搜索结果：Automating

To systematically evaluate the methodological quality and diagnostic performance of artificial intelligence (AI) applications, specifically machine learning (ML) and deep learning (DL), in the diagnosis of endometriosis through imaging and clinical symptomology. A systematic search was conducted across seven databases for studies published between 2015 and 2025. Inclusion criteria focused on primary research utilizing AI for endometriosis diagnosis via MRI, ultrasound, or patient-reported symptoms. Methodological quality was appraised using the QUADAS-2 tool. Study selection adhered to a double-blinded protocol to minimize selection bias. Clinical and methodological conflicts were addressed by a Professor of Radiography, while technical AI complexities were adjudicated by a Professor of Artificial Intelligence. AI models demonstrated high technical efficacy, with imaging-based algorithms achieving diagnostic accuracies up to 94.32% (MRI) and AUCs of 0.90 (Ultrasound). Symptom-based models reported accuracies reaching 95.95%, utilizing classifiers such as Random Forest and XGBoost. However, quality appraisal revealed significant clinical heterogeneity and systemic vulnerabilities. Spectrum bias was prevalent, as most models were trained on advanced-stage cohorts, limiting applicability for early-stage detection. Furthermore, symptom-based models often relied on self-reported data from social media, introducing significant selection and verification bias. While AI demonstrates high potential for automating endometriosis detection, current literature is constrained by retrospective designs and narrow patient selection. To move from experimental prototypes to clinical screening tools, future research must prioritize prospective validation in undifferentiated populations using a combination of diagnostic reference methods.

Rapid eye movement (REM) sleep behavior disorder (RBD) is a parasomnia, and its isolated form is of particular interest, as it is an early phase alpha-synucleinopathy. Machine learning (ML) and deep learning (DL) models offer potential for automated detection, prediction of phenoconversion, and phenotyping. This scoping review identified 75 studies applying ML/DL in RBD and evaluated their methodological and reporting quality using the APPRAISE-AI tool. Most studies (73.3%) focused on RBD detection and mainly used polysomnographic data for this, while 16% addressed prediction of phenoconversion, with imaging data being the most employed modality. Sample sizes were generally small (most studies including only 20-100 individuals). According to APPRAISE-AI scores, 80% of studies had moderate overall methodological and reporting quality. Common deficiencies included lack of transparency in data and code sharing (23.3%), and poor reporting of hyperparameter tuning (17.1%), bias assessment (26.9%), and error analysis (0.66%). Data leakage was observed in 32% of studies. These issues hinder clinical translation and prevent incremental progress between research groups. Without transparent reporting and shared resources, replication and model comparisons become nearly impossible. Future work should adopt open science principles and rigorous validation to advance AI-based tools in sleep medicine.

Attention-Deficit/Hyperactivity Disorder (ADHD) is a prevalent neurodevelopmental disorder that imposes significant personal and societal burdens. Traditional diagnostic approaches, which rely on behavioral assessments, are susceptible to subjectivity and variability, underscoring the need for objective and automated diagnostic tools. This study develops an ADHD-specific, biologically informed multi-stream deep learning framework for pediatric brain MRI classification, in which a Vision Transformer (ViT) and an Enhanced Convolutional Neural Network (ECNN) are integrated with Raw MRI, Phase Spectrum Transform (PST), and Quantile Histogram Equalization with Denoising (QHED) representations to capture complementary global and local neuroanatomical characteristics. The architecture leverages complementary modeling capacities by combining global contextual representations from ViT with localized discriminative features extracted by ECNN across a biologically informed multi-stream preprocessing strategy, including Raw MRI to preserve global anatomy, Phase Spectrum Transform (PST) to highlight cortical boundary irregularities, and Quantile Histogram Equalization with Denoising (QHED) to enhance subtle gray-white matter contrasts. Experimental evaluations conducted on a stratified pediatric MRI dataset demonstrated that the proposed ViT+ECNN model achieved a classification accuracy of 99.4%, precision of 99.3%, recall of 99.5%, and an F1-score of 0.99, substantially outperforming standalone ViT and ECNN configurations. These findings indicate that hybrid transformer-convolutional models can substantially enhance diagnostic accuracy and offer a promising approach for supporting early identification and intervention in ADHD.

