RESUMEN
Objective To explore the risk factors of depression and anxiety in adult patients with epilepsy and their relationship with quality of life.Methods From May 2022 to January 2023,patients diagnosed with epilepsy(aged≥18 years)in the department of neurology of our hospital were collected.General demographic data and disease-related information were recorded.Quality of life,depression and anxiety scales were measured in all patients.SPSS26.0 software was used for multiple linear regression analysis,multiple ordered Logistic regression analysis,rank sum test,Pearson correlation analysis,etc.Results Among the 111 patients,49.5%had depression and 43.2%had anxiety.Depression score and anxiety score were correlated with attack type,attack frequency,quality of life and right temporal lobe,and there was a significant negative correlation between life quality score and anxiety and depression score(P<0.01).Seizure frequency,seizure type and right temporal lobe were common risk factors for depression and anxiety in patients with epilepsy(P<0.05).Conclusion Epileptic depression and anxiety were affected by seizure frequency and seizure type,and this bad mood further affected the quality of life of patients.No clear link has been found between the lateralization of seizures and the presence of depression and anxiety states,and further research is needed.
RESUMEN
BACKGROUND:Traditional 3D dental segmentation methods usually utilize predefined spatial geometric features,such as curvature and normal vectors,as the reference information for tooth segmentation. OBJECTIVE:To propose an algorithm for complex 3D dental segmentation and deeply explore the correlation between segmentation results and application scenarios. METHODS:A 3D dental segmentation algorithm based on dual stream extraction of structural features and spatial features was established,and the modular design of split flow was used to avoid feature confusion.Among them,the attention mechanism on the structural feature flow was used to capture the fine-grained semantic information required for tooth segmentation,and the Tran Net based on the spatial feature flow was used to ensure the robustness of the model to complex tooth and jaw segmentation.This algorithm verified its effectiveness and reliability based on clinical datasets including healthy dental jaws and complex dental jaws such as missing teeth,malocclusion and dentition crowding.The segmentation performance of the model was measured in terms of overall accuracy,mean intersection over union,and directional cut discrepancy. RESULTS AND CONCLUSION:The overall segmentation accuracy of this algorithm in the clinical data set is 97.08%,and the segmentation effect is superior to that of other competitive methods from the qualitative and quantitative perspectives.It is verified that the structural feature flow designed in this paper can extract more precise local details of tooth shape from coordinate and normal information by constructing an attention aggregation mechanism,and the spatial feature flow designed in this paper can ensure the robustness of the model to complex teeth such as missing teeth,dislocated teeth,and crowded dentition by constructing a transformation network(Tran Net).Therefore,this tooth segmentation algorithm is highly reliable for clinicians'practical reference.
RESUMEN
Objective To optimize the ethanol extraction technology parameters of Jinlei Compound through orthogonal experiment combined with beetle antennae search(BAS)-back propagation(BP)neural network.Methods On the basis of the optimal extraction concentration obtained by single factor investigation,the ratio of solid to liquid,extraction time and extraction times were set as the orthogonal test factors.The entropy weight method was used to calculate the comprehensive scores of the yield of luteolin,kaempferol,swertianin and dry paste.Then,the BAS-BP neural network model was established,and the optimum extraction process was predicted by the BAS.Results BAS-BP neural network optimized Jinlei Compound alcohol extraction process was as follows:solid-liquid ratio 1:10,extraction time of 0.5 h,extraction times of 3,the comprehensive score was 96.352 6.The optimal process parameters obtained by orthogonal design were:solid-liquid ratio 1:10,extraction for 0.5 h,extraction for 3 times,the comprehensive score 90.988 0.The comprehensive score of BAS-BP neural network model was slightly better than that of orthogonal experiment,but the difference between the two was small.The optimal extraction process of Jinlei Compound was determined by comprehensive production practice as the ratio of solid to liquid 1:10,extraction for 0.5 h,extraction for 3 times.Conclusion The optimized process based on BAS-BP neural network has higher extraction efficiency and good stability,which can provide reference for subsequent development and quality control.
