搜索结果：automated

The aim of this study was to examine whether a novel semi-automated dataset based on electronic health record documentation can be used for surveillance of central venous catheter-related mechanical complications (failed catheterisation, bleeding, cardiac arrhythmia, pneumothorax and nerve injury) within 24 hours of catheterisation. The semi-automated dataset comprised a fully automated extraction of clinical documentation from the electronic health record supplemented with a minor manual review aimed at identifying pneumothoraces as these rarely are diagnosed at the time of insertion but rather after a postprocedural chest X-ray. To assess surveillance performance, we compared the number of mechanical complications between the semi-automated and manually evaluated datasets for the same cohort and study period, focusing on agreement in aggregate counts. Comparisons were made at the group level only, without enforcing insertion-by-insertion matching. A total of 12 667 insertions were included. Minor mechanical complications occurred in 615 (4.9%) of the insertions in the semi-automated dataset and in 645 (5.1%) of the insertions in the manually validated dataset. Major mechanical complications occurred in 44 (0.35%) of the insertions in the semi-automated dataset compared to 48 (0.38%) in the manually validated dataset. A semi-automated dataset based on electronic health record documentation provides sufficiently accurate surveillance of central catheterisation related mechanical complications at the group level. Despite minor discrepancies, the semi-automated method enhances efficiency, scalability, and supports continuous real-time quality assurance. The potential underestimation of complication rates is offset by the possibility of robust real-time quality assurance in registries and a substantial analytical power in scientific studies.

Grass seed crops are susceptible to yellow dwarf viruses transmitted by aphids. The Willamette Valley in Oregon USA is the leading producer of cool-season grass seed crops globally, and industry reports have attributed seed yield loss and shortened stand longevity to aphid-transmitted yellow dwarf viruses. Genetic resources are needed for effective and sustainable management of this pest, specifically the Rhopalosiphum padi-PAV pathosystem, in grass seed production to reduce foliar insecticide applications and maintain optimum seed yield potential. High-throughput phenotyping methods are needed to screen grass seed cultivars to identify resistant traits for traditional breeding programs. An automated video tracking procedure was optimized to evaluate host plant resistance in cool-season grass seed crops to R. padi-PAV with live plants and viruliferous and non-viruliferous aphid populations. Feeding behavior recorded with automated video tracking was strongly correlated with 'ground-truthed' observations by human observers. Partial resistance (antixenosis and antibiosis) and tolerance traits were detected in select perennial ryegrass and tall fescue cultivars evaluated with traditional phenotyping methods in a greenhouse setting and with high-throughput phenotyping using automated video tracking in the laboratory. Across grass cultivars, nonviruliferous aphids had greater fitness and preference for non-infected grass plants compared to viruliferous aphids. Automated video tracking can be used as a high-throughput phenotyping method for continued evaluation of host plant resistance in grasses grown for seed production, as well as to identify resistant genotypes in other grass crops susceptible to aphid-YDV virus-vector systems.

Wrong-side diagnostic imaging order errors are preventable errors that can delay diagnosis and cause patient harm yet remain underdetected due to limitations in existing reporting systems. To develop and validate an automated electronic health record (EHR)-based method for detecting potential wrong-side diagnostic imaging order errors using an adapted Retract-and-Reorder (RAR) approach and to identify associated risk factors. Retrospective cohort study. Six-facility health system comprising inpatient, outpatient and emergency room sites. We screened 355 000 imaging orders with side specified, placed during 2021 across our healthcare system. We adapted the RAR methodology, originally developed to detect near-miss medication errors, by extending detection windows to 24 hours and identifying any orders switching from one side to the contralateral side, accounting for multiprovider workflows inherent in imaging. We validated the method through chart review of 100 randomly selected RAR events, then applied the query across all imaging orders. Multivariate logistic regression was used to identify risk factors associated with RAR events. We identified 1667 RAR events (4.70 per 1000 orders). Validation yielded a positive predictive value of 87% (95% CI 79.0% to 92.2%), estimating 4.09 confirmed wrong-side errors per 1000 orders. The odds of an RAR event were significantly higher in outpatient settings compared with inpatient settings (OR 4.53; 95% CI 3.80 to 5.42) and among administrative staff compared with attending physicians (OR 2.08; 95% CI 1.73 to 2.49). CT scans showed 79% higher odds of an RAR event compared with X-rays (OR 1.79; 95% CI 1.34 to 2.39). This validated approach offers a scalable solution for automated detection of potential wrong-side diagnostic imaging order errors. The methodology leverages commonly available EHR data to support continuous surveillance and intervention evaluation for improved diagnostic safety.