暂无摘要（点击查看详情）

Early and accurate diagnosis of Alzheimer's Disease (AD) is crucial for effective clinical intervention. In this study, we propose a lightweight vision transformer architecture specifically designed for AD classification using 2D brain MRI slices. LICAUN-ViT incorporates three key innovations: Mono-Head Self-Attention (MOHSA) to reduce computational overhead, Uniformity Normalization (Uni-Norm) to mitigate oversmoothing and enhance feature diversity, and Context-Aware Convolution (CAC) to integrate long-range dependencies with local structural features. Evaluated on two benchmark datasets derived from the Alzheimer's Disease Neuroimaging Initiative (ADNI), our model achieves state-of-the-art performance with an accuracy of 93.03 % on axial slices and 94.15 % on sagittal slices, while maintaining relatively low floating-point operations (FLOPs) for efficient deployment. Extensive ablation studies and singular value analyses confirm the effectiveness and robustness of the proposed components. These results demonstrate that the proposed model offers a computationally efficient and promising solution for automated AD diagnosis, with strong potential for clinical integration.

We hope to update the reader on recent literature investigating the use of regional anesthesia for laparoscopic and robotic surgery. These peripheral nerve blocks are now supported by evidence in robotic or laparoscopic surgery: (1). transversus abdominis plane (TAP) block for robotic prostatectomy, (2) paravertebral block (PVB) for robotic mitral valve repair, (3) SAPB for robotic thymectomy, (4) TAP or erector spinae plane block (ESPB) for laparoscopic cholecystectomy, (5) quadratus lumborum block (QLB), ESPB, or PVB for percutaneous nephrolithotomy, (6) QLB or ESPB for partial or full nephrectomy, and (7) TAP or QLB for laparoscopic colectomy.

MRI is essential for diagnosing and monitoring neurological diseases. Conventional protocols require multiple sequences to obtain complementary contrasts, increasing scan time, cost, and tolerability. Generating multiple contrasts from a single acquisition may streamline workflow while maintaining clinical utility. To train attention-based convolutional neural networks (ACNNs) to generate clinical-quality Fluid-Attenuated-Inversion-Recovery (FLAIR), Magnetization-Prepared-Rapid-Gradient-Echo (MPRAGE), R2* maps, and derived contrasts from a single Gradient Echo Plural Contrast Imaging (GEPCI) acquisition. Retrospective. 43 MRI scans from individuals with multiple sclerosis (25/18 F/M, 49 ± 11 years-of-age). 3 T MRI, 3D GEPCI, MPRAGE, and FLAIR. Technical quality of AI-generated contrasts was evaluated against directly acquired MRI using structural similarity index (SSIM). Clinical image quality was assessed by physicians. Lesion volumes and counts were obtained using automated segmentation. One-sample one-sided Wilcoxon signed-rank test was used to establish the clinical quality of images. Agreement between native- and AI-derived lesion volume and lesion count measurements was assessed using intraclass correlation coefficients (ICC). Quantitative accuracy for R2* maps was evaluated using normalized root-mean-square error (NRMSE). AI-generated FLAIR and MPRAGE achieved mean SSIM values of 0.923 ± 0.028 and 0.935 ± 0.022, respectively. Generated R2* maps achieved a mean SSIM of 0.996 ± 0.006 and NRMSE of 0.031 ± 0.020. Physicians-assigned mean clinical quality ratings of 4.2 for GEPCI-FLAIR and 4.5 for GEPCI-MPRAGE exceeded the 4.0 clinical standard on a 1-to-5 scale. Lesion volume and count comparisons from automated segmentation showed strong agreement between AI-generated and ground-truth measurements: R2 = 0.988 and R2 = 0.933, ICC = 0.988 and ICC = 0.967, respectively. AI-GEPCI generated multiple clinically relevant MRI contrasts from a single GEPCI acquisition with high similarity to corresponding acquired images, supporting high-quality, intrinsically co-registered multi-contrast brain evaluation. 2. Stage 1. This study developed a new MRI method called AI‐GEPCI that can create many important brain images from a single scan. Normally, people with neurological diseases need several different MRI scans, which take a long time, are costly, and can be hard to tolerate. We trained artificial intelligence (AI) to turn one GEPCI scan into several clinical images (called FLAIR, MPRAGE, and R2*) that doctors routinely use. The AI‐generated images looked very similar to standard MRI images and were rated as high quality by physicians. This approach may allow faster, more comfortable, and more consistent MRI exams for brain diseases.

