Accuracy Of Disease Diagnosis Research Articles

Precise and automated lung cancer cell classification using deep neural network with multiscale features and model distillation

Lung diseases globally impose a significant pathological burden and mortality rate, particularly the differential diagnosis between adenocarcinoma, squamous cell carcinoma, and small cell lung carcinoma, which is paramount in determining optimal treatment strategies and improving clinical prognoses. Faced with the challenge of improving diagnostic precision and stability, this study has developed an innovative deep learning-based model. This model employs a Feature Pyramid Network (FPN) and Squeeze-and-Excitation (SE) modules combined with a Residual Network (ResNet18), to enhance the processing capabilities for complex images and conduct multi-scale analysis of each channel's importance in classifying lung cancer. Moreover, the performance of the model is further enhanced by employing knowledge distillation from larger teacher models to more compact student models. Subjected to rigorous five-fold cross-validation, our model outperforms existing models on all performance metrics, exhibiting exceptional diagnostic accuracy. Ablation studies on various model components have verified that each addition effectively improves model performance, achieving an average accuracy of 98.84% and a Matthews Correlation Coefficient (MCC) of 98.83%. Collectively, the results indicate that our model significantly improves the accuracy of disease diagnosis, providing physicians with more precise clinical decision-making support.

Ultrasound image denoising autoencoder model based on lightweight attention mechanism.

The presence of noise in medical ultrasound images significantly degrades image quality and affects the accuracy of disease diagnosis. The convolutional neural network-denoising autoencoder (CNN-DAE) model extracts feature information by stacking regularly sized kernels. This results in the loss of texture detail, the over-smoothing of the image, and a lack of generalizability for speckle noise. A lightweight attention denoise-convolutional neural network (LAD-CNN) is proposed in the present study. Two different lightweight attention blocks (i.e., the lightweight channel attention (LCA) block and the lightweight large-kernel attention (LLA) block are concatenated into the downsampling stage and the upsampling stage, respectively. A skip connection is included before the upsampling layer to alleviate the problem of gradient vanishing during backpropagation. The effectiveness of our model was evaluated using both subjective visual effects and objective evaluation metrics. With the highest peak signal-to-noise ratio (PSNR) and structural similarity (SSIM) values at all noise levels, the proposed model outperformed the other models. In the test of brachial plexus ultrasound images, the average PSNR of our model was 0.15 higher at low noise levels and 0.33 higher at high noise levels than the suboptimal model. In the test of fetal ultrasound images, the average PSNR of our model was 0.23 higher at low noise levels and 0.20 higher at high noise levels than the suboptimal model. The statistical analysis showed that the p values were less than 0.05, which indicated a statistically significant difference between our model and the other models. The results of this study suggest that the proposed LAD-CNN model is more efficient in denoising and preserving image details than both conventional denoising algorithms and existing deep-learning algorithms.

MRI super-resolution using similarity distance and multi-scale receptive field based feature fusion GAN and pre-trained slice interpolation network

Challenges arise in achieving high-resolution Magnetic Resonance Imaging (MRI) to improve disease diagnosis accuracy due to limitations in hardware, patient discomfort, long acquisition times, and high costs. While Convolutional Neural Networks (CNNs) have shown promising results in MRI super-resolution, they often don't look into the structural similarity and prior information available in consecutive MRI slices. By leveraging information from sequential slices, more robust features can be obtained, potentially leading to higher-quality MRI slices. We propose a multi-slice two-dimensional (2D) MRI super-resolution network that combines a Generative Adversarial Network (GAN) with feature fusion and a pre-trained slice interpolation network to achieve three-dimensional (3D) super-resolution. The proposed model requires consecutively acquired three low-resolution (LR) MRI slices along a specific axis, and achieves the reconstruction of the MRI slices in the remaining two axes. The network effectively enhances both in-plane and out-of-plane resolution along the sagittal axis while addressing computational and memory constraints in 3D super-resolution. The proposed generator has a in-plane and out-of-plane Attention (IOA) network that fuses both in-plane and out-plane features of MRI dynamically. In terms of out-of-plane attention, the network merges features by considering the similarity distance between features and for in-plane attention, the network employs a two-level pyramid structure with varying receptive fields to extract features at different scales, ensuring the inclusion of both global and local features. Subsequently, to achieve 3D MRI super-resolution, a pre-trained slice interpolation network is used that takes two consecutive super-resolved MRI slices to generate a new intermediate slice. To further enhance the network performance and perceptual quality, we introduce a feature up-sampling layer and a feature extraction block with Scaled Exponential Linear Unit (SeLU). Moreover, our super-resolution network incorporates VGG loss from a fine-tuned VGG-19 network to provide additional enhancement. Through experimental evaluations on the IXI dataset and BRATS dataset, using the peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM) and the number of training parameters, we demonstrate the superior performance of our method compared to the existing techniques. Also, the proposed model can be adapted or modified to achieve super-resolution for both 2D and 3D MRI data.