Fruits are valued for their nutritional benefits, providing essential carbohydrates, vitamins, and dietary fibre. However, assessing fruit ripeness remains challenging, as it is governed by complex physiological processes and environmental factors that are not always reflected in external appearance. This difficulty is exemplified by Nam Dok Mai Si Tong (NDMST) mangoes, which retain a largely uniform yellow skin during ripening, making visual maturity grading unreliable. To address this limitation, we propose AFMG-DLFF (Automated Fruit Maturity Grading using Deep Learning with Feature Fusion), a multimodal deep learning framework that integrates external RGB image features with internal biochemical attributes. Visual features are extracted using DenseNet201, Inception-ResNetV2, and EfficientNetV2, while intrinsic traits such as total soluble solids, titratable acidity, and BrimA are encoded via a dedicated neural network. These complementary feature spaces are fused and optimised using Glowworm Swarm Optimization (GSO) for hyperparameter tuning. The model is trained with an 80:20 train-test split and early-stopping-based validation, achieving a classification accuracy of 97.86% for NDMST mango maturity stages. The results demonstrate that AFMG-DLFF achieves strong performance relative to the evaluated deep-learning baselines and remains competitive with selected literature-reported fruit ripeness classification methods, while relying only on accessible RGB imaging and standard biochemical measurements. This highlights its potential as a practical, non-destructive, and cost-effective solution for automated fruit maturity grading in real-world supply chains.

Pen-and-paper cognitive assessment tools to detect dementia have higher rates of misdiagnosis amongst minority populations, especially those who complete the assessment in their second language. CognoSpeak is an automated cognitive assessment tool that uses machine learning to detect early signs of cognitive impairment from speech. We assess the utility of different pen-and-paper cognitive assessments and CognoSpeak in ethnic minority populations living in the UK. Research champions from four community centres across Yorkshire recruited cognitively healthy adults from their community: 51 Somali, 50 South Asian (South Yorkshire), 50 Chinese, and 49 South Asian (West Yorkshire). Participants completed the Montreal Cognitive Assessment (MoCA), Rowland Universal Dementia Assessment Scale (RUDAS), Multicultural Cognitive Examination (MCE), and CognoSpeak. A high percentage (47.5%) of participants recruited from ethnic minority community centres were misclassified as cognitively impaired with the MoCA, compared to just 3.4% in the RUDAS and 2% in the MCE. An acoustic-based SVM model analysis of responses to CognoSpeak achieved 83% accuracy in the ethnic minority cohort, at a similar rate to monolinguals (86%). Linguistic and text-based models showed higher levels of bias. Cognitive assessments, such as the MCE and RUDAS, may be superior to the MoCA in multilingual ethnic minority populations. Automated AI tools like CognoSpeak show promise in reducing healthcare burden in detecting dementia; however, additional work is required on managing implicit bias in any AI model before they could be clinically implemented.