Mesolimbic dopamine (DA) neurons are central to cue-guided reward seeking and action sequence learning. Yet, the mechanisms by which cue-induced DA neural activity drives goal-directed or habitual sequence execution remain unknown. We designed two novel tasks to isolate the effect of sequence-delineating cues on DA-driven behavioral strategies and learning. In the lever insertion fixed-ratio 5 task (LI5), the lever insertion marked sequence initiation. In the lever retraction fixed-ratio 5 task (LR5), the lever retraction served as both sequence termination and reward-predictive cue. We found that sequence initiation and termination cues differentially affect reward expectation during action sequences, with only the termination cue contributing to greater outcome devaluation insensitivity, automaticity and behavioral chunking. Mesolimbic fiber photometry recording revealed that this habit-like behavior was associated with a rapid backpropagation in DA signals from the reward to the immediately preceding cue and with attenuated DA reward prediction error signals, which reflected greater behavioral inflexibility. Finally, in absence of external cues, brief optogenetic stimulation of VTA DA neurons at sequence termination was sufficient to drive automaticity and, to some extent, behavioral chunking. Our results highlight the critical role of cue-evoked DA signals at sequence termination in driving the development of automated, habit-like sequence execution.

The hydro-spherical contamination of water resources by per- and polyfluoroalkyl substances (PFAS) demands analytical technologies that transcend the limitations of current methods, which struggle to simultaneously achieve ultra-sensitivity, specificity, portability, and low cost. Therefore, this review advances next-generation PFAS surveillance by proposing a framework built on the synergistic convergence of molecularly engineered sulfonate-MXenes, CRISPR-Cas12a, and microfluidic automation. The manuscript critically analyzes how sulfonate-terminated Ti3C2Tx MXenes achieve picomolar affinity and rapid preconcentration of PFAS through biomimetic binding architectures. It also details the mechanism by which CRISPR-Cas12a, guided by PFAS-specific aptamers, enables single-molecule discrimination with attomolar sensitivity. Finally, this review paper demonstrates how microfluidic networks orchestrate this synergy, miniaturizing the entire assay into a portable, multiplexed platform that reduces analysis time from hours to minutes. Synergistically unifying breakthroughs in nanomaterials, synthetic biology, and lab-on-a-chip design, this work provides both a methodological blueprint for ultrasensitive PFAS sensors and a relevant roadmap for implementing proactive, decentralized water quality monitoring.

Kinetics of microbially induced carbonate precipitation (MICP) in hypersaline matrices are difficult to resolve because bulk assays collapse nucleation and growth into endpoint metrics, thereby masking interfacial heterogeneity. Here, we use an in situ microfluidic platform with time-lapse imaging and automated crystal tracking to probe how ionic strength reshapes bacteria-mineral biointerface interactions and thereby regulates CaCO3 precipitation by Staphylococcus succinus J3 across 0-100 g L-1 NaCl. Microfluidic observations reveal an ionic-strength-dependent nucleation-growth switch: low salinity is associated with the rapid appearance of abundant microcrystals and early growth saturation, whereas high salinity yields fewer observable nuclei but sustained post-nucleation growth, producing sparse yet much larger, calcite-dominated crystals. These trends were interpreted qualitatively using literature-based concepts from classical nucleation theory and double-layer interactions, suggesting that low ionic strength favors nucleation, whereas high ionic strength suppresses nucleation but promotes growth on existing surfaces. To test this mechanism, we introduced low salinity biogenic nuclei into hypersaline produced water, thereby increasing hardness removal from 91.32% to 98.67% and generating larger particles with improved separability. Overall, this work provides a biointerface-based mechanistic rationale for tuning MICP under high ionic strength and highlights microfluidics as a practical tool for resolving biomineralization kinetics in complex fluids.

