Explore the vast range of titles available.

Many biological applications require the segmentation of cell bodies, membranes and nuclei from microscopy images. Deep learning has enabled great progress on this problem, but current methods are specialized for images that have large training datasets. Here we introduce a generalist, deep learning-based segmentation method called Cellpose, which can precisely segment cells from a wide range of image types and does not require model retraining or parameter adjustments. Cellpose was trained on a new dataset of highly varied images of cells, containing over 70,000 segmented objects. We also demonstrate a three-dimensional (3D) extension of Cellpose that reuses the two-dimensional (2D) model and does not require 3D-labeled data. To support community contributions to the training data, we developed software for manual labeling and for curation of the automated results. Periodically retraining the model on the community-contributed data will ensure that Cellpose improves constantly.Cellpose is a generalist, deep learning-based approach for segmenting structures in a wide range of image types. Cellpose does not require parameter adjustment or model retraining and outperforms established methods on 2D and 3D datasets.

Many facts come with an expiration date, from the name of the President to the basketball team Lebron James plays for. However, most language models (LMs) are trained on snapshots of data collected at a specific moment in time. This can limit their utility, especially in the closed-book setting where the pretraining corpus must contain the facts the model should memorize. We introduce a diagnostic dataset aimed at probing LMs for factual knowledge that changes over time and highlight problems with LMs at either end of the spectrum—those trained on specific slices of temporal data, as well as those trained on a wide range of temporal data. To mitigate these problems, we propose a simple technique for jointly modeling text with its timestamp. This improves memorization of seen facts from the training time period, as well as calibration on predictions about unseen facts from future time periods. We also show that models trained with temporal context can be efficiently “refreshed” as new data arrives, without the need for retraining from scratch.

Global climate models represent small-scale processes such as convection using subgrid models known as parameterizations, and these parameterizations contribute substantially to uncertainty in climate projections. Machine learning of new parameterizations from high-resolution model output is a promising approach, but such parameterizations have been prone to issues of instability and climate drift, and their performance for different grid spacings has not yet been investigated. Here we use a random forest to learn a parameterization from coarse-grained output of a three-dimensional high-resolution idealized atmospheric model. The parameterization leads to stable simulations at coarse resolution that replicate the climate of the high-resolution simulation. Retraining for different coarse-graining factors shows the parameterization performs best at smaller horizontal grid spacings. Our results yield insights into parameterization performance across length scales, and they also demonstrate the potential for learning parameterizations from global high-resolution simulations that are now emerging. Machine learning has been used to represent small-scale processes, such as clouds, in atmospheric models but this can lead to instability in simulations of climate. Here, the authors demonstrate a use of machine learning in an atmospheric model that leads to stable simulations of climate at a range of grid spacings.

Background Running-related injuries are prevalent among distance runners. Changing step rate is a commonly used running retraining strategy in the management and prevention of running-related injuries. Objective The aims of this review were to synthesise the evidence relating to the effects of changing running step rate on injury, performance and biomechanics. Design Systematic review and meta-analysis. Data Sources MEDLINE, EMBASE, CINAHL, and SPORTDiscus. Results Thirty-seven studies were included that related to injury ( n  = 2), performance ( n  = 5), and biomechanics ( n  = 36). Regarding injury, very limited evidence indicated that increasing running step rate is associated with improvements in pain (4 weeks: standard mean difference (SMD), 95% CI 2.68, 1.52 to 3.83; 12 weeks: 3.62, 2.24 to 4.99) and function (4 weeks: 2.31, 3.39 to 1.24); 12 weeks: 3.42, 4.75 to 2.09) in recreational runners with patellofemoral pain. Regarding performance, very limited evidence indicated that increasing step rate increases perceived exertion ( − 0.49,  − 0.91 to − 0.07) and awkwardness (− 0.72, − 1.38 to − 0.06) and effort (− 0.69, − 1.34, − 0.03); and very limited evidence that an increase in preferred step rate is associated with increased metabolic energy consumption (− 0.84, − 1.57 to − 0.11). Regarding biomechanics, increasing running step rate was associated with strong evidence of reduced peak knee flexion angle (0.66, 0.40 to 0.92); moderate evidence of reduced step length (0.93, 0.49 to 1.37), peak hip adduction (0.40, 0.11 to 0.69), and peak knee extensor moment (0.50, 0.18 to 0.81); moderate evidence of reduced foot strike angle (0.62, 034 to 0.90); limited evidence of reduced braking impulse (0.64, 0.29 to 1.00), peak hip flexion (0.42, 0.10 to 0.75), and peak patellofemoral joint stress (0.56, 0.07 to 1.05); and limited evidence of reduced negative hip (0.55, 0.20 to 0.91) and knee work (0.84, 0.48 to 1.20). Decreasing running step rate was associated with moderate evidence of increased step length (− 0.76, − 1.31 to − 0.21); limited evidence of increased contact time (− 0.95, − 1.49 to − 0.40), braking impulse (− 0.73, − 1.08 to − 0.37), and negative knee work (− 0.88, − 1.25 to − 0.52); and limited evidence of reduced negative ankle work (0.38, 0.03 to 0.73) and negative hip work (0.49, 0.07 to 0.91). Conclusion In general, increasing running step rate results in a reduction (or no change), and reducing step rate results in an increase (or no change), to kinetic, kinematic, and loading rate variables at the ankle, knee and hip. At present there is insufficient evidence to conclusively determine the effects of altering running step rate on injury and performance. As most studies included in this review investigated the immediate effects of changing running step rate, the longer-term effects remain largely unknown. Prospero Registration CRD42020167657.

