An improved lightweight object detection algorithm for YOLOv5.

Object detection based on deep learning has made great progress in the past decade and has been widely used in various fields of daily life. Model lightweighting is the core of deploying target detection models on mobile or edge devices. Lightweight models have fewer parameters and lower computational costs, but are often accompanied by lower detection accuracy. Based on YOLOv5s, this article proposes an improved lightweight target detection model, which can achieve higher detection accuracy with smaller parameters. Firstly, utilizing the lightweight feature of the Ghost module, we integrated it into the C3 structure and replaced some of the C3 modules after the upsample layer on the neck network, thereby reducing the number of model parameters and expediting the model's inference process. Secondly, the coordinate attention (CA) mechanism was added to the neck to enhance the model's ability to pay attention to relevant information and improved detection accuracy. Finally, a more efficient Simplified Spatial Pyramid Pooling-Fast (SimSPPF) module was designed to enhance the stability of the model and shorten the training time of the model. In order to verify the effectiveness of the improved model, experiments were conducted using three datasets with different features. Experimental results show that the number of parameters of our model is significantly reduced by 28% compared with the original model, and mean average precision (mAP) is increased by 3.1%, 1.1% and 1.8% respectively. The model also performs better in terms of accuracy compared to existing lightweight state-of-the-art models. On three datasets with different features, mAP of the proposed model achieved 87.2%, 77.8% and 92.3%, which is better than YOLOv7tiny (81.4%, 77.7%, 90.3%), YOLOv8n (84.7%, 77.7%, 90.6%) and other advanced models. When achieving the decreased number of parameters, the improved model can successfully increase mAP, providing great reference for deploying the model on mobile or edge devices.

Improved deep learning image classification algorithm based on Swin Transformer V2.

While convolutional operation effectively extracts local features, their limited receptive fields make it challenging to capture global dependencies. Transformer, on the other hand, excels at global modeling and effectively captures global dependencies. However, the self-attention mechanism used in Transformers lacks a local mechanism for information exchange within specific regions. This article attempts to leverage the strengths of both Transformers and convolutional neural networks (CNNs) to enhance the Swin Transformer V2 model. By incorporating both convolutional operation and self-attention mechanism, the enhanced model combines the local information-capturing capability of CNNs and the long-range dependency-capturing ability of Transformers. The improved model enhances the extraction of local information through the introduction of the Swin Transformer Stem, inverted residual feed-forward network, and Dual-Branch Downsampling structure. Subsequently, it models global dependencies using the improved self-attention mechanism. Additionally, downsampling is applied to the attention mechanism's Q and K to reduce computational and memory overhead. Under identical training conditions, the proposed method significantly improves classification accuracy on multiple image classification datasets, showcasing more robust generalization capabilities.

Fabrication and Separation of EGaIn Microparticles from Human Blood Based on Dielectrophoresis Force and a W-Type Electrode.

Harmful particles such as heavy metal particles in the human body can cause many problems such as kidney stones, gallstones, and cerebrovascular diseases. Therefore, it is critical to separate them from the blood and perform a systematic analysis as early as possible. Here, we apply eutectic gallium indium (EGaIn) microparticles as a model to study the separation of particles from blood, thanks to their properties of low toxicity, excellent degradability, and negligible vapor pressure. In particular, the dielectrophoresis (DEP) separation method is employed to separate EGaIn of different sizes and characteristics in blood. First, the screen-printing method is used to create EGaIn microparticles with diameters of 15, 23, 18, and 11 µm. According to the lifetime test, these microparticles can last more than 1 month, as evidenced by their surface oxidation characteristics. Moreover, a DEP platform with W-type electrodes is developed to sort EGaIn particles from whole human blood. The results show that a sorting efficiency of 95% can be attained, which is similar to the separation efficiency of 98% achieved by finite element analysis (FEA) using COMSOL software based on the orthogonal array experiment method. The proposed study successfully validates the use of the DEP method to separate particles from human blood, providing insights into heavy metal particle separating, drug screening, and cell sorting and potentially broadening the applications in environmental analysis, food engineering, and bioengineering.

