in cisplatin-induced AKI mice. While the anti-ferroptosis effect of FG-4592 were largely eliminated by tiR-Lys inhibitor. Notably, tiR-Lys directly alleviated cell death and MDA accumulation induced by the ferroptosis inducer Erastin, accompanied with restored expression of GPX4. RNA-Pulldown and RIP-qPCR results revealed that tiR-Lys can interact with the RNA-binding protein GRSF1.tiR-lys overexpression can preserve protein expression of GRSF1 decreased by cisplatin. Inhibiting Grsf1 via shRNA eliminated the upregulation of GPX4 by tiR-Lys. In conclusion, our study demonstrates that HIF-1α-induced tiR-Lys is protective in cisplatin-induced AKI, primarily by upregulating the level of GPX4 through interaction with GRSF1, thereby inhibiting ferroptosis in renal tubular epithelial cells.

Identification of key genes as potential diagnostic biomarkers in sepsis by bioinformatics analysis.

Background: Sepsis, an infection-triggered inflammatory syndrome, poses a global clinical challenge with limited therapeutic options. Our study is designed to identify potential diagnostic biomarkers of sepsis onset in critically ill patients by bioinformatics analysis. Methods: Gene expression profiles of GSE28750 and GSE74224 were obtained from the Gene Expression Omnibus (GEO) database. These datasets were merged, normalized and de-batched. Weighted gene co-expression network analysis (WGCNA) was performed and the gene modules most associated with sepsis were identified as key modules. Functional enrichment analysis of the key module genes was then conducted. Moreover, differentially expressed gene (DEG) analysis was conducted by the "limma" R package. Protein-protein interaction (PPI) network was created using STRING and Cytoscape, and PPI hub genes were identified with the cytoHubba plugin. The PPI hub genes overlapping with the genes in key modules of WGCNA were determined to be the sepsis-related key genes. Subsequently, the key overlapping genes were validated in an external independent dataset and sepsis patients recruited in our hospital. In addition, CIBERSORT analysis evaluated immune cell infiltration and its correlation with key genes. Results: By WGCNA, the greenyellow module showed the highest positive correlation with sepsis (0.7, p = 2e - 19). 293 DEGs were identified in the merged datasets. The PPI network was created, and the CytoHubba was used to calculate the top 20 genes based on four algorithms (Degree, EPC, MCC, and MNC). Ultimately, LTF, LCN2, ELANE, MPO and CEACAM8 were identified as key overlapping genes as they appeared in the PPI hub genes and the key module genes of WGCNA. These sepsis-related key genes were validated in an independent external dataset (GSE131761) and sepsis patients recruited in our hospital. Additionally, the immune infiltration profiles differed significantly between sepsis and non-sepsis critical illness groups. Correlations between immune cells and these five key genes were assessed, revealing that plasma cells, macrophages M0, monocytes, T cells regulatory, eosinophils and NK cells resting were simultaneously and significantly associated with more than two key genes. Conclusion: This study suggests a critical role of LTF, LCN2, ELANE, MPO and CEACAM8 in sepsis and may provide potential diagnostic biomarkers and therapeutic targets for the treatment of sepsis.

Urea transporter UT-A1 as a novel drug target for hyponatremia.

Hyponatremia is the most common disorder of electrolyte imbalances. It is necessary to develop new type of diuretics to treat hyponatremia without losing electrolytes. Urea transporters (UT) play an important role in the urine concentrating process and have been proved as a novel diuretic target. In this study, rat and mouse syndromes of inappropriate antidiuretic hormone secretion (SIADH) models were constructed and analyzed to determine if UTs are a promising drug target for treating hyponatremia. Experimental results showed that 100 mg/kg UT inhibitor 25a significantly increased serum osmolality (from 249.83 ± 5.95 to 294.33 ± 3.90 mOsm/kg) and serum sodium (from 114 ± 2.07 to 136.67 ± 3.82 mmol/L) respectively in hyponatremia rats by diuresis. Serum chemical examination showed that 25a neither caused another electrolyte imbalance nor influenced the lipid metabolism. Using UT-A1 and UT-B knockout mouse SIADH model, it was found that serum osmolality and serum sodium were lowered much less in UT-A1 knockout mice than in UT-B knockout mice, which suggest UT-A1 is a better therapeutic target than UT-B to treat hyponatremia. This study provides a proof of concept that UT-A1 is a diuretic target for SIADH-induced hyponatremia and UT-A1 inhibitors might be developed into new diuretics to treat hyponatremia.

