Objective To establish an analytical strategy covering multi-dataset processing, recursive feature elimination (RFE) screening and multi-model evaluation based on multiple machine learning algorithms, so as to screen radiation-sensitive genes and verify the feasibility of the evaluation strategy. Methods Qualified radiation transcriptome datasets were retrieved from public gene expression databases. Following standardized data preprocessing and feature preselection, 13 machine learning algorithms were adopted to construct models. The performance of each model was compared and validated in independent datasets. Results A total of 38 eligible datasets were included. Sixteen differentially expressed genes unreported in existing literature were screened out, among which ugcrhl, pdcl3, mct4, h2-g2 and fam120aos were correlated with radiation phenotypes. Ensemble learning algorithms including random forest and gradient boosting exhibited the optimal comprehensive performance. Independent dataset verification confirmed that the screened genes overlapped with known radiation-sensitive genes, and the model performance was consistent with the findings. Conclusion The machine learning strategy constructed in this study can effectively explore potential radiation-sensitive genes, and provides methodological support for subsequent relevant studies.

Objective To establish an analytical strategy covering multi-dataset processing, recursive feature elimination (RFE) screening and multi-model evaluation based on multiple machine learning algorithms, so as to screen radiation-sensitive genes and verify the feasibility of the evaluation strategy. Methods Qualified radiation transcriptome datasets were retrieved from public gene expression databases. Following standardized data preprocessing and feature preselection, 13 machine learning algorithms were adopted to construct models. The performance of each model was compared and validated in independent datasets. Results A total of 38 eligible datasets were included. Sixteen differentially expressed genes unreported in existing literature were screened out, among which ugcrhl, pdcl3, mct4, h2-g2 and fam120aos were correlated with radiation phenotypes. Ensemble learning algorithms including random forest and gradient boosting exhibited the optimal comprehensive performance. Independent dataset verification confirmed that the screened genes overlapped with known radiation-sensitive genes, and the model performance was consistent with the findings. Conclusion The machine learning strategy constructed in this study can effectively explore potential radiation-sensitive genes, and provides methodological support for subsequent relevant studies.

Objective To establish an analytical strategy covering multi-dataset processing, recursive feature elimination (RFE) screening and multi-model evaluation based on multiple machine learning algorithms, so as to screen radiation-sensitive genes and verify the feasibility of the evaluation strategy. Methods Qualified radiation transcriptome datasets were retrieved from public gene expression databases. Following standardized data preprocessing and feature preselection, 13 machine learning algorithms were adopted to construct models. The performance of each model was compared and validated in independent datasets. Results A total of 38 eligible datasets were included. Sixteen differentially expressed genes unreported in existing literature were screened out, among which ugcrhl, pdcl3, mct4, h2-g2 and fam120aos were correlated with radiation phenotypes. Ensemble learning algorithms including random forest and gradient boosting exhibited the optimal comprehensive performance. Independent dataset verification confirmed that the screened genes overlapped with known radiation-sensitive genes, and the model performance was consistent with the findings. Conclusion The machine learning strategy constructed in this study can effectively explore potential radiation-sensitive genes, and provides methodological support for subsequent relevant studies.