RESUMEN
Objective:To investigate the effectiveness and feasibility of a 3D U-Net in conjunction with a three-phase CT image segmentation model in the automatic segmentation of GTVnx and GTVnd in nasopharyngeal carcinoma.Methods:A total of 645 sets of computed tomography (CT) images were retrospectively collected from 215 nasopharyngeal carcinoma cases, including three phases: plain scan (CT), contrast-enhanced CT (CTC), and delayed CT (CTD). The dataset was grouped into a training set consisting of 172 cases and a test set comprising 43 cases using the random number table method. Meanwhile, six experimental groups, A1, A2, A3, A4, B1, and B2, were established. Among them, the former four groups used only CT, only CTC, only CTD, and all three phases, respectively. The B1 and B2 groups used phase fine-tuning CTC models. The Dice similarity coefficient (DSC) and 95% Hausdorff distance (HD95) served as quantitative evaluation indicators.Results:Compared to only monophasic CT (group A1/A2/A3), triphasic CT (group A4) yielded better result in the automatic segmentation of GTVnd (DSC: 0.67 vs. 0.61, 0.64, 0.64; t = 7.48, 3.27, 4.84, P < 0.01; HD95: 36.45 vs. 79.23, 59.55, 65.17; t = 5.24, 2.99, 3.89, P < 0.01), with statistically significant differences ( P < 0.01). However, triphasic CT (group A4) showed no significant enhancement in the automatic segmentation of GTVnx compared to monophasic CT (group A1/A2/A3) (DSC: 0.73 vs. 0.74, 0.74, 0.73; HD95: 14.17 mm vs. 8.06, 8.11, 8.10 mm), with no statistically significant difference ( P > 0.05). For the automatic segmentation of GTVnd, group B1/B2 showed higher automatic segmentation accuracy compared to group A1 (DSC: 0.63, 0.63 vs. 0.61, t = 4.10, 3.03, P<0.01; HD95: 58.11, 50.31 mm vs. 79.23 mm, t = 2.75, 3.10, P < 0.01). Conclusions:Triphasic CT scanning can improve the automatic segmentation of the GTVnd in nasopharyngeal carcinoma. Additionally, phase fine-tuning models can enhance the automatic segmentation accuracy of the GTVnd on plain CT images.
RESUMEN
OBJECTIVE Optimizing the water extraction technology of Xiangqin jiere granules. METHODS The orthogonal test of 3 factors and 3 levels was designed, and comprehensive scoring was conducted for the above indexes by using G1-entropy weight to obtain the optimized water extraction technology of Xiangqin jiere granules with water addition ratio, extraction time and extraction times as factors, using the contents of forsythoside A, baicalin, phillyrin, oroxylin A-7-O-β-D-glycoside, wogonoside, baicalein and wogonin, and extraction rate as evaluation indexes. BP neural network modeling was used to optimize the network model and water extraction process using the results of 9 groups of orthogonal tests as test and training data, the water addition multiple, decocting time and extraction times as input nodes, and the comprehensive score as output nodes. Then the two analysis methods were compared by verification test to find the best water extraction process parameters. RESULTS The water extraction technology optimized by the orthogonal test was 8-fold water, extracting 3 times, extracting for 1 h each time. Comprehensive score was 96.84 (RSD=0.90%). The optimal water extraction technology obtained by BP neural network modeling included 12-fold water, extracting 4 times, extracting for 0.5 h each time. The comprehensive score was 92.72 (RSD=0.77%), which was slightly lower than that of the orthogonal test. CONCLUSIONS The water extraction technology of Xiangqin jiere granules is optimized successfully in the study, which includes adding 8-fold water, extracting 3 times, and extracting for 1 hour each time.