Cost-effectiveness of a multicomponent-adherence intervention in fracture liaison services.

BACKGROUND: This study aims to assess the lifetime cost-effectiveness of a multi-component adherence intervention (MCAI), including a patient decision aid and motivational interviewing, compared to usual care in patients with a recent fracture attending fracture liaison services (FLS) and eligible for anti-osteoporosis medication (AOM). RESEARCH DESIGN AND METHODS: Data on AOM initiation and one-year persistence were collected from a quasi-experimental study conducted between 2019 and 2023 in two Dutch FLS centers. An individual level, state-transition Markov model was used to simulate lifetime costs and quality-adjusted life years (QALYs) with a societal perspective of MCAI vs usual care. One-way and probabilistic sensitivity analyses were conducted including variation in additional FLS and MCAI costs (no MCAI cost in baseline). RESULTS: MCAI was associated with gain in QALYs (0.0012) and reduction in costs (-€16) and is therefore dominant. At the Dutch willingness-to-pay threshold of €50,000/QALY, MCAI remained cost-effective when increasing costs of the FLS visit or the yearly maintenance cost for MCAI up to +€60. Probabilistic sensitivity analysis demonstrated MCAI to be dominant in 54% of the simulations and cost-effective in 87% with a threshold of €50,000/QALY. CONCLUSIONS: A MCAI implemented in FLS centers may lead to cost-effective allocation of resources in FLS care, depending on extra costs.

SGLT2i Alleviates Atherosclerosis by Inhibiting NHE1 Activation to Protect against Macrophage Senescence Induced by Angiotensin II.

BACKGROUND: Sodium-dependent glucose transporter (SGLT2) inhibitors (SGLT2i) have been found to have anti-atherosclerotic effects in clinical treatment. OBJECTIVES: The aim of this study was to explore whether angiotensin II (Ang II) induces changes in the expression of Na+/H+ exchanger of cytoplasmic membrane channel proteins (NHE1) and SGLT2 in macrophages and whether dapagliflozin (DAPA), an SGLT2i, protects against Ang II induced macrophage senescence by inhibiting NHE1 activation to alleviate Atherosclerosis (AS). METHODS: After intervention with DAPA plus gavage or feeding them a high-fat diet, the mice's aortas were dissected, and oil red O staining was performed. Cell proliferation and toxicity detection, western blot, immunofluorescence, and ß-galactosidase staining methods were adopted to detect cell activity, expressions of senescence-related genes, and number of senescent cells after different concentrations of Ang II or DAPA or plasmid NHE1 were treated with RAW264.7 cells. RESULTS: (1) The formation of AS plaques in ApoE -/- mice showed a downward trend under DAPA. (2) After the intervention of Ang II, the cell activity of RAW264.7 decreased, and the expression of senescent cells and related genes increased. (3) Under the Ang II condition, the expression of SGLT2 and NHE1 increased, and SGLT2, NHE1, and senescence-related genes decreased with the addition of DAPA. (4) The expression of NHE1, senescent cells and related genes decreased in RAW264.7 cells after DAPA treatment with plasmid NHE1 intervention. CONCLUSION: SGLT2i alleviates atherosclerosis by inhibiting NHE1 activation to protect against macrophage senescence induced by Ang II.

Expanding the clinical phenotype and genetic spectrum of GEMIN5 disorders: Early-infantile developmental and epileptic encephalopathies.