Microparticle Manipulation Based on the Bulk Acoustic Wave Combined with the Liquid Crystal Backflow Effect Driving in 2D/3D Platforms.

Microparticle manipulation has been widely used in clinical diagnosis, cell separation, and biochemical analysis via optics, electronics, magnetics, or acoustic wave driving. Among them, the bulk acoustic wave (BAW) driving method has been increasingly adopted because of non-contact, easy control, and precise manipulation. However, its low manipulation efficiency limits the usage of the BAW driving in high viscosity solutions. Therefore, in order to obtain larger driving force and more flexible manipulation of microparticles, both two-dimensional (2D) and three-dimensional (3D) platforms based on the BAW and liquid crystal backflow effect (LCBE) driving in liquid crystal (LC) solutions are proposed. The driving forces applied on the microparticles allow for the change of microparticle moving direction, which is also ascertained through theory analysis combined with various driving methods. Specifically, the maximum moving speed (68.78 µm/s) of the polystyrene particles is obtained by the BAW (13 Vpp) combined with LCBE (30 V) at a low frequency of 7.2 kHz in the 2D platform. Precise position manipulation in 3D is also fulfilled through a programmable logic control model using polystyrene particles as a demonstration. In addition, red blood cells mixed with LC solutions are arranged in a line or gathered in the pressure nodes of the BAW forces along with sinusoid signals generated by various transducer combinations. Therefore, it is approved that the LC solution that induces the LCBE force could increase the microparticle manipulation efficiency in both 2D and 3D platforms. The proposed method will open up new avenues in particle manipulation and benefit a variety of applications in cell separation, drug synthesis, analytical chemistry, and others.

A plantar wearable pressure sensor based on hybrid lead zirconate-titanate/microfibrillated cellulose piezoelectric composite films for human health monitoring.

Flexible and wearable electronic sensors hold great promise for improving the quality of life, especially in the field of healthcare monitoring, owing to their low cost, flexibility, high electromechanical coupling performance, high sensitivity, and biocompatibility. To achieve high piezoelectric performance similar to that of rigid materials while satisfying the flexible requirements for wearable sensors, we propose novel hybrid films based on lead zirconate titanate powder and microfibrillated cellulose (PZT/MFC) for plantar pressure measurements. The flexible films made using the polarization process are tested. It was found that the maximum piezoelectric coefficient was 31 pC N-1 and the maximum tensile force of the flexible films was 26 N. A wide range of bending angles between 15° and 180° proves the flexibility capability of the films. In addition, the charge density shows a proportional relation with the applied mechanical force, and it could sense stress of 1 kPa. Finally, plantar pressure sensors are arranged and packaged with a film array followed by connection with the detection module. Then, the pressure curves of each point on the plantar are obtained. Through analysis of the curve, several parameters of human body motions that are important in the rehabilitation of diabetic patients and the detection of sports injury can be performed, including stride frequency, length and speed. Overall, the proposed PZT/MFC wearable plantar pressure sensor has broad application prospects in the field of sports injury detection and medical rehabilitation training.

Influence of different postures under vertical impact load on thoracolumbar burst fracture.

Clinical studies have extensively shown that burst fractures can cause severe and long-term neurological deficits. However, the mechanism of burst fracture is not clear, and the influence of different spinal postures on burst fracture is still unidentified. The study aimed at investigating the influence of different postures under vertical impact load on thoracolumbar burst fracture. A detailed nonlinear finite element model of T12-L2 segment was developed to investigate these problems. In this work, a rigid ball was used to vertically impact the finite element spinal segment, which emulated the process of burst fracture as in experimental condition. During the process, amounting to 9 different postures (normal, flexion, extension, right/left lateral bending of 8°, right/left axial rotation of 4° and 8°) were studied. Totally five failure modes were observed. Six different parameters, including vertebral height, vertebral bulge, interpedicular widening, vertebral kyphotic angle, posterior vertebral body angle, and joint facet contact force, were observed to evaluate the corresponding severity of burst fracture. Burst fracture in extension was the severest, and the loss of vertebral height in flexion was the most. The different postures in these simulations changed the morphology of intervertebral disc and facet joints force, resulting in different types of fracture.