Automated tools quantifying multiple sclerosis (MS) imaging biomarkers often require non-routine MRI sequences and lack MS reference data. We developed an open-source quantitative report (QReport) that integrates validated 3D T2-FLAIR quantification methods with multi-centre MS and healthy reference models, and presents outputs in a structured graphical report to support contextualised interpretation of clinically relevant biomarkers. 2516 cross-sectional 3D T2-FLAIR scans from people with MS (pwMS) and healthy controls (HC) were retrospectively collected from 14 centres within Magnetic Resonance Imaging in MS (MAGNIMS) and affiliated sites, as well as open-source datasets. Validated T2-FLAIR-based algorithms quantified total and regional lesion count (LC), lesion volume (LV), brain volume (BV), and brain age gap estimation (BrainAGE). Distributions in pwMS and HC were estimated using quantile regression. A QReport was designed to present biomarkers and reference models in graphical formats. Four neuroradiologists assessed agreement between QReport outputs and their visual assessment, and evaluated its usefulness, in 22 cases. We analysed scans from 1723 HC (age, mean ± SD: 54.5 ± 16.0; range: 18-75; F/M: 949/774) and 793 pwMS (age, mean ± SD: 43.0 ± 11.1; range: 18-75; F/M: 538/255) across 14 centres. The QReport presents single-subject measures contextualised against the 95th, 50th, and 5th percentile distributions in pwMS and HC, and includes BrainAGE. In 94% of evaluations, QReport outputs demonstrated Moderate-to-Complete agreement with visual assessment and were rated as useful in 82%. We developed an MS QReport requiring only 3D T2-FLAIR, integrating validated quantification algorithms and incorporating BrainAGE within a clinically interpretable framework.

Modern high-density neural recordings demand spike sorting algorithms that can handle diverse probe geometries and complex, neuron-specific drift, yet existing methods often rely on rigid geometric assumptions and one-dimensional drift models. Here, we introduce KIASORT (Knowledge-Integrated Automated Spike Sorting), a geometry-free approach for per-neuron drift tracking. KIASORT builds channel-specific sorting models from a hybrid linear-nonlinear sample-sorting stage, using representative template banks or supervised classifiers. These channel-specific models then sort spikes by independently tracking each neuron, unconstrained by probe layout. Biophysical simulations showed that even sub-micron probe displacements induce neuron-specific waveform distortions that standard drift models cannot correct. In ground-truth benchmarks with heterogeneous, neuron-specific drift, KIASORT outperformed Kilosort4 in recovering high-quality units, while maintaining real-time performance on standard CPUs. Its robustness was further illustrated on both primate and mouse data. KIASORT combines automated sorting with manual curation in a unified graphical interface, offering a complete and user-friendly spike sorting platform. The software is freely available at https://kiasort.com.Significance Statement Accurate spike sorting remains a fundamental challenge in systems neuroscience, particularly as recording technologies advance toward simultaneous monitoring of thousands of neurons and the next generations of recording probes. Current methods often rely on rigid assumptions about probe geometry and uniform drift patterns for different neurons which often fail in real-world recordings. We introduce KIASORT with capability to track neurons in a geometry-free framework, as a fundamentally new approach that addresses these critical limitations with many use cases which are not supported by other existing methods.

High-performance cell-free protein synthesis has transformative potential for synthetic biology, yet the prohibitive costs of Protein synthesis Using Recombinant Elements kits and the labor intensity of in-house preparation have restricted accessibility and scalability. We developed Purified components Optimized for Flexible protein expression using in vitro-produced translation factors (i-POPFLEX), a modular cell-free protein synthesis system in which 34 translation factors and a split T7 RNA polymerase are individually synthesized in vitro and assembled using automated liquid handling. This workflow minimizes manual input and supports parallelized production, generating complete, ready-to-use systems within 2 days. i-POPFLEX achieves up to 8.4-fold higher protein yields and a 96% cost reduction (27-fold lower cost) compared with commercial kits. Its flexible architecture also enables selective component inclusion for genetic code reprogramming and site-specific incorporation of noncanonical amino acids. By coupling modular design with automation, i-POPFLEX provides an accessible, customizable, and economically viable platform for next-generation biomanufacturing workflows.