暂无摘要（点击查看详情）

The global spread of drug-resistant bacteria represents a critical public health challenge, underscoring the need for rapid and reliable antibiotic susceptibility testing. Recently developed drug susceptibility testing microfluidic (DSTM) devices enable rapid assessment by analyzing antibiotic-induced bacterial morphological changes. However, interpretation of these changes is complex and subjective, necessitating automation. In this study, we developed an image analysis-based interpretation system for DSTM devices and evaluated its performance and practical utility. Using reference strains and 975 Enterobacterales isolates collected in Japan, we developed an algorithm that assigns susceptibility categories based on DSTM results. Eight antibiotics were evaluated: ampicillin (ABPC), cefazolin (CEZ), cefmetazole (CMZ), cefotaxime (CTX), cefepime (CFPM), meropenem (MEPM), levofloxacin (LVFX), and amikacin (AMK). Performance was assessed using 59 clinical isolates and 10 standard strains, with results compared with those obtained using the Clinical and Laboratory Standards Institute M07 broth microdilution method. Category agreement was highest for CEZ and LVFX (98.6%) and lowest for CMZ (82.6%). Major error rates were 0% for CFPM, LVFX, MEPM, and AMK, whereas ABPC and CTX showed the highest major error rate (7.2%). No very major errors were observed for CEZ; however, MEPM showed a relatively high very major error rate (13.0%). Although discrepancies occurred between algorithm-based and morphology-based interpretations for some antibiotics, morphological transitions generally corresponded to susceptibility profiles in most strains. Image analysis-based automation reduced subjectivity associated with conventional visual evaluation and yielded highly reproducible results, enabling quickly determination of Enterobacterales antimicrobial susceptibility within 3 h.

Accurately calculating the Ki-67 index, a critical biomarker for cellular proliferation, is pivotal in breast cancer (BC) treatment personalization. Variability in manual assessments of this index can lead to inconsistent clinical decisions, underscoring the need for precise, automated methods. This systematic review addresses this imperative by examining advanced Deep Learning (DL) techniques applied to BC histopathology images to predict the Ki-67 index. Following the PRISMA framework, an exhaustive search was conducted across multiple databases without starting-time restrictions, including studies up to February 1, 2025, with stringent inclusion criteria. The review evaluates various DL architectures, such as PathoNet, highlighting their potential to streamline and enhance the accuracy of Ki-67 index predictions. The analysis includes a comprehensive discussion of primary research databases and their accessibility, pointing to significant advancements in the automation of Ki-67 predictions. These advancements are crucial for fostering standardized treatment approaches across oncology practices. Despite notable progress, challenges persist, including the need for more extensive databases and the development of innovative DL models. Clinical adoption remains limited due to difficulties in interpretability, validation, and integration into pathology workflows. Advancing in explainability, standardized validation, and dataset expansion is crucial for facilitating clinical implementation and improving personalized BC treatment.

Breast ultrasound imaging is widely used for the early detection of breast cancer due to its accessibility and effectiveness, particularly in dense breast tissues. However, its diagnostic performance is often affected by operator dependency, speckle noise, low contrast, and variability in data quality. Although deep learning methods have shown promise in automated tumor segmentation and classification, their clinical applicability remains limited due to challenges such as small and imbalanced datasets, inconsistent annotations, and the lack of integrated learning strategies. In this work, we propose a Multi-Task U-Net framework that jointly performs lesion segmentation and tumor classification by leveraging shared feature representations. The proposed method incorporates a deterministic oversampling strategy for handling class imbalance, a prediction-refinement module to ensure consistency between segmentation and classification outputs, and an attention-guided feature learning mechanism to enhance lesion localization. Additionally, a curated version of the BUSI dataset is constructed by removing duplicate and inconsistent samples to ensure reliable evaluation. The proposed model achieves a Dice score of up to 0.81 in comparative evaluation, along with classification accuracy of up to 0.96-0.98, demonstrating improved performance over baseline methods. The consistent performance across both segmentation and classification tasks indicates good generalization capability despite dataset limitations. Finally, the proposed multi-task framework provides an effective and reliable solution for automated breast cancer detection in ultrasound images and shows strong potential for clinical application.