OBJECTIVES: To propose a single repetition time (TR) quantitative magnetic susceptibility imaging method for the lumbar spine using bipolar readout gradient, and compare the quantitative magnetic susceptibility measurement using single TR and dual TR methods for the lumbar spine with different bone densities. METHODS: A translation correction method was proposed to correct spatial misalignment along the frequency encoding direction between positive and negative gradient readout images, and the phase difference between the images was eliminated using a phase correction method. The data of lumbar vertebrae L1-L5 were collected using single TR and dual TR methods from 6 normal individuals, 2 patients with osteopenia, and 2 patients with osteoporosis. The magnetic susceptibility map was reconstructed, the quantitative results of single TR before and after correction were compared with those of the dual TR method. RESULTS: The linear regression result of the lumbar spine magnetic susceptibility values obtained by the single TR method before calibration and the dual TR method is Y=0.64*X-11.61. The linear regression result of the lumbar spine magnetic susceptibility values corrected by the single TR method and the dual TR method is Y=1.03*X+0.25. The results of the corrected single TR method were highly consistent with those of the dual TR method, and the calibrated single TR method could effectively distinguish osteopenia and osteoporosis patients from normal individuals. CONCLUSIONS The calibrated single TR bipolar readout gradient method can generate artifact-free lumbar spine quantitative magnetic susceptibility distribution maps and reduce data acquisition time by 50%.
Humans ; Lumbar Vertebrae/pathology* ; Magnetic Resonance Imaging/methods* ; Female ; Middle Aged ; Male ; Osteoporosis/diagnosis* ; Adult ; Bone Density ; Aged ; Bone Diseases, Metabolic/diagnosis*

Objective To explore the value of improved ResNet18 lightweight deep learning(DL)models for automatically detecting gouty arthritis(GA)based on ultrasonogram of the first metatarsophalangeal joint(MTP1).Methods A total of 2 401 ultrasonograms obtained from 260 patients with suspected gout who underwent MTP1 ultrasound examination were included and divided into training set(1 910 ultrasonograms from 209 cases)and test set(491 ultrasonograms from 51 cases)at the ratio of 4∶1.GA lesions on ultrasonograms were manually labeled.After preprocessing,ResNet18 lightweight network was used to construct DL models for identifying the ultrasonogram category was normal or abnormal(with any manifestation of GA).Five-fold cross-validation method was adopted to evaluate the efficacy of the DL models constructed with 2,3,4 or 6 residual blocks,i.e.model 1,2,3 and 4,respectively,and the computational cost and the amount of parameters of each model were recorded.The efficacy of the models were verified using test set,and the best DL model was screened.Results The computational cost of model 1,2,3 and 4 was 7 558.27,2 963.73,4 012.33 and 6 093.39 M,respectively,while the amount of parameters was 4.61,4.91,4.91 and 5.28 M,respectively.Model 2 had the least computational cost with parameters only slightly more than model 1.In test set,no significant difference of accuracy nor the area under the curve was found among 4 models(all P＞0.05).The sensitivity of model 2 was higher than that of model 3,while its specificity was lower only than that of model 3(both P＜0.05),hence model 2 was the best DL model.Conclusion Improved ResNet18 lightweight DL models could be used for automatically detecting GA based on ultrasonogram of MTP1,among which model 2 was the best one.

Objective To explore the value of improved ResNet18 lightweight deep learning(DL)models for automatically detecting gouty arthritis(GA)based on ultrasonogram of the first metatarsophalangeal joint(MTP1).Methods A total of 2 401 ultrasonograms obtained from 260 patients with suspected gout who underwent MTP1 ultrasound examination were included and divided into training set(1 910 ultrasonograms from 209 cases)and test set(491 ultrasonograms from 51 cases)at the ratio of 4∶1.GA lesions on ultrasonograms were manually labeled.After preprocessing,ResNet18 lightweight network was used to construct DL models for identifying the ultrasonogram category was normal or abnormal(with any manifestation of GA).Five-fold cross-validation method was adopted to evaluate the efficacy of the DL models constructed with 2,3,4 or 6 residual blocks,i.e.model 1,2,3 and 4,respectively,and the computational cost and the amount of parameters of each model were recorded.The efficacy of the models were verified using test set,and the best DL model was screened.Results The computational cost of model 1,2,3 and 4 was 7 558.27,2 963.73,4 012.33 and 6 093.39 M,respectively,while the amount of parameters was 4.61,4.91,4.91 and 5.28 M,respectively.Model 2 had the least computational cost with parameters only slightly more than model 1.In test set,no significant difference of accuracy nor the area under the curve was found among 4 models(all P＞0.05).The sensitivity of model 2 was higher than that of model 3,while its specificity was lower only than that of model 3(both P＜0.05),hence model 2 was the best DL model.Conclusion Improved ResNet18 lightweight DL models could be used for automatically detecting GA based on ultrasonogram of MTP1,among which model 2 was the best one.