BACKGROUND: Several biallelic truncating and missense variants of the gem nuclear organelle-associated protein 5 (GEMIN5) gene have been reported to cause neurodevelopmental disorders characterized by cerebellar atrophy, intellectual disability, and motor dysfunction. However, the association between biallelic GEMIN5 variants and early-infantile developmental and epileptic encephalopathies (EIDEEs) has not been reported. PURPOSE: This study aimed to expand the phenotypic spectrum of GEMIN5 and explore the correlations between epilepsy and molecular sub-regional locations. METHODS: We performed whole-exome sequencing in two patients with EIDEE with unexplained etiologies. The damaging effects of variants were predicted using multiple in silico tools and modeling. All reported patients with GEMIN5 pathogenic variants and detailed neurological phenotypes were analyzed to evaluate the genotype-phenotype relationship. RESULTS: Novel biallelic GEMIN5 variants were identified in two unrelated female patients with EIDEE, including a frameshift variant (Hg19, chr5:154284147-154284148delCT: NM_015465: c.2551_c.2552delCT: p.(Leu851fs*30)), a nonsense mutation (Hg19, chr5:154299603-154299603delTinsAGA: NM_015465: c.1523delTinsAGA: p.(Leu508*)), and two missense variants (Hg19, chr5:154282663T > A: NM_015465: c.2705T > A: p.(Leu902Gln) and Hg19, chr5:154281002C > G: NM_015465: c.2911C > G: p.(Gln971Glu)), which were inherited from asymptomatic parents and predicted to be damaging or probably damaging using in silico tools. Except p.Leu508*, all these mutations are located in tetratricopeptide repeat (TPR) domain. Our two female patients presented with seizures less than 1 month after birth, followed by clusters of spasms. Brain magnetic resonance imaging suggests dysgenesis of the corpus callosum and cerebellar hypoplasia. Video electroencephalogram showed suppression-bursts. Through a literature review, we found 5 published papers reporting 48 patients with biallelic variants in GEMIN5. Eight of 48 patients have epilepsy, and 5 patients started before 1 year old, which reminds us of the relevance between GEMIN5 variants and EIDEE. Further analysis of the 49 GEMIN5 variants in those 50 patients demonstrated that variants in TPR-like domain or RBS domain were more likely to be associated with epilepsy. CONCLUSIONS: We found novel biallelic variants of GEMIN5 in two individuals with EIDEE and expanded the clinical phenotypes of GEMIN5 variants. It is suggested that the GEMIN5 gene should be added to the EIDEE gene panel to aid in the clinical diagnosis of EIDEE and to help determine patient prognosis.

Taking SCFAs produced by Lactobacillus reuteri orally reshapes gut microbiota and elicits antitumor responses.

BACKGROUND: Colorectal cancer (CRC) incidence is increasing in recent years due to intestinal flora imbalance, making oral probiotics a hotspot for research. However, numerous studies related to intestinal flora regulation ignore its internal mechanisms without in-depth research. RESULTS: Here, we developed a probiotic microgel delivery system (L.r@(SA-CS)2) through the layer-by-layer encapsulation technology of alginate (SA) and chitosan (CS) to improve gut microbiota dysbiosis and enhance anti-tumor therapeutic effect. Short chain fatty acids (SCFAs) produced by L.r have direct anti-tumor effects. Additionally, it reduces harmful bacteria such as Proteobacteria and Fusobacteriota, and through bacteria mutualophy increases beneficial bacteria such as Bacteroidota and Firmicutes which produce butyric acid. By binding to the G protein-coupled receptor 109A (GPR109A) on the surface of colonic epithelial cells, butyric acid can induce apoptosis in abnormal cells. Due to the low expression of GPR109A in colon cancer cells, MK-6892 (MK) can be used to stimulate GPR109A. With increased production of butyrate, activated GPR109A is able to bind more butyrate, which further promotes apoptosis of cancer cells and triggers an antitumor response. CONCLUSION: It appears that the oral administration of L.r@(SA-CS)2 microgels may provide a treatment option for CRC by modifying the gut microbiota.