Kitchen waste (KW) is generated in enormous quantities, while conventional composting often performs poorly in winter because low temperatures suppress self-heating and biodegradation. Although hyperthermophilic composting (HC) can markedly promote KW degradation, the degradation mechanisms of specific KW components under cold-climate conditions remain insufficiently understood. A single-substrate enrichment strategy was established to compare the degradation of four representative fractions: cellulose, hemicellulose, lignin, and fats. After 30 days of composting, fat showed the highest degradation efficiency (49.51%), followed by cellulose (47.64%) and hemicellulose (45.29%), whereas lignin showed the lowest degradation efficiency (29.61%). Firmicutes played a dominant role across all four component-enriched systems. Tepidimicrobium was dominant in fat degradation, whereas Gracilibacillus and Ammoniibacillus were mainly associated with the degradation of cellulose, hemicellulose, and lignin. EEM-PARAFAC showed that aromatic proteins and soluble microbial by-products in all four systems were gradually transformed into humic-like substances, whereas lignin-derived dissolved organic matter showed weaker humification. Partial least-squares path modelling further indicated that hemicellulose and fat degradation were mainly driven by microbial activity, whereas cellulose and lignin degradation were more strongly influenced by environmental factors. These findings reveal component-specific microbial degradation and humification mechanisms during winter HC and provide a mechanistic basis for inoculum selection and process regulation for kitchen waste treatment in cold regions.

The timeliness of treatment for out-of-hospital cardiac arrest (OHCA) is critical for patient survival. Automated External Defibrillators (AEDs) are a proven effective intervention, yet China's rapidly developing Public Access Defibrillation (PAD) program may be accompanied by significant spatial inequities in AED distribution. This study developed a comprehensive multi-dimensional evaluation model to assess the spatial equity of AED allocation in four first-tier Chinese cities: Beijing, Shanghai, Guangzhou, and Shenzhen. The model integrated four dimensions: resource allocation (supply-demand ratio), spatial coverage (service coverage index), opportunity accessibility (accessibility index via an enhanced Gaussian two-step floating catchment area method), and spatial distribution (Gini coefficient). These dimensions were aggregated into a Comprehensive Equity Index (CEI) using the Entropy Weight Method (EWM). Leveraging high-resolution gridded population data and precise AED locations, our analysis captures fine-scale spatial variations often obscured in aggregate statistics. Furthermore, to uncover the spatially heterogeneous drivers of equity, we employed an integrated Principal Component Analysis and Geographically Weighted Regression (PCA-GWR) framework to analyze socioeconomic and urban environmental factors. The results indicate that: (1) Overall comprehensive equity was low across all cities (mean CEI&#x2009;<&#x2009;0.3). Shenzhen exhibited the highest equity (mean CEI: 0.252), followed by Beijing (0.207), with Shanghai and Guangzhou lagging. (2) A significant "core-periphery" disparity was observed in all cities, with core districts showing markedly higher equity than suburban districts, a gap particularly pronounced in Beijing and Shanghai. (3) The PCA-GWR analysis revealed pronounced spatial heterogeneity in the associations between external factors and AED equity. Degree of urbanization showed a generally positive association, which was consistently weaker in urban cores. Public service facility provision exhibited inconsistent (often negative) associations, while the wealth-population density trade-off demonstrated marked city-specific variation. This study provides a systematic, multidimensional assessment of AED allocation equity in major Chinese cities. By employing a spatially nuanced PCA-GWR framework, it reveals that equity is shaped by complex, location-specific interactions of urban development, service provision, and socioeconomic structure. The findings underscore the necessity for spatially differentiated policy interventions within China's PAD program to achieve more equitable and efficient deployment of these lifesaving resources.