Applications of anaerobic ammonium oxidation (anammox) processes for treating municipal wastewater have promising benefits in reducing carbon emissions and have been heavily investigated recently. However, the practical operation of mainstream anammox processes still suffers from unclear external regulation strategies. Therefore, this study utilized the H2O automated machine learning (AutoML) algorithm and interpretable analysis to capture the internal relationships existing in the big dataset collected from anammox-based studies for treating municipal wastewater. The eXtreme Gradient Boosting and gradient boosting machine models automatically generated by the AutoML algorithm provided the most accurate prediction results (R2 = 0.814-0.993). Moreover, the optimal models showed good generalization ability (R2 = 0.725-0.945) for unseen data collected from this study. The appropriate ranges of operation conditions and influent characteristics response to the high removal efficiency of nitrogen pollutants and high nitrogen removal rate through the anammox reaction pathway were revealed by the one-dimensional, and two-dimensional partial dependence plots. This work would advance the understanding of how to improve the practical operation of mainstream anammox-based nitrogen removal processes for treating municipal wastewater.

Carotid atherosclerotic plaque burden is a well-established biomarker of cerebrovascular and cardiovascular risk, yet its quantitative assessment from ultrasound imaging remains highly operator-dependent and poorly standardized. This study proposes a deep multi-task attention-based learning framework for automated, reproducible quantification of carotid plaque area from routine B-mode ultrasound images. The model integrates a ResNet-50 backbone with convolutional block attention modules (CBAM) to jointly learn plaque presence and normalized plaque area within a unified architecture. A total of 1,100 expert-annotated carotid ultrasound images were used, with 200 images reserved for independent validation. On the validation set, the proposed approach achieved a mean absolute error (MAE) of 0.001324 (corresponding to 2.44 mm&#xb2; clinically), a root mean square error (RMSE) of 0.001891, and a Pearson correlation coefficient of 0.712 between predicted and reference plaque areas. The model demonstrated 65.2% improvement in MAE over U-Net segmentation pipelines and 37.0% improvement over ResNet-50 regression-only approaches. Bland-Altman analysis revealed minimal bias (mean difference: 0.0003) with narrow limits of agreement (-0.0034 to 0.0039), while intraclass correlation coefficient (ICC) reached 0.85, indicating excellent measurement reliability. Clinical assessment showed that 85% of measurements fell within clinically acceptable error thresholds (<4.0 mm&#xb2;), enabling detection of plaque progression exceeding the 5.0 mm&#xb2; minimal detectable change with 95% confidence. These findings demonstrate that attention-guided deep multi-task learning enables accurate, reproducible quantification of carotid plaque burden, directly supporting the reliable detection of clinically significant plaque progression and advancing clinical translation through agreement-centered validation.

Recent advances in connectomics have been led by high-resolution reconstruction of large volumes of neural tissues using electron microscopy (EM), providing unprecedented insights into brain structure and function. Dendritic spines-dynamic protrusions on neuronal dendrites-play crucial roles in synaptic plasticity, influencing learning, memory, and various neurological disorders. However, current spine analysis methods often rely on manual annotation of subcellular features, limiting their ability to handle the complexity of spines in dense dendritic networks. This paper introduces a novel automated computational framework that integrates discrete differential geometry, machine learning, and 3D image processing to analyze dendritic spines in these intricate environments. By generating distributions of spine morphology from high resolution images including many thousands of spines, our approach captures subtle variations in spine shapes, offering a nuanced understanding of their roles in synaptic function. This framework is tested on multiple EM datasets, with the aim of enhancing our understanding of synaptic plasticity and its alterations in disease states. The proposed method is poised to accelerate neuroscience research by providing a scalable, objective, and comprehensive solution for spine analysis, uncovering insights into the role of spine geometry for neural function.