BACKGROUND:Due to the sudden release and the rapid removal by proteases,platelet-rich plasma hydrogel leads to shorter residence times of growth factors at the wound site.In recent years,researchers have focused on the use of hydrogels to encapsulate platelet-rich plasma in order to improve the deficiency of platelet-rich plasma hydrogels. OBJECTIVE:To prepare self-assembled polypeptide-platelet-rich plasma hydrogel and to explore its effects on the release of bioactive factors of platelet-rich plasma. METHODS:The self-assembled polypeptide was synthesized by the solid-phase synthesis method,and the solution was prepared by D-PBS.Hydrogels were prepared by mixing different volumes of polypeptide solutions with platelet-rich plasma and calcium chloride/thrombin solutions,so that the final mass fraction of polypeptides in the system was 0.1%,0.3%,and 0.5%,respectively.The hydrogel state was observed,and the release of growth factors in platelet-rich plasma was detected in vitro.The polypeptide self-assembly was stimulated by mixing 1%polypeptide solution with 1%human serum albumin solution,so that the final mass fraction of the polypeptide was 0.1%,0.3%,and 0.5%,respectively.The flow state of the liquid was observed,and the rheological mechanical properties of the self-assembled polypeptide were tested.The microstructure of polypeptide(mass fraction of 0.1%and 0.001%)-human serum albumin solution was observed by scanning electron microscope and transmission electron microscope. RESULTS AND CONCLUSION:(1)Hydrogels could be formed between different volumes of polypeptide solution and platelet-rich plasma.Compared with platelet-rich plasma hydrogels,0.1%and 0.3%polypeptide-platelet-rich plasma hydrogels could alleviate the sudden release of epidermal growth factor and vascular endothelial growth factor,and extend the release time to 48 hours.(2)After the addition of human serum albumin,the 0.1%polypeptide group still exhibited a flowing liquid,the 0.3%polypeptide group was semi-liquid,and the 0.5%polypeptide group stimulated self-assembly to form hydrogel.It was determined that human serum albumin in platelet-rich plasma could stimulate the self-assembly of polypeptides.With the increase of the mass fraction of the polypeptide,the higher the storage modulus of the self-assembled polypeptide,the easier it was to form glue.(3)Transmission electron microscopy exhibited that the polypeptide nanofibers were short and disordered before the addition of human serum albumin.After the addition of human serum albumin,the polypeptide nanofibers became significantly longer and cross-linked into bundles,forming a dense fiber network structure.Under a scanning electron microscope,the polypeptides displayed a disordered lamellar structure before adding human serum albumin.After the addition of human serum albumin,the polypeptides self-assembled into cross-linked and densely arranged porous structures.(4)In conclusion,the novel polypeptide can self-assemble triggered by platelet-rich plasma and the self-assembly effect can be accurately adjusted according to the ratio of human serum albumin to polypeptide.This polypeptide has a sustained release effect on the growth factors of platelet-rich plasma,which can be used as a new biomaterial for tissue repair.