Despite recent advances in medical informatics, extracting tumor information from pathology reports remains a challenge in modern cancer registry and surveillance workflows. These documents often have an unstructured format, complex medical content, and a considerably lengthy context, creating significant challenges for automated phenotypic information extraction. Although some recent language models such as BERT, GatorTron, and GPT-4 have demonstrated efficacy in medical applications, they are either constrained by sequence length limitations or cloud-based computing that violates the handling of protected health information. We introduce two oncology pathology-optimized transformer models OncoPT, based on Longformer and BigBird architectures and trained on real-world pathology reports. OncoPT efficiently processes reports up to 4,096 tokens, making it suitable for hospitals' onsite deployment with limited resources. We apply OncoPT to a common malignancy (exemplified by breast cancer) and a rare malignancy (exemplified by gastric cancer), across five key tumor phenotypes: Subsite, Histology, Grade, Stage, and Laterality. The results demonstrate that OncoPT achieves state-of-the-art weighted F-1 on a private pathology dataset and surpasses commercial chatbots (ChatGPT 4o and o1) on the public CORAL dataset (up to 30% improvement). These findings highlight the robustness of OncoPT models with the added benefit of preserving the privacy of patient health information.

To compare accuracy, precision, recall, F1 and time spent using commercial tools to identify physiotherapy trials based on title and abstract, compared with a human approach. This study compared two approaches for title and abstract screening of 10,793 newly published records. In the reference standard human approach, two reviewers independently screened records using pre-specified rules to assess relevance to physiotherapy. A third person resolved disagreements. We evaluated three LLMs (gpt-4o, gpt-4.5, gpt-4-turbo) within two commercial, web-based tools (ChatGPT and Co-pilot). Outcomes were accuracy (proportion of records that model correctly identified as relevant or irrelevant), precision (proportion of records identified as relevant that were considered as relevant by human approach), recall (the proportion of all actual relevant records that the model successfully identified), F1 (harmonic mean of precision and recall) and time spent. Exploratory analyses compared the performance of the commercial tools with local approaches, including local LLMs implementation, machine learning and natural language processing. Commercial tools showed comparable performance across all metrics (ChatGPT vs Copilot: accuracy: 83% vs 86%; precision: 44% vs 48%; recall: 88% vs 87%; F1: 59% vs 62%). The total time spent using commercial tools with a labelled dataset was equivalent to 37% of the time required for the human-only screening process. Exploratory analysis showed that the API-based implementation has comparable performance (accuracy: 82%; precision: 42%; recall: 93%; F1: 58%). Yet, LLM-based models demonstrated lower performance compared with other local, custom-adapted automation approaches such as machine learning and natural language processing. This proof-of-concept study demonstrates that commercial web-based LLMs may have sufficient accuracy to support title and abstract screening and substantially reduce the time to identify field-specific trials. However, alternative approaches, including machine learning or natural language processing, could achieve screening performance similar to or slightly higher than that of commercial tools, yet they require a series of pre-processing steps for implementation.

Body-related images (e.g., idealized thin/muscular bodies and more diverse body types) are used to study appearance-related affect, body dissatisfaction, and social comparison. Preparing such stimuli typically requires human ratings of core emotional dimensions (e.g., valence and arousal), which is labor-intensive and difficult to scale across contexts. Multimodal large language models (LLMs), such as ChatGPT, offer the potential to automate preliminary norming tasks through multimodal training and adaptability, but their validity for such rating remains unknown. Across two studies, we tested whether ChatGPT models could approximate human emotional ratings (valence and arousal) of body-related images, and whether performance depended on temperature settings and few-shot prompting. Using 80 established images of ideal and non-ideal male and female bodies, Study 1 evaluated the temperature effect by testing GPT-4o-mini and GPT-5-chat-latest across four temperature settings. Study 2 examined the few-shot effect using GPT-4o-mini, GPT-5-chat-latest, and GPT-5 with 0-5 example shots. Intraclass correlations, Pearson correlations, and error metrics were computed to examine AI-human alignment. Results revealed that across models and settings, valence ratings aligned more closely with human scores than arousal ratings. Few-shot prompting significantly improved alignment, particularly for GPT-5. However, improvements were inconsistent across both temperatures and few-shot learning. The models showed promising alignment when guided by a few examples but exhibited sex-specific biases. Findings suggest that LLMs can support preliminary emotional norming of body-related images, particularly for valence when guided by examples, but sex-specific asymmetries highlight the need to develop bias-reducing strategies to ensure responsible use in future applications.