RESUMEN
Objective To build a neural network based on the Unet infrastructure for recognition and segmentation of two-dimensional calcium imaging fluorescence images.Methods The in vivo miniaturized two-photon microscope(mTPM)was used for brain calcium imaging in freely moving mice.The imaging data was motion corrected using the NoRMCorre algorithm and processed using ImageJ software to obtain the original images after correction,and the labels were produced using the Labelme software.The neural network HDCGUnet was built using the original images and labels for training,and optimized to improve the model structure according to the training effect.Finally,the evaluation indexes were selected and compared with those of other models to verify the utility of this model.Results The HDCGUnet model,which was collected and made on our own,performed best in the two-photon calcium imaging dataset compared to other models,and performed well on the BBBC dataset either.Conclusion The HDCGUnet model provides a novel alternative for the recognition and segmentation of two-photon calcium imaging images.
RESUMEN
Objective To explore the establishment of a prognostic model based on machine learning algorithm to predict primary graft dysfunction (PGD) in patients with idiopathic pulmonary fibrosis (IPF) after lung transplantation. Methods Clinical data of 226 IPF patients who underwent lung transplantation were retrospectively analyzed. All patients were randomly divided into the training and test sets at a ratio of 7:3. Using regularized logistic regression, random forest, support vector machine and artificial neural network, the prognostic model was established through variable screening, model establishment and model optimization. The performance of this prognostic model was assessed by the area under the receiver operating characteristic curve (AUC), positive predictive value, negative predictive value and accuracy. Results Sixteen key features were selected for model establishment. The AUC of the four prognostic models all exceeded 0.7. DeLong and McNemar tests found no significant difference in the performance among different models (both P>0.05). Conclusions Based on four machine learning algorithms, the prognostic model for grade 3 PGD after lung transplantation is preliminarily established. The overall prediction performance of each model is similar, which may predict the risk of grade 3 PGD in IPF patients after lung transplantation.
RESUMEN
ObjectiveDirect continuous monitoring of arterial blood pressure is invasive and continuous monitoring cannot be achieved by traditional cuffed indirect blood pressure measurement methods. Previously, continuous non-invasive arterial blood pressure monitoring was achieved by using photoplethysmography (PPG), but it is discrete values of systolic and diastolic blood pressures rather than continuous values constructing arterial blood pressure waves. This study aimed to reconstruct arterial blood pressure wave signal based on CNN-LSTM using PPG to achieve continuous non-invasive arterial blood pressure monitoring. MethodsA CNN-LSTM hybrid neural network model was constructed, and the PPG and arterial blood pressure wave synchronized recorded signal data from the Medical Information Mart for Intensive Care (MIMIC) were used. The PPG signals were input to this model after noise reduction, normalization, and sliding window segmentation. The corresponding arterial blood pressure waves were reconstructed from PPG by using the CNN-LSTM hybrid model. ResultsWhen using the CNN-LSTM neural network with a window length of 312, the error between the reconstructed arterial blood pressure values and the actual arterial blood pressure values was minimal: the values of mean absolute error (MAE) and root mean square error (RMSE) were 2.79 mmHg and 4.24 mmHg, respectively, and the cosine similarity is the optimal. The reconstructed arterial blood pressure values were highly correlated with the actual arterial blood pressure values, which met the Association for the Advancement of Medical Instrumentation (AAMI) standards. ConclusionCNN-LSTM hybrid neural network can reconstruct arterial blood pressure wave signal using PPG to achieve continuous non-invasive arterial blood pressure monitoring.