A recursive framework for improving the performance of multi-objective differential evolution algorithms for gene selection

Gene selection is a pivotal process in machine-learning-driven medical diagnostics, where the goal is to identify a subset of genes from microarray expression profiles that can enhance the predictive accuracy of classifiers for disease diagnosis. The two key objectives of gene selection are to reduce the dimensionality of the data and to improve the accuracy of disease diagnosis, which is typically a multi-objective optimization problem. In recent years, multi-objective evolutionary algorithms (MOEAs) have gained wide attention in feature selection research, and several related algorithms have been produced. However, most algorithms tend to get stuck in local optimality when searching for solutions from a high-dimensional space. To solve the gene selection problem effectively, this study introduces a recursive multi-objective differential evolution algorithm with elite recursive strategy (RMODE-E) and a recursive multi-objective differential evolution algorithm with Pareto front recursive strategy (RMODE-P). RMODE-E amalgamates the features selected by the top E elite individuals, RMODE-P consolidates the features selected by the Pareto front set, and the combined features then serve as the foundation for subsequent recursive rounds of searching. The proposed feature subspace combination strategy not only reduces the recursive search space but also improves the capacity to find globally optimal feature subsets. Extensive experiments were conducted to compare our proposed algorithms with eight state-of-the-art evolutionary algorithms to validate their effectiveness. Experimental results demonstrate that RMODE-P has better global search capability as it achieves better best classification accuracy, mean classification accuracy, and minimal gene subset size.

Early Detection of Parkinson’s Disease Using Machine Learning

A hallmark of Parkinson's disease is the degeneration of dopaminergic neurons in the midbrain's substantia nigra pars compacta. However, machine learning is needed to design and implement early Parkinson's disease detection. by using machine learning methods such as CNN and SVM, which can reliably identify voice signals and spiral images to identify early indications of Parkinson's disease. Using machine learning on handwriting, tremor, and gait datasets, the method addresses the shortcomings of individual analyses for a more complete diagnosis solution by investigating relationships between symptoms. This increases accuracy. When voice and spiral drawing data were combined, Parkinson's disease diagnosis accuracy showed promise, with the machine learning model successfully differentiating between unaffected patients and those who were affected. This observation suggests a viable path for precise Parkinson's disease diagnosis: an integrated method that combines machine learning techniques with data from spiral drawings and voice.

DesTrans: A medical image fusion method based on Transformer and improved DenseNet

Medical image fusion can provide doctors with more detailed data and thus improve the accuracy of disease diagnosis. In recent years, deep learning has been widely used in the field of medical image fusion. The traditional method of medical image fusion is to operate by superimposing and other methods of pixels. The introduction of deep learning methods has improved the effectiveness of medical image fusion. However, these methods still have problems such as edge blurring and information redundancy. In this paper, we propose a deep learning network model based on Transformer and an improved DenseNet network module integration that can be applied to medical images and solve the above problems. At the same time, the method can be moved to natural images. The use of Transformer and dense concatenation enhances the feature extraction capability of the method by limiting the feature loss which reduces the risk of edge blurring. We compared several representative traditional methods and more advanced deep learning methods with this method. The experimental results show that the Transformer and the improved DenseNet network module have a strong capability of feature extraction. The method yields good results both in terms of visual quality and objective image evaluation metrics.