SoxB family genes delay regeneration and cause abnormal movement in Dugesia japonica.

SoxB subfamily is an important branch of Sox family and plays a key role in animal physiological process, but little is known about their function in planarian regeneration. This study aims to evaluate the function of DjSoxB family genes in intact and regenerating planarians Dugesia japonica. Here, we amplify the full-length cDNA of DjSoxB1 and DjSoxB2 in D. japonica by rapid amplification of the cDNA ends (RACE), detect the expression of DjSoxB family genes in planarian. The results show that DjSoxBs are expressed in parenchymal tissue and the hybridization signals partially disappear after irradiation indicates DjSoxB family genes are expressed in neoblasts. After the RNA interference (RNAi) of DjSoxB1, DjSoxB2 and DjSoxB3 separately, the numbers of proliferative cells are all reduced that causes planarians show slower growth of blastema in the early stage of regeneration, and nerves of planarians are affected that the movement speed of planarians decreases in varying degrees. Specially, planarians in the DjSoxB3 RNAi group show shrinkage and twisting. Overall, this study reveals that DjSoxB family genes play a role in cell proliferation during regeneration. They also play an important role in the maintenance of normal nerve function and nerve regeneration. These results provide directions for the functional study of SoxB family genes and provide an important foundation for planarian regeneration.

Causal association between 25-Hydroxyvitamin D and risk of abnormal spermatozoa: A bidirectional Mendelian randomization study.

Previous studies indicated conflicting findings regarding the association between vitamin D and abnormal spermatozoa. Herein, we assessed the causal association between circulating 25-Hydroxyvitamin D (25OHD) levels and the risk of abnormal spermatozoa by utilizing bidirectional Mendelian randomization (MR) analysis. Genome-wide association study summary statistics for 25OHD and abnormal spermatozoa were obtained from publicly accessible databases. Single nucleotide polymorphisms (SNPs) associated with 25OHD and SNPs associated with abnormal spermatozoa were used as instrumental variables (IVs) for forward MR analysis and reverse MR analysis, respectively. Inverse variance weighted (IVW) was the main MR approach, while weighted median, MR-Egger, and maximum likelihood methods were employed to supplement IVW. In addition, several sensitivity tests assessed the reliability of MR analysis. Forward MR analysis showed that elevated 25OHD levels significantly reduced abnormal spermatozoa risk (odds ratio [OR] = 0.75, 95 % confidence interval [CI]: 0.56-1.00, P = 4.98E-02), and the effect remained statistically significant after excluding SNPs associated with confounders (OR = 0.73, 95 % CI: 0.54-0.98, P = 3.83E-02) or only utilizing SNPs located near 25OHD-associated genes only as IVs (OR = 0.58, 95 % CI: 0.41-0.81, P = 1.67E-03). Reverse MR analysis indicated abnormal spermatozoa not affecting 25OHD level (P > 0.05). Sensitivity tests showed that MR analyses were not affected by heterogeneity and horizontal polytropy. Overall, the present MR study supports that elevated 25OHD levels reduce the risk of abnormal spermatozoa. Therefore, ensuring adequate vitamin D intake and maintaining stable levels of 25OHD may be effective strategies to optimize reproductive outcomes.

Mendelian randomization study supports positive bidirectional causal relationships between genetically predicted insomnia symptom and liability to benign prostatic hyperplasia.