Finite element modeling and static/dynamic validation of thoracolumbar-pelvic segment.

Finite element method is an efficient tool to investigate the biomechanics of human spine. The key to finite element method is to reconstruct a complete and accurate finite element model. In this study, a three-dimensional finite element model of thoracolumbar structure including complete pelvis (T12-pelvis) was built using computed tomography technology. The modeling process has been explained in detailed. During the process of validation, the model was assigned with non-linear material property for static or dynamic analyses. In static analysis, the vertebral geometry parameters of T12-L5, the axial displacement, the posterior disc bulge and the intradiscal pressure of intervertebral disc, range of motion under six loading cases and facet joint forces were obtained and compared with the experimental data. In dynamic analysis, motion segments were loaded with sinusoidal displacement at 1 Hz in the anterior-posterior and axial directions to verify the reaction force. The first-order resonant frequencies in the vertical direction from one motion segment and two motion segments to the entire model were obtained. The study provides a detailed and accurate method of validation to verify the finite element model of thoracolumbar spine.

A biomechanical investigation of thoracolumbar burst fracture under vertical impact loads using finite element method.

BACKGROUND: A sudden vertical impact load on spine can cause spinal burst fracture, especially in the thoracolumbar junction region. This study aimed at investigating the mechanism of spinal burst fracture under different energy vertical impact loads, producing the failure risk region to understand burst fracture, reducing nervous system damage and guiding clinical treatment. METHODS: A nonlinear finite element model of T12-L1 motion segment was created to analyze the response of the vertical impact load. A rigid ball was used to impact the segment vertically to simulate the vertical impact load in practice. There were three different mass balls to represent the different loads: low energy, intermediate energy and high energy (respectively 13 J, 30 J and 56 J). The results of impact force, vertical displacement, stress, intradiscal pressure and contact force were obtained during the process. FINDINGS: At low energy condition, the rigid ball rebounded rapidly. At intermediate energy condition, fractures were initiated in vertebral foramen and left rear regions on the superior cortical bone near the superior endplate of L1. At high energy condition, burst fracture occurred and a part of L1 was isolated from the model. INTERPRETATION: The fracture occurred on the L1 segment only at the intermediate energy and high energy. The strength of vertebral body under low and intermediate energy was enough to support the impact. The burst fracture pattern at high energy was also observed in clinical practice. The findings may explain the mechanism of burst fracture.

Prognostic significance of claudin-1 and cyclin B1 protein expression in patients with hypopharyngeal squamous cell carcinoma.

Claudin-l and cyclin B1 are abnormally expressed in certain malignancies, but their expression in hypopharyngeal squamous cell carcinoma (HSCC) has not been reported thus far. Studying the expression levels of claudin-1 and cylin B1 in HSCC tissues and their association with clinical stage, pathological grade and prognosis in patients with HSCC may provide a theoretical basis and guide future research on HSCC targeted therapy. The protein expression levels of the above two biomarkers was immunohistochemically detected in 97 HSCC cases and 90 matched adjacent tissue samples. The correlation between the expression levels of claudin-1 and cylin B1 and the patients' clinical parameters was analyzed via Pearson's χ2 test, while survival analysis was performed using a log-rank test. The results of the current study revealed that claudin-1 and cyclin B1 were highly expressed in HSCC tissues, and the expression of claudin-1 was associated with tumor differentiation degree and lymph node metastasis, while cyclin B1 expression was associated with tumor differentiation degree. Furthermore, Kaplan-Meier analysis revealed that claudin-1 expression correlated with survival (P=0.003), and the expression levels of claudin-1 and cyclin B1 were observed to be positively correlated, in patients with HSCC. Cyclin B1 and claudin-1 exhibited an elevated expression in HSCC specimens, thus suggesting their use as tumor markers. Therefore, the joint detection of claudin-1 and cyclin B1 may aid to guide cancer therapy and to determine prognosis in HSCC. Furthermore, claudin-1 may be used as an HSCC-monitoring index, and may serve as a therapeutic target.