Ketepatan Pengkodean Diagnosis Demam Berdarah Dengue Berdasarkan Hasil Pemeriksaan Klinis di Rumah Sakit Singaparna Medika Citrautama

One of the supporting units in the hospital, namely medical records, is a manual / electronic record carried out by health care providers such as diagnosis. One of the medical record activities is to code the disease. Coding is the provision of alphanumeric symbols carried out by the coder. The diagnosis coding provided must be correct in accordance with the codes listed in ICD-10, the correct coding will affect hospital health financing. The research was conducted at Medika Citra Utama Hospital. This study aims to describe the coding procedures and accuracy of disease diagnosis using a cross sectional design. The population of cases in this study were medical records of dengue haemorrhagic fever patients from January to December 2022. The number of samples was 98 medical records. The sampling technique used was simple random sampling. The results of the study of diagnosis coding procedures carried out by the coder have not been in accordance with the steps that have been determined by ICD-10 volume 2 due to the busyness of officers in carrying out other activities and there are 78.6% of medical records that are appropriate and in accordance with the results of clinical examinations, 21.4% of medical records that are not appropriate and do not match the results of clinical examinations resulting in pending claims at the hospital. The conclusion of this study is that coder officers are expected to do coding referring to ICD-Volume 2 and pay attention to clinical examination in order to get accurate coding so that hospital claims are claimed on time.

Development of a Dual Fluorescence Signal-Enhancement Immunosensor Based on Substrate Modification for Simultaneous Detection of Interleukin-6 and Procalcitonin.

High-sensitivity detection of biomarkers is of great significance to improve the accuracy of disease diagnosis and the rate of occult disease diagnosis. Using a substrate modification and two-color quantum dot (QD) nanobeads (QBs), we have developed a dual fluorescence signal-enhancement immunosensor for sensitive, simultaneous detection of interleukin 6 (IL-6) and procalcitonin (PCT) at low volumes (∼20 μL). First, the QBs compatible with QDs with different surface ligands were prepared by optimizing surfactants based on the microemulsion method. Through the use of a fluorescence-linked immunosorbent assay (FLISA), the feasibility of a dual signal-enhancement immunosensor was verified, and a 5-fold enhancement of fluorescence intensity was achieved after the directional coating of the antibodies on sulfhydryl functionalization (-SH) substrates and the preparation of QBs by using a polymer and silica double-protection method. Next, a simple polydimethylsiloxane (HS-PDMS) immunosensor with a low volume consumption was prepared. Under optimal conditions, we achieved the simultaneous detection of IL-6 and PCT with a linear range of 0.05-50 ng/mL, and the limit of detection (LOD) was 24 and 32 pg/mL, respectively. The result is comparable to two-color QBs-FLISA with a sulfhydryl microplate, even though only 20% of its volume was used. Thus, the dual fluorescence signal-enhancement HS-PDMS immunosensor offers the capability of early microvolume diagnosis of diseases, while the detection of inflammatory factors is clinically important for assisting disease diagnosis and determining disease progression.

VSG-GAN: A high-fidelity image synthesis method with semantic manipulation in retinal fundus image