RESUMEN
BACKGROUND:Rapid developments in the field of bioinformatics have provided new methods for the diagnosis of osteoarthritis.Artificial neural networks have powerful data computing and classification capabilities,which have shown better performance in disease diagnosis. OBJECTIVE:To establish a new diagnostic predictive model of osteoarthritis based on artificial neural network and to verify the diagnostic value of the model in osteoarthritis with an external dataset. METHODS:The eligible osteoarthritis-related data sets were downloaded through GEO database search and divided into Train group and Test group.The gene expression matrix of the Train group was analyzed to screen the differentially expressed genes.GO and KEGG enrichment analyses were performed on the differentially expressed genes.Through Lasso regression model,support vector machine model and random forest tree model,the key genes of osteoarthritis were further identified from the differentially expressed genes.The R software"Neuralnet"package was then used to construct the osteoarthritis diagnosis model based on artificial neural network,and the model performance was evaluated by the five-fold cross-validation.Two independent data sets in the Test group were used to verify their diagnostic results. RESULTS AND CONCLUSION:A total of 90 differentially expressed genes related to osteoarthritis were obtained by differential analysis,of which 33 were down-regulated and 57 were up-regulated.GO enrichment analysis showed that the differentially expressed genes were mainly involved in the following biological processes,including leukocyte-mediated immunity,leukocyte migration in bone marrow and chemokine production.KEGG enrichment analysis showed that these genes were mainly enriched in rheumatoid arthritis,interleukin-17 signaling pathway and osteoclast differentiation pathway.Five key genes for the diagnosis of osteoarthritis,HMGB2,GADD45A,SLC19A2,TPPP3 and FOLR2,were identified by three machine learning methods.The artificial neural network model of five key genes in the Train group showed that the accuracy was 96.36%and the area under the curve was 0.997.The five-fold cross validation of the neural network model showed that the average area under the curve was greater than 0.9 and the model was of robustness.Two independent data sets in the Test group showed its area under the curve was 0.814 and 0.788 respectively.Therefore,the establishment of an artificial neural network model for the diagnosis of osteoarthritis has a certain diagnostic value.
RESUMEN
BACKGROUND:With the continuous improvement and progress of artificial intelligence technology in the treatment of spinal deformity,a large number of studies have been invested in this field,but the main research status,hot spots and development trends are still unclear. OBJECTIVE:To visually analyze the literature related to artificial intelligence in the field of spinal deformities by using bibliometrics,identify the research hotspots and shortcomings in this field,and provide references for future research. METHODS:The core database of Web of Science was used to search the articles related to artificial intelligence in the field of spinal deformities published from inception to 2023.The data were visually analyzed by Citespace 5.6.R5 and VOSviewer 1.6.19. RESULTS AND CONCLUSION:(1)A total of 165 papers were included,and the number of papers published in this field showed a fluctuating upward trend.The author with the largest number of articles is Lafage V,and the country with the largest number of articles is China.(2)Keyword analysis results show that adolescent scoliosis,deep learning,classification,precision and robot are the main keywords.(3)The in-depth analysis results of co-cited and highly cited articles show that artificial intelligence has three hotspots in the field of spinal deformities,including the use of U-shaped architecture(a mature mode of deep learning convolutional neural networks)to automatically measure imaging parameters(Cobb angle and accurate segmentation of paraspinal muscles),multi-view correlation network architecture(i.e.,spine curvature assessment framework),and robot-guided spinal surgery.(4)In the field of artificial intelligence treatment of spinal malformations,the mechanism research such as genomics is very weak.In the future,unsupervised hierarchical clustering and other machine learning techniques can be used to study the basic mechanism of susceptibility genes in the field of spinal deformities by genome-wide association analysis and other genomics research methods.