Purpose To compare the effect of three differently spaced retraining schedules (1-day, 2-day, and 1-week intervals) on the acquisition of basic arthroscopic skills and skill retention after 3 months. Methods Thirty orthopaedic residents without arthroscopic experience were enrolled in a double-blind, randomised, parallel-controlled trial. Spaced retaining schedules were divided into massed training and retraining phases. Participants were required to obtain perfect scores in all tasks on the simulator in the massed training phase, followed by a pretest to evaluate the training effect. During the retraining phase, participants were randomly assigned to Groups A (1-day interval), B (2-day interval) or C (1-week interval). A posttest was used to evaluate the effect of different retraining patterns. Follow-up evaluations were conducted at 1 week, 1 month and 3 months after the completion of spaced retraining schedules to measure skill retention. One-way ANOVA and paired-sample t tests were used for statistical analysis. Results Significant between-group differences in diagnostic arthroscopy (137.0 ± 24.8 vs. 140.1 ± 21.3 vs. 175.3 ± 27.4 s, P (A−C)  = 0.005, P (B−C)  = 0.010) and loose body removal (193.1 ± 33.9 vs. 182.0 ± 32.1 vs. 228.7 ± 42.9 s, P (B−C)  = 0.025) completion times were observed. No significant differences were found in other posttest metrics. An assessment of skill retention after the 3-month follow-up (Evaluation 3) showed significant differences in diagnostic arthroscopy completion time (202.5 ± 53.3 vs. 172.0 ± 27.2 vs. 225.5 ± 42.1 s, P (B−C)  = 0.026). No significant differences were found in other Evaluation 3 metrics. Conclusion The 2-day retraining schedule was the most effective for the acquisition and retention of basic arthroscopic skills and could be integrated into arthroscopic skills curricula. After a 3-month follow-up, residents who followed this schedule showed better skill retention than those who followed the 1-week interval schedule. Level of evidence Level I.

Remote sensing can transform the speed, scale, and cost of biodiversity and forestry surveys. Data acquisition currently outpaces the ability to identify individual organisms in high resolution imagery. We outline an approach for identifying tree-crowns in RGB imagery while using a semi-supervised deep learning detection network. Individual crown delineation has been a long-standing challenge in remote sensing and available algorithms produce mixed results. We show that deep learning models can leverage existing Light Detection and Ranging (LIDAR)-based unsupervised delineation to generate trees that are used for training an initial RGB crown detection model. Despite limitations in the original unsupervised detection approach, this noisy training data may contain information from which the neural network can learn initial tree features. We then refine the initial model using a small number of higher-quality hand-annotated RGB images. We validate our proposed approach while using an open-canopy site in the National Ecological Observation Network. Our results show that a model using 434,551 self-generated trees with the addition of 2848 hand-annotated trees yields accurate predictions in natural landscapes. Using an intersection-over-union threshold of 0.5, the full model had an average tree crown recall of 0.69, with a precision of 0.61 for the visually-annotated data. The model had an average tree detection rate of 0.82 for the field collected stems. The addition of a small number of hand-annotated trees improved the performance over the initial self-supervised model. This semi-supervised deep learning approach demonstrates that remote sensing can overcome a lack of labeled training data by generating noisy data for initial training using unsupervised methods and retraining the resulting models with high quality labeled data.