In recent years, advancements in retinal image analysis, driven by machine learning and deep learning techniques, have enhanced disease detection and diagnosis through automated feature extraction. However, challenges persist, including limited data set diversity due to privacy concerns and imbalanced sample pairs, hindering effective model training. To address these issues, we introduce the vessel and style guided generative adversarial network (VSG-GAN), an innovative algorithm building upon the foundational concept of GAN. In VSG-GAN, a generator and discriminator engage in an adversarial process to produce realistic retinal images. Our approach decouples retinal image generation into distinct modules: the vascular skeleton and background style. Leveraging style transformation and GAN inversion, our proposed hierarchical variational autoencoder module generates retinal images with diverse morphological traits. In addition, the spatially adaptive denormalization module ensures consistency between input and generated images. We evaluate our model on MESSIDOR and RITE data sets using various metrics, including structural similarity index measure, inception score, Fréchet inception distance, and kernel inception distance. Our results demonstrate the superiority of VSG-GAN, outperforming existing methods across all evaluation assessments. This underscores its effectiveness in addressing data set limitations and imbalances. Our algorithm provides a novel solution to challenges in retinal image analysis by offering diverse and realistic retinal image generation. Implementing the VSG-GAN augmentation approach on downstream diabetic retinopathy classification tasks has shown enhanced disease diagnosis accuracy, further advancing the utility of machine learning in this domain.

Bladder cancer risk stratification with the Oncuria 10-plex bead-based urinalysis assay using three different Luminex xMAP instrumentation platforms

BackgroundNo single marker of bladder cancer (BC) exists in urine samples with sufficient accuracy for disease diagnosis and treatment monitoring. The multiplex Oncuria BC assay noninvasively quantifies the concentration of 10 protein analytes in voided urine samples to quickly generate a unique molecular profile with proven BC diagnostic and treatment-tracking utility. Test adoption by diagnostic and research laboratories mandates reliably reproducible assay performance across a variety of instrumentation platforms used in different laboratories.MethodsWe compared the performance of the clinically validated Oncuria BC multiplex immunoassay when data output was generated on three different analyzer systems. Voided urine samples from 36 subjects (18 with BC and 18 Controls) were reacted with Oncuria test reagents in three 96-well microtiter plates on Day 1, and consecutively evaluated on the LED/image-based MagPix, and laser/flow-based Luminex 200 and FlexMap 3D (all xMAP instruments from Luminex Corp., Austin, TX) on Day 2. The BC assay uses magnetic bead-based fluorescence technology (xMAP, Multi-analyte profiling; Luminex) to simultaneously quantify 10 protein analytes in urine specimens [i.e., angiogenin (ANG), apolipoprotein E (ApoE), carbonic anhydrase IX (CA9), CXCL8/interleukin-8 (IL-8), matrix metalloproteinase-9 (MMP-9), matrix metalloproteinase-10 (MMP-10), serpin A1/alpha-1 anti-trypsin (A1AT), serpin E1/plasminogen activator inhibitor-1 (PAI-1), CD138/syndecan-1 (SDC1), and vascular endothelial growth factor-A (VEGF-A)]. All three analyzers quantify fluorescence signals generated by the Oncuria assay.ResultsAll three platforms categorized all 10 analytes in identical samples at nearly identical concentrations, with variance across systems typically < 5%. While the most contemporary instrument, the FlexMap 3D, output higher raw fluorescence values than the two comparator systems, standard curve slopes and analyte concentrations determined in urine samples were concordant across all three units. Forty-four percent of BC samples registered ≥ 1 analyte above the highest standard concentration, i.e., A1AT (n = 7/18), IL-8 (n = 5), and/or ANG (n = 2), while only one control sample registered an analyte (A1AT) above the highest standard concentration.ConclusionMultiplex BC assays generate detailed molecular signatures useful for identifying BC, predicting treatment responsiveness, and tracking disease progression and recurrence. The similar performance of the Oncuria assay across three different analyzer systems supports test adaptation by clinical and research laboratories using existing xMAP platforms.Trial Registration: This study was registered at ClinicalTrials.gov as NCT04564781, NCT03193528, NCT03193541, and NCT03193515.

The Expression and Prognostic Significance of ICOS in NSCLC Integrated Pan-Cancer and Multi-Omics Analyses.