Background::Hypothermia therapy has been suggested to attenuate myocardial necrosis; however, the clinical implementation as a valid therapeutic strategy has failed, and new approaches are needed to translate into clinical applications. This study aimed to assess the feasibility, safety, and efficacy of a novel selective intracoronary hypothermia (SICH) device in mitigating myocardial reperfusion injury.Methods::This study comprised two phases. The first phase of the SICH was performed in a normal porcine model for 30 minutes ( n = 5) to evaluate its feasibility. The second phase was conducted in a porcine myocardial infarction (MI) model of myocardial ischemia/reperfusion which was performed by balloon occlusion of the left anterior descending coronary artery for 60 minutes and maintained for 42 days. Pigs in the hypothermia group ( n = 8) received hypothermia intervention onset reperfusion for 30 minutes and controls ( n = 8) received no intervention. All animals were followed for 42 days. Cardiac magnetic resonance analysis (five and 42 days post-MI) and a series of biomarkers/histological studies were performed. Results::The average time to lower temperatures to a steady state was 4.8 ± 0.8 s. SICH had no impact on blood pressure or heart rate and was safely performed without complications by using a 3.9 F catheter. Interleukin-6 (IL-6), tumor necrosis factor-α, C-reactive protein (CRP), and brain natriuretic peptide (BNP) were lower at 60 min post perfusion in pigs that underwent SICH as compared with the control group. On day 5 post MI/R, edema, intramyocardial hemorrhage, and microvascular obstruction were reduced in the hypothermia group. On day 42 post MI/R, the infarct size, IL-6, CRP, BNP, and matrix metalloproteinase-9 were reduced, and the ejection fraction was improved in pigs that underwent SICH.Conclusions::The SICH device safely and effectively reduced the infarct size and improved heart function in a pig model of MI/R. These beneficial effects indicate the clinical potential of SICH for treatment of myocardial reperfusion injury.

Objective:To design an improved lamina hook system and compare its biomechanical properties with traditional lamina hook system in fixation of lumbar spondylolysis.Methods:The thin layer CT data of the lumbosacral vertebrae of 20 healthy young male servicemen who underwent physical examination in the outpatient department of the 940th Hospital of Joint Logistics Support Force of PLA from January 2021 to August 2022 were collected. The age of the subjects was 20-30 years [(25.0±3.0)years]. A 3-dimensional model of the L 5 vertebral body was constructed using the 3-dimensional modeling software. The new improved lamina hook was designed according to the measurements including the thickness of the middle area, the longest longitudinal diameter, the curvature radius of the lower edge, the angle between the upper and lower tail ends, the thickness of the lower edge, and the longest diameter of the lower edge of the bilateral L 5 vertebral plates. One serviceman was selected from the aforementioned group to construct a linear finite element model of segments L 4-S using the 3-dimensional virtual software (normal model, model A), based on which, the L 5 bilateral spondylolysis model (model B), improved lamina hook model (model C) and traditional lamina hook models (model D) were designed. By constraining both sides of the sacrum and applying a longitudinal load of 400 N on the L 4 vertebral body, the upper 1/3 gravity of the body was simulated, and with a bending moment of 10 N·m along the X, Y, and Z directions, motions of forward flexion, backward extension, lateral bending, rotation, etc were simulated. The range of motion of segment L 4/5 and L 5/S 1 of model A was evaluated and compared with the findings of the previous researches to verify its effectiveness. The overall range of motion of models A, B, C, and D, the range of motion of segment L 4/5 and L 5/S 1, the maximum overall displacement, the maximum displacement and stress of the isthmus, the stress distribution and maximum stress of internal fixation of models C and D, and the stress distribution and maximum stress of the vertebral body of models C and D were compared. Results:(1) During forward flexion, backward extension, lateral bending and rotation, the range of motion of model A was 5.01°, 4.03°, 3.91° and 1.42° in segment L 4/5, and was 4.62°, 2.51°, 2.40° and 1.23° in segment L 5/S 1. (2) The overall range of motion, range of motion of segment L 4/5 and L 5/S 1 and maximum overall displacement of models A, C, and D were similar in axial compression, forward flexion, backward extension, left bending, and left rotation, while those of model B were significantly increased. (3) There was no significant difference in the maximum displacement of the isthmus of models A, C, and D under different motion modes, while the maximum displacement of model B in the isthmus was significantly larger than that of models A, C, and D, especially during rotation, increased by 295%, 277%, and 276% respectively. The maximum stress of the isthmus of model C was 0.938 MPa, 1.698 MPa, 0.410 MPa, 2.775 MPa, and 1.554 MPa respectively. The maximum stress in the isthmus of model D was 0.590 MPa, 1.297 MPa, 0.520 MPa, 3.088 MPa, and 2.072 MPa respectively. The maximum stress of the isthmus of models C and D was similar during axial compression and forward flexion, while the stress of the isthmus of model C was smaller than that of model D during backward extension, lateral bending, and rotation, decreased by 21.1%, 10.2%, and 25.0% respectively compared with model D. (4) The maximum stress of internal fixation in models C and D during forward flexion, backward extension, left bending, and left rotation was 135.220 MPa, 130.180 MPa, 200.940 MPa and 306.340 MPa respectively, and was 131.840 MPa, 112.280 MPa, 349.980 MPa and 370.140 MPa respectively. The maximum stress of internal fixation in the two models of internal fixation during forward flexion and backward extension was similar, while it was decreased by 42.6% and 17.2% in model C during left bending and left rotation, compared with model D. (5) The maximum stress of the vertebral body during forward flexion, backward extension, left bending, and left rotation was 79.787 MPa, 36.857 MPa, 37.943 MPa and 96.965 MPa respectively in model C, but was 80.104 MPa, 64.236 MPa, 196.010 MPa and 193.020 MPa respectively in model D. The maximum stress of models C and D was all distributed in the contact area with the internal fixation, and especially during backward extension, left bending, and left rotation, when it was reduced by 42.6%, 80.6%, and 49.8% of model C respectively, compared with that of model D. Conclusions:The improved laminar hook is more consistent with the Chinese anatomized structure of the lamina. Compared with the traditional lamina hook system, the improved lamina hook system can effectively reduce the displacement in all directions and range of motion of lumbar spondylolysis, therefor can significantly reduce the stress of internal fixation and vertebral body and has better biomechanical performance.