RESUMEN
Objective To evaluate the relationship between diabetic nephropathy(DN)and diabetic retinopathy(DR)in patients with type 2 diabetes mellitus(T2DM)based on imaging and clinical testing data.Methods Totally 600 T2DM patients who visited the First People's Hospital of Ziyang from March 2021 to December 2022 were included.The fundus photography and fundus fluorescein angiography were performed on all these patients and their age,gender,T2DM duration,cardiovascular diseases,cerebrovascular disease,hypertension,smoking history,drinking history,body mass in-dex,systolic blood pressure,diastolic blood pressure and other clinical data were collected.The levels of fasting blood glu-cose(FPG),triglyceride(TG),total cholesterol(TC),high-density lipoprotein cholesterol(HDL-C),low-density lipo-protein cholesterol(LDL-C),glycosylated hemoglobin(HbA1c),24 h urinary albumin(UAlb),urinary albumin to creati-nine ratio(ACR),serum creatinine(Scr)and blood urea nitrogen(BUN)were measured.Logistic regression was used to analyze the risk factors associated with DR.DR staging was performed according to fundus images,and the convolutional neural network(CNN)algorithm was used as an image analysis method to explore the correlation between DR and DN based on emission computed tomography(ECT)and clinical testing data.Results The average lesion area rates of DR and DN detected by the CNN in the non-DR,mild-non-proliferative DR(NPDR),moderate-NPDR,severe-NPDR and pro-liferative DR(PDR)groups were higher than those obtained by the traditional algorithm(TCM).As DR worsened,the Scr,BUN,24 h UAlb and ACR gradually increased.Besides,the incidence of DN in the non-DR,mild-NPDR,moderate-NPDR,severe-NPDR and PDR groups was 1.67%,8.83%,16.16%,22.16%and 30.83%,respectively.Logistic regression analysis showed that the duration of T2DM,smoking history,HbA1c,TC,TG,HDL-C,LDL-C,24 h UAlb,Scr,BUN,ACR and glomerular filtration rate(GFR)were independent risk factors for DR.Renal dynamic ECT analysis demonstrated that with the aggravation of DR,renal blood flow perfusion gradually decreased,resulting in diminished renal filtration.Conclusion The application of CCN in the early stage DR and DN image analysis of T2DM patients will improve the diag-nosis accuracy of DR and DN lesion area.The DN is worsening as the aggravation of DR.
RESUMEN
Pathology is the gold standard for diagnosis of neoplastic diseases.Whole slide imaging turns traditional slides into digital images,and artificial intelligence has shown great potential in pathological image analysis,especially deep learning models.The application of artificial intelligence in whole slide imaging of lung cancer involves many aspects such as histopathological classification,tumor microenvironment analysis,efficacy and survival prediction,etc.,which is expected to assist clinical decision-making of accurate treatment.Limitations in this field include the lack of precisely annotated data and slide quality varying among institutions.Here we summarized recent research in lung cancer pathology image analysis leveraging artificial intelligence and proposed several future directions.
RESUMEN
Objective:By analyzing the cost change trend,internal structure change and main influencing factors of diabetes inpatients'hospitalization expenses,it provides empirical basis for promoting the reform of medical service prices,optimizing the internal structure of hospitalization expenses,effectively controlling hospitalization expenses,and reducing the economic burden of diabetes inpatients.Methods:Using the first page data of medical records of 13 426 diabetes inpatients in the target hospital from 2017 to 2021,it analyzes the structural change of diabetes inpatients'hospitalization expenses by using structural change degree and grey correlation degree methods,and analyzes the influencing factors of hospitalization expenses by using linear regression and BP neural network model.Results:Drug expenses and medical technology expenses are the top two in the proportion of total hospitalization expenses of discharged patients with diabetes,and they account for a large proportion in the total hospitalization expenses.The results of structural change and grey correlation show that drug expenses and medical technology expenses are these two factors that cause changes in the total hospitalization cost structure and have a high correlation with the total hospitalization cost,with a cumulative contribution rate of 90.50%.According to the results of linear regression and neural network model,the length of stay is the most important factor affecting the total cost of hospitalization of diabetes patients,followed by the number of operations/procedures and diagnoses.Conclusion:The internal composition of hospitalization expenses for diabetes patients is unreasonable.The proportion of drug expenses and medical technology expenses is too high.The proportion of medical and nursing expenses reflecting the technical labor value of medical personnel is relatively low.The structure of medical income needs to be further optimized.The length of stay is the most critical factor affecting the hospitalization expenses of diabetes patients.Reasonable control of the length of stay can effectively control the unreasonable growth of medical expenses and reduce the economic burden of diabetes patients.