Background: Inducible co-stimulator (ICOS) shows great potential in the regulation of innate and adaptive immunity. However, previous studies of ICOS have often been limited to one or two levels. Methods: Using the data from the online database, the immunohistochemistry, and enzyme-linked immunosorbent assays, we investigated the role of ICOS / PD-L1 on patients with NSCLC at the mRNA, protein, and serum levels. Results: Our data revealed that unlike most solid tumors, the mRNA expression of ICOS was down-regulated in NSCLC. In addition, our data also showed that mRNA expression levels in ICOS are negatively associated with poor clinicopathologic grading but positively associated with better prognostic outcomes and higher Tregs infiltration level. Immunohistochemistry showed that ICOS correlated negatively with the T stage; while PD-L1 levels correlated positively with the N stage and FOXP3 levels. Serological biomarker analysis showed that patients with NSCLC had lower sICOS levels, which increased significantly post-surgery, and combined sICOS and sPD-L1 diagnosis improved efficacy and accuracy of disease diagnosis. Conclusion: Our findings support that ICOS suggests lower pathological staging and better prognosis. ICOS is a potential diagnostic and prognostic biomarker for NSCLC.

KNN-SVM Classifiers in Complex Diagnosis

In many applications, classification plays an indispensable role due to its powerful detection and diagnosis function. Especially in real data on disease, the detection of important factors and the diagnosis of the result usually bring huge contributions to patients. Simultaneously, complex problems in real data such as imbalanced data and missing data also lead to more challenges and difficulties. The ignorance of missing data will undermine study efficiency, and sometimes introduce substantial bias. Imbalanced data tends to be overwhelmed by the majority classes and ignoring the minority ones. The paper develops new support vector machine classifiers using k-nearest neighbors’ information (KNN-SVM), to impute missing data by calculating k-nearest neighbors’ statistical characteristic values and to interpolate some new samples between k-nearest minority class examples. As comparisons, the paper uses different kernel functions in KNN-SVM classifiers to show the different performances in disease diagnosis accuracy.

Shining a light on liver health: advancements in fluorescence-enhanced enzyme biosensors for early disease detection.

Early detection of liver diseases holds paramount importance in optimizing treatment outcomes and prognosis, thereby significantly enhancing the likelihood of recovery while mitigating the risk of progression to liver cancer. Liver diseases encompass a spectrum of conditions, each potentially manifesting distinct enzymatic profiles. Monitoring these enzymes in situ facilitates timely intervention and therapeutic management. In recent years, the field of biosensor technology has witnessed remarkable advancement, owing to strides in biomedicine and computational sciences. Biosensors have garnered widespread utility across medical and biological domains, spanning the detection of disease biomarkers, drug release tracking, ion imaging, and fluorescence imaging within living organisms. These applications have markedly enhanced imaging resolution and have the potential to refine disease diagnosis accuracy for clinicians. A pivotal aspect in the successful application of this technology lies in the construction of fluorescence probes adept at swiftly and selectively identifying target enzymes by amalgamating liver disease enzymes with fluorescence probe technology. However, research in this niche area remains relatively scarce. Building upon this foundational understanding, the present review delineates the utilization of biosensors in the early diagnosis of liver disease. Serving as a theoretical framework, this review envisages the development of high-performance biosensors tailored for the early detection of liver cancer. Furthermore, it offers insights into the potential of biosensor technology to progress and broaden its practical applications, thus contributing to the advancement of diagnostic methodologies in liver disease management.

Optimizing genetic prediction: Define-by-run DL approach in DNA sequencing

Abstract The utility of DNA sequencing in diagnosing and prognosis of diseases is vital for assessing the risk of genetic disorders, particularly for asymptomatic individuals with a genetic predisposition. Such diagnostic approaches are integral in guiding health and lifestyle decisions and preparing families with the necessary foreknowledge to anticipate potential genetic abnormalities. The present study explores implementing a define-by-run deep learning (DL) model optimized using the Tree-structured Parzen estimator algorithm to enhance the precision of genetic diagnostic tools. Unlike conventional models, the define-by-run model bolsters accuracy through dynamic adaptation to data during the learning process and iterative optimization of critical hyperparameters, such as layer count, neuron count per layer, learning rate, and batch size. Utilizing a diverse dataset comprising DNA sequences from two distinct groups: patients diagnosed with breast cancer and a control group of healthy individuals. The model showcased remarkable performance, with accuracy, precision, recall, F1-score, and area under the curve metrics reaching 0.871, 0.872, 0.871, 0.872, and 0.95, respectively, outperforming previous models. These findings underscore the significant potential of DL techniques in amplifying the accuracy of disease diagnosis and prognosis through DNA sequencing, indicating substantial advancements in personalized medicine and genetic counseling. Collectively, the findings of this investigation suggest that DL presents transformative potential in the landscape of genetic disorder diagnosis and management.