BACKGROUND: Sleep quality may be related to benign prostatic hyperplasia (BPH), however causal associations have not been established. This study aimed to evaluate causal relationships between six sleep traits ([i] day time napping, [ii] daytime sleepiness, [iii] insomnia, [iv] long sleep duration, [v] short sleep duration, and [vi] sleep duration per hour) and BPH through a bidirectional Mendelian randomization (MR) study. METHODS: Genome-wide association summary statistics of sleep traits and BPH were downloaded from public databases. Inverse variance weighting (IVW) was used as the main approach for causal inference. For causal estimates identified by IVW, various sensitivity analyses were performed to assess the reliability of the results: (i) four additional MR methods to complement IVW; (ii) Cochran's Q test to assess heterogeneity; (iii) MR-Egger intercept test and MR-PRESSO global test to assess horizontal pleiotropy; and (iv) leave-one-out method to assess stability. RESULTS: Forward MR analyses indicated that genetically predicted insomnia symptom significantly increased BPH risk (OR = 1.267, 95% CI: 1.003-1.601, P = 0.048), while reverse MR analyses identified that genetically predicted liability to BPH significantly increased the incidence of insomnia (OR = 1.026, 95% CI: 1.000-1.052, P = 0.048). In a replicate MR analysis based on summary statistics including exclusively male participants, the finding of increased risk of BPH due to genetically predicted insomnia symptom was further validated (OR = 1.488, 95% CI: 1.096-2.022, P = 0.011). No further causal links were identified. In addition, sensitivity tests demonstrated the reliability of the MR results. CONCLUSION: This study identified that a higher prevalence of genetically predicted insomnia symptoms may significantly increase the risk of BPH, while genetically predicted liability to BPH may in turn increase the incidence of insomnia symptom. Therefore, improving sleep quality and reducing the risk of insomnia could be a crucial approach for the prevention of BPH.

Numerical simulation of rock blasting under different in-situ stresses and joint conditions.

High primary rock stress can limit the generation of rock cracks caused by blasting, and blasting usually shows different rock breaking states under different primary rock stress conditions. There are a large number of naturally formed joints in rock mass, due to the limitations of laboratory tests, a numerical model of jointed rock mass was established using LS-DYNA software to investigate the evolution of blasting damage under various in-situ stresses and open joints. In this simulation, using the Lagrange-Euler (ALE) procedure and the equation of state (JWL) that defines explosive materials, the study considered different joint thicknesses (2cm, 4cm, and 6cm), joint angles (0°, 30°, 60°, and 90°), and in-situ stress conditions (lateral stress coefficients of 0.5, 1, and 2, with vertical in-situ stresses of 10MPa and 20MPa), through stress analysis and damage area comparison, the relationship between damage crack propagation and horizontal and vertical stress difference is explored. The research aimed to understand the mechanisms underlying crack initiation and propagation. The results show that: (1) The presence of joints exerts a barrier effect on the expansion and penetration of cracks. When explosion stress waves reach the joint surface, their propagation is impeded, leading to the diffusion of wing cracks at the joint ends. When the lateral stress coefficient and joint angle are the same, an increase in initial in-situ stress results in a reduction in the area of the blasting damage zone. (2) Under the same initial in-situ stress conditions, the area of the blasting damage zone initially increases and then decreases with an increasing joint angle. However, it remains larger than that without a joint, and there exists an optimal angle that maximizes the damage area. In the simulated conditions, the area of damage cracks is greatest when the joint angle is 60° dip angle. (3) The presence of initial in-situ stress has a certain impact on the initiation and expansion of blasting cracks. The degree and nature of this influence are not solely related to the lateral stress coefficient but also depend on the joint's angle and thickness. When in-situ stress is present, the initial in-situ stress field's pressure is not conducive to the initiation and propagation of blasting cracks. However, the existence of a joint has a noticeable guiding and promoting effect on crack propagation, and the pattern of crack propagation is influenced by both joint and in-situ stress conditions.

Fecal microbiota transplantation for treatment of non-alcoholic fatty liver disease: Mechanism, clinical evidence, and prospect.