RESUMEN
Objective:To explore the value of cardiodynamicsgram (CDG) obtained from electrocardiogram (ECG) data by radial basis functionradial basis function (RBF) neural network in early diagnosis of patients with acute coronary syndrome (ACS).Methods:Retrospective analysis method was used. Patients with chest pain as the main initial symptom in the emergency department of Baoan District People's Hospital of Shenzhen from October 2021 to September 2022 were enrolled. Baseline data were collected, including gender, age, smoking history, family history of coronary heart disease and history of hypertension, diabetes, hyperlipidemia, and atherosclerosis. The first 12-lead ECG was recorded after admission to the emergency department, and electrocardiodynamics analysis was performed to generate CDG. Receiver operator characteristic curve (ROC curve) was plotted to analyze the value of CDG and ECG in the early diagnosis of ACS and non-ST segment elevation ACS (NSTE-ACS). Sensitivity, specificity, area under the ROC curve (AUC), and 95% confidence interval (95% CI) were calculated. CDG and coronary angiography results of 3 patients with ACS with normal ECG were observed and analyzed. Non-ACS patients with normal ECG but positive CDG were followed for 30 days for adverse cardiovascular events. Results:A total of 384 patients with chest pain were included, including 169 patients with ACS and 215 patients without ACS. The proportion of male (87.0% vs. 53.0%), smoking history (37.9% vs. 12.1%), hypertension (46.2% vs. 22.3%), diabetes (24.3% vs. 7.9%), hyperlipidemia (55.0% vs. 14.0%) and history of atherosclerosis (22.5% vs. 2.3%) in ACS group were significantly higher than those in non-ACS group (all P < 0.05). The ROC curve showed that the AUC of CDG diagnosis of ACS was higher than that of ECG [AUC (95% CI): 0.88 (0.66-0.76) vs. 0.71 (0.84-0.92)], the sensitivity was 92.8%, 78.6%, and the specificity was 83.3%, 64.2%, respectively. The AUC of CDG diagnosis of NSTE-ACS was higher than that of ECG [AUC (95% CI): 0.85 (0.80-0.90) vs. 0.63 (0.56-0.69)], the sensitivity was 87.1%, 61.3%, and the specificity was 83.3%, 64.2%, respectively. CDG of 3 patients with ACS with normal ECG showed disordered state, and coronary angiography showed ≥70% stenosis of major coronary branches. Of 215 non-ACS patients, 20 had a normal ECG but positive CDG, and 3 developed ST segment elevation myocardial infarction (STEMI) within 30 days, and 2 developed unstable angina (UA) within 30 days. Conclusion:CDG has high value in early diagnosis of ACS patients and is expected to become an important means of early diagnosis of ACS in emergency.
RESUMEN
Aiming at the problem of large force tracking errors caused by environmental stiffness changes when dual-arm robot is assisting in opening soft tissues in head and neck surgery,an adaptive admittance control strategy based on radial basis function(RBF)neural network is proposed for reducing force tracking error and improving system response speed.By using RBF neural network to adjust admittance parameters online during surgery,the adaptability of the robotic arm to different contact conditions and operation requirements is improved,thereby realizing fast and accurate force tracking.The simulation experiment introduces the adaptive admittance control strategy based on RBF neural network into the dual-arm force synchronous admittance control system and compares it with the traditional fixed-parameter admittance control to prove its contact force control effect under the condition of variable contact environment parameters.The results demonstrate that the adaptive admittance control strategy based on RBF neural network can effectively improve the force tracking accuracy,response speed and anti-interference capability of dual-arm surgical robot.