Identification of brain tumors based on digitized parameters from magnetic resonance imaging results

A methodology is proposed for identifying brain tumors by dividing the database into four parts. The results obtained from the study of sample specimens for each type of brain tumor showed a high degree of similarity in recognition. This methodology can be applied in healthcare facilities to improve the accuracy of disease diagnosis.

Integrated wearable foam modified with WS2 nanosheets@MoS2 quantum dots for oral disease diagnosis and healthcare monitoring

To develop a sensitive point-of-care platform for screening hidden caries, periodontitis and functional dyskinesia is highly desirable. Breath analysis is considered as a promising method for non-invasive clinical diagnostics. When further combined with monitoring the movement disorders of maxillofacial region, the accuracy of disease diagnosis can be greatly improved. However, the design and investigation of integrated sensing platform for multiple stimuli are currently limited. In this study, a multifunctional heterostructure combined p-type WS2 nanosheets with n-type MoS2 quantum dots (WS2/MoS2 heterostructure) is designed for realizing multi-parameters sensing via assistance of different substrate. The gas sensor based on an optimized WS2/MoS2 heterostructure (W:Mo = 300:1) exhibits an enhanced response (73.1 % to 1 ppm) and the limit of detection is 10 ppb at room temperature (25 °C) under UV assistance. The improvements in gas sensing performance can be attributed to the synergistic effect of heterostructure, sensitization of QDs and UV photoactivation. The electromechanical sensor achieved by modified WS2/MoS2 heterostructure on conductive TPU foam exhibits excellent linearity, as well as a wide sensing range of 10 %–160 % strain and 5–250 kPa pressure, respectively. In addition, it is effective in perceiving subtle or large oral and maxillofacial movements. Finally, a multifunctional wearable platform is successfully employed to differentiate the respiratory biomarkers of healthy individuals from simulated exhaled breath in patients with caries and periodontitis, and to discriminate bite force of incisors and molars in diverse populations. This wearable device offers a promising, practical, and sustainable strategy for mobile treatment platforms and personalized healthcare.

Evolutionary Optimization Algorithm for Classification of Microarray Datasets with Mayfly and Whale Survival

In the field of bioinformatics, a vast amount of biological data has been generated thanks to the digitalization of high-throughput devices at a reduced cost. Managing such large datasets has become a challenging task for identifying disease-causing genes. Microarray technology enables the simultaneous monitoring of gene expression levels, thereby improving disease diagnosis accuracy for conditions like diabetes, hepatitis, and cancer. As these complex datasets become more accessible, innovative data analytics approaches are necessary to extract meaningful knowledge. Machine learning and data mining techniques can be employed to leverage big and heterogeneous data sources, facilitating biomedical research and healthcare delivery. Data mining has emerged as a vital tool in the medical field, providing insights into illnesses and treatments and enhancing the efficiency of healthcare systems. This thesis aims to present a novel hybrid technique for feature selection using amalgamation wrappers. The proposed approach combines the Mayfly and whale survival strategies, leveraging the strengths of both algorithms. The model was evaluated using various datasets and assessment criteria, including precision, accuracy, recall, F1-score, and specificity. The simulation results demonstrated that the proposed integrated optimization model exhibits improved classification performance with 12% higher accuracy in disease diagnosis.

Performance of Ultrasonography in the Diagnosis of Acute Colonic Diverticulitis.