Introduction Vocal cord dysfunction (VCD) is a complex disorder characterized by episodic adduction of the vocal folds during inspiration and expiration, which can lead to dyspnea, wheezing, cough, and acute‐onset respiratory distress. Currently, there is a lack of standardized criteria among treating physicians across multiple disciplines, including otolaryngologists, pulmonologists, allergists, and speech and language pathologists, for diagnosis and treatment of VCD, although laryngeal‐respiratory retraining therapy (LRT) has emerged as the preferred treatment modality. Objective In the present study, we examined the efficacy of LRT in patients presenting with a clinical diagnosis of VCD in the presence and absence of laryngeal adduction on laryngoscopy. Results Overall, 74.1% of the cohort showed a response to LRT, of which 62.1% were partial and 12.1% were significant responses. When comparing between patients with and without laryngeal adduction on laryngoscopy, there were no significant differences in the number of sessions of LRT undertaken, mean time to response, and overall response rate between the groups. Conclusion Our findings suggest that LRT should be utilized for all patients presenting with symptoms of VCD, even in the absence of laryngeal adduction on laryngoscopy. In this retrospective study, we found that abnormal or normal laryngoscopy findings did not predict patient‐reported outcomes in response to speech therapy among a cohort of patients with vocal cord dysfunction (VCD). This finding is significant because it provides further evidence for the utility of speech therapy even in patients with only a clinical presumption of VCD with objectively normal laryngoscopy findings. To date, the literature regarding outcomes for VCD is still lacking, and thus we believe that this manuscript would be interesting and relevant for your audience.

Medical image classification plays an essential role in clinical treatment and teaching tasks. However, the traditional method has reached its ceiling on performance. Moreover, by using them, much time and effort need to be spent on extracting and selecting classification features. The deep neural network is an emerging machine learning method that has proven its potential for different classification tasks. Notably, the convolutional neural network dominates with the best results on varying image classification tasks. However, medical image datasets are hard to collect because it needs a lot of professional expertise to label them. Therefore, this paper researches how to apply the convolutional neural network (CNN) based algorithm on a chest X-ray dataset to classify pneumonia. Three techniques are evaluated through experiments. These are linear support vector machine classifier with local rotation and orientation free features, transfer learning on two convolutional neural network models: Visual Geometry Group i.e., VGG16 and InceptionV3, and a capsule network training from scratch. Data augmentation is a data preprocessing method applied to all three methods. The results of the experiments show that data augmentation generally is an effective way for all three algorithms to improve performance. Also, Transfer learning is a more useful classification method on a small dataset compared to a support vector machine with oriented fast and rotated binary (ORB) robust independent elementary features and capsule network. In transfer learning, retraining specific features on a new target dataset is essential to improve performance. And, the second important factor is a proper network complexity that matches the scale of the dataset.

Despite prolonged scientific efforts to unravel the effects of ions on the structure and dynamics of water, many open questions remain, in particular concerning the spatial extent of this effect (i.e., the number of water molecules affected) and the origin of ion-specific effects. A combined terahertz and femtosecond infrared spectroscopic study of water dynamics around different ions (specifically magnesium, lithium, sodium, and cesium cations, as well as sulfate, chloride, iodide, and perchlorate anions) reveals that the effect of ions and counterions on water can be strongly interdependent and nonadditive, and in certain cases extends well beyond the first solvation shell of water molecules directly surrounding the ion.

Infant cardiopulmonary resuscitation (iCPR) is often poorly performed, predominantly because of ineffective learning, poor retention and decay of skills over time. The aim of this study was to investigate whether an individualized, competence-based approach to simulated iCPR retraining could result in high skill retention of infant chest compressions (iCC) at follow-up. An observational study with 118 healthcare students was conducted over 12 months from November 2019. Participants completed pediatric resuscitation training and a 2-min assessment on an infant mannequin. Participants returned for monthly assessment until iCC competence was achieved. Competence was determined by passing assessments in two consecutive months. After achieving competence, participants returned just at follow-up. For each ‘FAIL’ during assessment, up to six minutes of practice using real-time feedback was completed and the participant returned the following month. This continued until two consecutive monthly ‘PASSES’ were achieved, following which, the participant was deemed competent and returned just at follow-up. Primary outcome was retention of competence at follow-up. Descriptive statistics were used to analyze demographic data. Independent t-test or Mann–Whitney U test were used to analyze the baseline characteristics of those who dropped out compared to those remaining in the study. Differences between groups retaining competence at follow-up were determined using the Fisher exact test. On completion of training, 32 of 118 participants passed the assessment. Of those achieving iCC competence at month 1, 96% retained competence at 9–10 months; of those achieving competence at month 2, 86% demonstrated competence at 8–9 months; of those participants achieving competence at month 3, 67% retained competence at 7–8 months; for those achieving competence at month 4, 80% demonstrated retention at 6–7 months.    Conclusion : Becoming iCC competent after initial training results in high levels of skill retention at follow-up, regardless of how long it takes to achieve competence. What is Known: • Infant cardiopulmonary resuscitation (iCPR) is often poorly performed and skills decay within months after training . • Regular iCPR skills updates are important, but the optimal retraining interval considering individual training needs has yet to be established . What is New: • Infant chest compression (iCC) competence can be achieved within one to four months after training and once achieved, it can be retained for many months . • With skill reinforcement of up to 28 minutes after initial training, 90% of individuals were able to achieve competence in iCC and 86% retained this competence at follow-up .