Study on the interaction between HIV reverse transcriptase and its non-nucleoside inhibitor nevirapine by capillary electrophoresis.

The HIV reverse transcriptase (RT) is an important antiviral target for the chemotherapy of AIDS because of its key role in virus replication. Nevirapine is a first generation of non-nucleoside reverse transcriptase inhibitors (NNRTIs), which is usually used for the therapy of AIDS. In this study, a high-performance analytical method based on capillary electrophoresis (CE) to investigate interactions between HIV RT and nevirapine was developed. Samples containing HIV RT and nevirapine at various ratios were incubated at 37 degrees C for 45 min and then separated by CE with Tris-acetate buffer at pH 7.3 containing 0.15% SDS. Both qualitative and quantitative characterizations of the binding were determined by CE for the first time. The binding constants of the interactions between HIV RT and nevirapine were calculated as (3.25+/-0.16)x10(4) and (1.25+/-0.07)x10(2) M(-1) by Scatchard analysis. HIV RT and nevirapine have two binding sites. The presented methodology should be generally applicable to study the interactions between HIV RT and nevirapine quantitatively and qualitatively.

Study on steady-state kinetics of nucleotide analogues incorporation by non-gel CE.

A method for studying steady-state kinetics of nucleotide analogues incorporation into DNA strand by non-gel CE (NGCE) with LIF was developed. Nucleoside analogue is a kind of antiviral agent used to inhibit viral replication in infected cells, especially HIV. Steady-state parameter K(m) for nucleotide analogues is determined to imply the relationship between nucleoside analogues and the enzyme in the DNA chain elongation and predict the antiviral efficacy in vivo. Samples were prepared by single nucleotide incorporation assays catalyzed by Taq DNA polymerase at 58 degrees C and HIV reverse transcriptase (RT) at 37 degrees C, and then were separated using NGCE under optimized conditions: 25 mmol/L Tris-boric-EDTA buffer (pH 8.0) with 7 mmol/L urea in the presence of 20% w/v PEG 35000 at 30 degrees C and -20 kV. K(m(dTTP)), K(m(d4TTP)) and K(m(AZTTP)) were measured by NGCE for the first time and their values for Taq DNA polymerase were 0.29+/-0.04, 32.1+/-3.3 and 74.5+/-6.6 micromol/L, respectively. For HIV RT, the values were 0.15+/-0.05, 0.31+/-0.03 and 0.17+/-0.03 micromol/L, respectively. The trend of data for HIV RT measured by NGCE was consistent with that measured by PAGE. The reported method by NGCE for the K(m) determination was powerful, sensitive and fast, and required less amounts of reagents compared with PAGE. It be employed as a reliable alternative method and further applied in other relative studies of nucleoside analogue substrates and DNA polymerases or RTs.

[The application of mastoid obliteration with homograft tooth in open method tympanoplasty].

OBJECTIVE: To investigate the effect of open method tympanoplasty with mastoid obliteration with homograft tooth and reconstruction of the attic wall. METHOD: Fifty-two cases with cholesteatoma or skeletal ulcer otitis media were performed the open method tympanoplasty with mastoid obliteration with homograft tooth and reconstruction of the attic wall after radical mastoidectomy. RESULT: External auditory canals of 48 patients were normal in appearance. The transplanted membrane in 46 ears was survived. The dry-ears rate was 92.31% and the average time waited till dry-ears was 17.56 +/- 4.16 days. The air-conducting hearing levels of average language frequency have been enhanced over 15 dB HL in 41 ears. An air-bone gap of less than 20 dB HL was achieved in 31 ears. CONCLUSION: The open method tympanoplasty with mastoid obliteration with homograft tooth and reconstruction of the attic wall can recovery the anatomic structure and physiological function of external auditory canal and middle ear better. The long-term effect on improving hearing was significant.