Objective:To explore classification models for radiosensitive lipid metabolites in human peripheral blood by combining lipidomics with multiple machine learning (ML) algorithms.Methods:Totally 97 peripheral blood samples were collected from 25 leukemia cases admitted to a general hospital in Beijing from March to September 2023 who were ready to undergo bone marrow transplantation, including 0 Gy blood samples before irradiation in the control group ( n=24), and 73 blood samples after irradiation at doses of 4, 8 and 12 Gy in the radiation group ( n=73), and the targeted lipidomic based on the ultra-high performance liquid chromatography-mass spectrometry (UHPLC-MS) platform method to analyze the differences of different lipids between control and radiation groups. Then, lipids responsive to radiation doses of 0-12 Gy were identified using linear regression. Finally, classification models were constructed using five ML algorithms based on the training set, followed by the validation and evaluation of these models using the validation set. Results:Compared with the control group, the differences in the concentration changes of 62 lipids in 9 classes of lipid metabolites sensitive to radiation group were statistically significant ( t=-4.91 to 4.74, P<0.05), including sphingomyelins(SMs), cholesteryl esters(CEs), ceramides(Cers), phosphatidylinositols(PIs), hexosylceramides(HexCers), lysophosphatidylcholines (LysoPCs), phosphatidylcholines (PCOs), phosphatidylethanolamines (PEs), and lysophosphatidylethanolamines (LysoPEs). Twenty lipids responsive to radiation doses of 0-12 Gy were identified, namely 11 SMs, 7 CEs, 1 Cer, and 1 PI. The five models based on ML algorithms of decision tree (DT), support vector machine (SVM), light gradient boosting machine (Light GBM), random forest (RF), and K-nearest neighbors (KNN) all exhibited high goodness of fit (F1=0.69-1.00) and high sensitivity. The evaluation and validation metrics revealed that the RF-based model yielded the optimal radiation classification discrimination (sensitivity: 1.00; accuracy: 0.72; F1 score: 0.80). Conclusions:Lipid metabolites responsive to radiation and lipids responsive to radiation dose in human samples were identified using targeted lipidomics. The RF-based model can provide new ideas for exploring models of human radiosensitive lipid metabolites.