Computed tomography is regarded as the reference-standard imaging modality for the assessment of acute left-sided colonic diverticulitis (ALCD). However, its utility may be impaired by cost issues, limited availability, radiation exposure, and contrast-related adverse effects. Ultrasonography is increasingly advocated as an alternative technique for evaluating ALCD, although there is variation regarding its accuracy in disease diagnosis and staging and in determining alternative diagnoses. The aim of this study was to assess the performance of ultrasonography in diagnosing ALCD, differentiating complicated from non-complicated disease and defining alternative diseases related to left lower quadrant pain. Within a 2-year period, all consecutive adult patients with clinically suspected ALCD and available abdominal computed tomography were prospectively evaluated and planned to undergo an abdominal ultrasonographic examination, tailored to the assessment of left lower quadrant. Computed tomography (CT) was regarded as the reference standard. A total of 132 patients (60 males, 72 females; mean age: 61.3 ± 11 years) were included. The sensitivity, specificity, and area under curve of ultrasonography for diagnosing ALCD were 88.6, 84.9, and 86.8%, with positive and negative predictive values of 89.7 and 83.3%, respectively. The method had sensitivity, specificity, and area under curve of 77.8, 100, and 88.9%, respectively, for defining complicated disease. The area under the curve for the identification of alternative diseases in patients with left lower quadrant pain was 90.9%. Ultrasonography has high diagnostic accuracy for diagnosing ALCD, differentiating complicated from non-complicated disease and establishing alternative diagnoses related to left lower quadrant pain. A low threshold to get a CT should be maintained as not to miss cases that may mimic ALCD.

Extracellular Vesicles: Techniques and Biomedical Applications Related to Single Vesicle Analysis.

Extracellular vesicles (EVs) are extensively dispersed lipid bilayer membrane vesicles involved in the delivery and transportation of molecular payloads to certain cell types to facilitate intercellular interactions. Their significant roles in physiological and pathological processes make EVs outstanding biomarkers for disease diagnosis and treatment monitoring as well as ideal candidates for drug delivery. Nevertheless, differences in the biogenesis processes among EV subpopulations have led to a diversity of biophysical characteristics and molecular cargos. Additionally, the prevalent heterogeneity of EVs has been found to substantially hamper the sensitivity and accuracy of disease diagnosis and therapeutic monitoring, thus impeding the advancement of clinical applications. In recent years, the evolution of single EV (SEV) analysis has enabled an in-depth comprehension of the physical properties, molecular composition, and biological roles of EVs at the individual vesicle level. This review examines the sample acquisition tactics prior to SEV analysis, i.e., EV isolation techniques, and outlines the current state-of-the-art label-free and label-based technologies for SEV identification. Furthermore, the challenges and prospects of biomedical applications based on SEV analysis are systematically discussed.

A Soft-Reference Breast Ultrasound Image Quality Assessment Method That Considers the Local Lesion Area.

The quality of breast ultrasound images has a significant impact on the accuracy of disease diagnosis. Existing image quality assessment (IQA) methods usually use pixel-level feature statistical methods or end-to-end deep learning methods, which focus on the global image quality but ignore the image quality of the lesion region. However, in clinical practice, doctors' evaluation of ultrasound image quality relies more on the local area of the lesion, which determines the diagnostic value of ultrasound images. In this study, a global-local integrated IQA framework for breast ultrasound images was proposed to learn doctors' clinical evaluation standards. In this study, 1285 breast ultrasound images were collected and scored by experienced doctors. After being classified as either images with lesions or images without lesions, they were evaluated using soft-reference IQA or bilinear CNN IQA, respectively. Experiments showed that for ultrasound images with lesions, our proposed soft-reference IQA achieved PLCC 0.8418 with doctors' annotation, while the existing end-to-end deep learning method that did not consider the local lesion features only achieved PLCC 0.6606. Due to the accuracy improvement for the images with lesions, our proposed global-local integrated IQA framework had better performance in the IQA task than the existing end-to-end deep learning method, with PLCC improving from 0.8306 to 0.8851.