This study aimed to clinically evaluate a digital biomarker, the Finger Fold Index (FFI), derived from the ratio of joint diameter to finger fold surface area in hand photographs, for the assessment of joint swelling in inflammatory arthritis. Smartphone hand photographs from patients with rheumatoid (RA) and psoriatic arthritis (PsA) were analyzed using a machine learning pipeline for automated detection and processing of joint diameter and finger folds at the proximal interphalangeal (PIP) joints. The FFI was clinically evaluated by correlation with joint swelling scores (0-3) and DAS28-CRP. A healthy cohort was used to establish FFI cut-offs, which were then compared to the arthritis cohorts. A total of 1275 PIP joint images of 124 arthritis patients and 53 healthy individuals were included. FFI values correlated with swelling scores in the arthritis population with r&#x2009;=&#x2009;0.443 (95% CI 0.384-0.498), while mean FFI values were weakly correlated with DAS28-CRP dichotomized at 3.2 (r&#x2009;=&#x2009;0.310; 95% CI 0.123-0.475). ROC analysis showed moderate discriminative performance of the prediction models for PIP joint swelling (2-4), with AUCs ranging from 0.664 to 0.786 (95% CI 0.615-0.868). Notably, diagnostic performance was characterized by low specificity. FFI values exceeding the healthy cut-offs were associated with swelling (Cramer's V&#x2009;=&#x2009;0.400-0.631; p&#x2009;<&#x2009;0.001) with the strongest association observed in cases of more pronounced swelling (grade 3). Longitudinal studies are needed to assess sensitivity to change and to establish whether this biomarker can be used for remote patient monitoring. Further refinement of the algorithm is warranted, as current diagnostic accuracy remains insufficient for clinical implementation. However, integration with additional clinical information, such as patient-reported outcomes, may improve performance and support future applicability in clinical practice.

Malocclusion and the treatment process can affect the person's quality of life, which includes physical and psychological effects. Classifying the patient accurately is critical to achieving the desired output. This study aimed to improve the current diagnostic process with artificial-intelligence algorithms using the lateral cephalogram. The study sample consisted of 1014 Arab patients diagnosed with skeletal class I, II, or III. In this study, we used linear discriminant analysis (LDA), random forest (RF), decision tree (DT), K-nearest neighbors (KNN), support vector machine (SVM), and naive Bayes (NB) as classification models. In addition, we calculated the parameters' importance using two techniques-the impurity decreases and the leave-one-feature-out (LOFO) technique. Finally, we applied an artificial neural network (ANN) to classify the patients accurately. One of the influential models presented in this study was the model that included only the parameters Wits appraisal, SNB, SNA, and ML-NSL angles. It could classify the patients with an accuracy of 0.98. In addition, we applied the leave-one-feature-out technique (LOFO) for multiple random forest models and found that the Calculated_ANB (ANB angle-individualized ANB) and Wits appraisal were the most important parameters in the random forest models. Besides, age and gender were in 8th and 21st places (out of 26 variables). Furthermore, the decision tree results demonstrated the distinct characteristics of this ethnic group, which were presented by different ranges of ANB angles that define every skeletal class. The results showed that the tree's root classified the patient as skeletal class III when the ANB angle is less than 0.084 degrees, and skeletal class II if the ANB angle is greater than 1.23. In summary, this research presented a model enabling orthodontists to precisely classify orthodontic patients. Further research should include different ethnic groups to validate our findings.

Endometrial health is a key determinant of female fertility and successful pregnancy outcomes, making the accurate diagnosis of endometrial lesions essential for the success of assisted reproductive technology. While hysteroscopy remains the gold standard for uterine cavity evaluation, interpretation can vary based on clinical expertise. To address this, we developed a deep learning-based clinical decision support system to classify hysteroscopic images from high-resolution (4&#xa0;K) videos into three categories: normal endometrium, endometrial polyps, and endometritis. Endometritis cases were classified based on hysteroscopic features suggestive of inflammation; no histopathological confirmation was obtained. Utilizing a dataset of 1500 expert-annotated images from 200 clinical videos, we applied transfer learning across four architectures: VGG-16, VGG-19, DenseNet-121, and EfficientNet-B0. Our results show that the models achieved classification accuracies between 85 and 89%, with DenseNet-121 demonstrating superior performance, specifically achieving a sensitivity of 93% and an AUC of 98.8% for polyp detection, alongside a precision of 90% for endometritis. Furthermore, Grad-CAM visualization confirmed that the networks focused on clinically relevant morphological features, enhancing model interpretability. These findings suggest that deep learning may serve as a supportive tool to assist clinicians in hysteroscopic analysis, pending validation on external datasets and with pathologically confirmed labels.