The population of non-alcoholic fatty liver disease (NAFLD) patients along with relevant advanced liver disease is projected to continue growing, because currently no medications are approved for treatment. Fecal microbiota transplantation (FMT) is believed a novel and promising therapeutic approach based on the concept of the gut-liver axis in liver disease. There has been an increase in the number of pre-clinical and clinical studies evaluating FMT in NAFLD treatment, however, existing findings diverge on its effects. Herein, we briefly summarized the mechanism of FMT for NAFLD treatment, reviewed randomized controlled trials for evaluating its efficacy in NAFLD, and proposed the prospect of future trials on FMT.

Flexible microfluidic colorimetric detection chip integrated with ABTS·+ and Co@MnO2 nanozyme catalyzed TMB reaction systems for bio-enzyme free detection of sweat uric acid.

BACKGROUND: The development of wearable detection devices that can achieve noninvasive, on-site and real-time monitoring of sweat metabolites is of great demand and practical significance for point-of-care testing and healthcare monitoring. Monitoring uric acid (UA) content in sweat provides a simple and promising way to reduce the risk of gout and hyperuricemia. Traditional bioenzyme based UA assays suffer from high cost, poor stability, inconvenience for storage and easy deactivation of bioenzymes. Wearable microfluidic colorimetric detection device for sweat UA detection has not been reported. The development of novel wearable microfluidic colorimetric detection chip with no requirement of bioenzymes for sweat UA detection is of great importance for health care monitoring. RESULTS: Firstly, Co@MnO2 nanozyme with high oxidase-like activity was synthesized and characterized. Co@MnO2 can catalyze the oxidation of 3,3',5,5'-tetramethylbenzidine (TMB) directly to generate blue-green colored ox-TMB. Green colored 2,2'-Azinobis-(3-ethylbenzthiazoline-6-sulphonate) radical (ABTS·+) was produced by the oxidation of ABTS by potassium persulfate. UA exhibits distinct quenching effect on Co@MnO2 catalyzed TMB colorimetric reaction system and ABTS·+ based colorimetric system, leading to obvious color fading of the two colorimetric systems. Then, a flexible microfluidic colorimetric detection chip for UA detection was fabricated by assembling Co@MnO2/TMB modified paper chips and ABTS·+ modified paper chips into a polydimethylsiloxane (PDMS) microfluidic chip. The fabricated microfluidic colorimetric detection chip exhibits good linear relationship for sweat UA detection. The linear range is from 20 to 200 µmol/L with detection limit as low as 6.6 µmol/L. Good results were obtained for the detection of UA in actual sweat from three volunteers. SIGNIFICANCE: This work provides two bio-enzyme free colorimetric detection systems for UA detection. Furthermore, a simple, low-cost and selective flexible wearable microfluidic colorimetric detection chip was fabricated for noninvasive and on-site detection of sweat UA, which holds great application potential for personal health monitoring and point-of-care testing.

Skin Conductance-Based Acupoint and Non-Acupoint Recognition Using Machine Learning.

Acupoints (APs) prove to have positive effects on disease diagnosis and treatment, while intelligent techniques for the automatic detection of APs are not yet mature, making them more dependent on manual positioning. In this paper, we realize the skin conductance-based APs and non-APs recognition with machine learning, which could assist in APs detection and localization in clinical practice. Firstly, we collect skin conductance of traditional Five-Shu Point and their corresponding non-APs with wearable sensors, establishing a dataset containing over 36000 samples of 12 different AP types. Then, electrical features are extracted from the time domain, frequency domain, and nonlinear perspective respectively, following which typical machine learning algorithms (SVM, RF, KNN, NB, and XGBoost) are demonstrated to recognize APs and non-APs. The results demonstrate XGBoost with the best precision of 66.38%. Moreover, we also quantify the impacts of the differences among AP types and individuals, and propose a pairwise feature generation method to weaken the impacts on recognition precision. By using generated pairwise features, the recognition precision could be improved by 7.17%. The research systematically realizes the automatic recognition of APs and non-APs, and is conducive to pushing forward the intelligent development of APs and Traditional Chinese Medicine theories.