RESUMEN
A spatial-temporal convolutional neural network-based method is proposed for schizophrenia classification.Unlike the mainstream methods that only analyze the temporal frequency features in EEG and ignore the spatial features between brain regions,the model mainly obtains the spatial-frequency features by convolving the adjacency matrix composed of wavelet coherence coefficients between different channels and EEG sequences,and then extracts the temporal-frequency features through one-dimensional temporal convolution.The processed matrix is flattened after multiple convolutions and input to the classification model.Experimental results show that the method has a classification accuracy of 96.32%on the publicly available dataset Zenodo,demonstrating its effectiveness and exhibiting the advantages of fusing temporal-frequency and spatial-frequency features for schizophrenia diagnosis.
RESUMEN
In order to address issues such as the decline in diagnostic performance of deep learning models due to imbalanced data distribution in psoriasis vulgaris,a VGG13-based deep convolutional neural network model is proposed by integrating the processing capability of the improved fuzzy KMeans clustering algorithm for highly clustered complex data and the predictive capability of VGG13 deep convolutional neural network model.The model is applied to the diagnosis of psoriasis vulgaris,and the experimental results indicate that compared with VGG13 and resNet18,the proposed approach based on deep learning and improved fuzzy KMeans is more suitable for identifying psoriasis features.
RESUMEN
Objective To establish a hybrid deep learning lung sound classification model based on convolutional neural network(CNN)-long short-term memory(LSTM)for electronic auscultation.Methods Wavelet transform was used to extract features from the dataset,transforming lung sound signals into energy entropy,peak value and other features.On this basis,a classification model based on hybrid algorithm incorporating CNN and LSTM neural network was constructed.The features extracted by wavelet transform were input into CNN module to obtain the spatial features of the data,and then the temporal features were detected through LSTM module.The fusion of the two types of features enabled the classification of lung sounds through the model,thereby assisting in the diagnosis of pulmonary diseases.Results The accuracy rate and F1 score of CNN-LSTM hybrid model were significantly higher than those of other single models,reaching 0.948 and 0.950.Conclusion The proposed CNN-LSTM hybrid model demonstrates higher accuracy and more precise classification,showcasing broad potential application value in intelligent auscultation.
RESUMEN
To address the problem of low accuracy in multi-classification recognition of motor imagery electroencephalogram(EEG)signals,a recognition method is proposed based on differential entropy and convolutional neural network for 4-class classification of motor imagery.EEG signals are extracted into 4 frequency bands(Alpha,Beta,Theta,and Gamma)through the filter,followed by the computation of differential entropy for each frequency band.According to the spatial characteristics of brain electrodes,the data structure is reconstructed into three-dimensional EEG signal feature cube which is input into convolutional neural network for 4-class classification.The method achieves an accuracy of 95.88%on the BCI Competition IV-2a public dataset.Additionally,a 4-class classification motor imagery dataset is established in the laboratory for the same processing,and an accuracy of 94.50%is obtained.The test results demonstrate that the proposed method exhibits superior recognition performance.
RESUMEN
Aldehyde oxidase (AOX) is a molybdoenzyme that is primarily expressed in the liver and is involved in the metabolism of drugs and other xenobiotics. AOX-mediated metabolism can result in unexpected outcomes, such as the production of toxic metabolites and high metabolic clearance, which can lead to the clinical failure of novel therapeutic agents. Computational models can assist medicinal chemists in rapidly evaluating the AOX metabolic risk of compounds during the early phases of drug discovery and provide valuable clues for manipulating AOX-mediated metabolism liability. In this study, we developed a novel graph neural network called AOMP for predicting AOX-mediated metabolism. AOMP integrated the tasks of metabolic substrate/non-substrate classification and metabolic site prediction, while utilizing transfer learning from 13C nuclear magnetic resonance data to enhance its performance on both tasks. AOMP significantly outperformed the benchmark methods in both cross-validation and external testing. Using AOMP, we systematically assessed the AOX-mediated metabolism of common fragments in kinase inhibitors and successfully identified four new scaffolds with AOX metabolism liability, which were validated through in vitro experiments. Furthermore, for the convenience of the community, we established the first online service for AOX metabolism prediction based on AOMP, which is freely available at https://aomp.alphama.com.cn.