Assessing robustness and generalization of a deep neural network for brain MS lesion segmentation on real-world data.

OBJECTIVES: Evaluate the performance of a deep learning (DL)-based model for multiple sclerosis (MS) lesion segmentation and compare it to other DL and non-DL algorithms. METHODS: This ambispective, multicenter study assessed the performance of a DL-based model for MS lesion segmentation and compared it to alternative DL- and non-DL-based methods. Models were tested on internal (n = 20) and external (n = 18) datasets from Latin America, and on an external dataset from Europe (n = 49). We also examined robustness by rescanning six patients (n = 6) from our MS clinical cohort. Moreover, we studied inter-human annotator agreement and discussed our findings in light of these results. Performance and robustness were assessed using intraclass correlation coefficient (ICC), Dice coefficient (DC), and coefficient of variation (CV). RESULTS: Inter-human ICC ranged from 0.89 to 0.95, while spatial agreement among annotators showed a median DC of 0.63. Using expert manual segmentations as ground truth, our DL model achieved a median DC of 0.73 on the internal, 0.66 on the external, and 0.70 on the challenge datasets. The performance of our DL model exceeded that of the alternative algorithms on all datasets. In the robustness experiment, our DL model also achieved higher DC (ranging from 0.82 to 0.90) and lower CV (ranging from 0.7 to 7.9%) when compared to the alternative methods. CONCLUSION: Our DL-based model outperformed alternative methods for brain MS lesion segmentation. The model also proved to generalize well on unseen data and has a robust performance and low processing times both on real-world and challenge-based data. CLINICAL RELEVANCE STATEMENT: Our DL-based model demonstrated superior performance in accurately segmenting brain MS lesions compared to alternative methods, indicating its potential for clinical application with improved accuracy, robustness, and efficiency. KEY POINTS: • Automated lesion load quantification in MS patients is valuable; however, more accurate methods are still necessary. • A novel deep learning model outperformed alternative MS lesion segmentation methods on multisite datasets. • Deep learning models are particularly suitable for MS lesion segmentation in clinical scenarios.

2020-2021 field seasons of Maize GxE project within the Genomes to Fields Initiative.

OBJECTIVES: This release note describes the Maize GxE project datasets within the Genomes to Fields (G2F) Initiative. The Maize GxE project aims to understand genotype by environment (GxE) interactions and use the information collected to improve resource allocation efficiency and increase genotype predictability and stability, particularly in scenarios of variable environmental patterns. Hybrids and inbreds are evaluated across multiple environments and phenotypic, genotypic, environmental, and metadata information are made publicly available. DATA DESCRIPTION: The datasets include phenotypic data of the hybrids and inbreds evaluated in 30 locations across the US and one location in Germany in 2020 and 2021, soil and climatic measurements and metadata information for all environments (combination of year and location), ReadMe, and description files for each data type. A set of common hybrids is present in each environment to connect with previous evaluations. Each environment had a collaborator responsible for collecting and submitting the data, the GxE coordination team combined all the collected information and removed obvious erroneous data. Collaborators received the combined data to use, verify and declare that the data generated in their own environments was accurate. Combined data is released to the public with minimal filtering to maintain fidelity to the original data.

Automatic and unbiased segmentation and quantification of myofibers in skeletal muscle.

Skeletal muscle has the remarkable ability to regenerate. However, with age and disease muscle strength and function decline. Myofiber size, which is affected by injury and disease, is a critical measurement to assess muscle health. Here, we test and apply Cellpose, a recently developed deep learning algorithm, to automatically segment myofibers within murine skeletal muscle. We first show that tissue fixation is necessary to preserve cellular structures such as primary cilia, small cellular antennae, and adipocyte lipid droplets. However, fixation generates heterogeneous myofiber labeling, which impedes intensity-based segmentation. We demonstrate that Cellpose efficiently delineates thousands of individual myofibers outlined by a variety of markers, even within fixed tissue with highly uneven myofiber staining. We created a novel ImageJ plugin (LabelsToRois) that allows processing of multiple Cellpose segmentation images in batch. The plugin also contains a semi-automatic erosion function to correct for the area bias introduced by the different stainings, thereby identifying myofibers as accurately as human experts. We successfully applied our segmentation pipeline to uncover myofiber regeneration differences between two different muscle injury models, cardiotoxin and glycerol. Thus, Cellpose combined with LabelsToRois allows for fast, unbiased, and reproducible myofiber quantification for a variety of staining and fixation conditions.

Brain volumes quantification from MRI in healthy controls: Assessing correlation, agreement and robustness of a convolutional neural network-based software against FreeSurfer, CAT12 and FSL.

BACKGROUND AND PURPOSE: There are instances in which an estimate of the brain volume should be obtained from MRI in clinical practice. Our objective is to calculate cross-sectional robustness of a convolutional neural network (CNN) based software (Entelai Pic) for brain volume estimation and compare it to traditional software such as FreeSurfer, CAT12 and FSL in healthy controls (HC). MATERIALS AND METHODS: Sixteen HC were scanned four times, two different days on two different MRI scanners (1.5 T and 3 T). Volumetric T1-weighted images were acquired and post-processed with FreeSurfer v6.0.0, Entelai Pic v2, CAT12 v12.5 and FSL v5.0.9. Whole-brain, grey matter (GM), white matter (WM) and cerebrospinal fluid (CSF) volumes were calculated. Correlation and agreement between methods was assessed using intraclass correlation coefficient (ICC) and Bland Altman plots. Robustness was assessed using the coefficient of variation (CV). RESULTS: Whole-brain volume estimation had better correlation between FreeSurfer and Entelai Pic (ICC (95% CI) 0.96 (0.94-0.97)) than FreeSurfer and CAT12 (0.92 (0.88-0.96)) and FSL (0.87 (0.79-0.91)). WM, GM and CSF showed a similar trend. Compared to FreeSurfer, Entelai Pic provided similarly robust segmentations of brain volumes both on same-scanner (mean CV 1.07, range 0.20-3.13% vs. mean CV 1.05, range 0.21-3.20%, p = 0.86) and on different-scanner variables (mean CV 3.84, range 2.49-5.91% vs. mean CV 3.84, range 2.62-5.13%, p = 0.96). Mean post-processing times were 480, 5, 40 and 5 min for FreeSurfer, Entelai Pic, CAT12 and FSL respectively. CONCLUSION: Based on robustness and processing times, our CNN-based model is suitable for cross-sectional volumetry on clinical practice.

COVID-19 pneumonia accurately detected on chest radiographs with artificial intelligence.

PURPOSE: To investigate the diagnostic performance of an Artificial Intelligence (AI) system for detection of COVID-19 in chest radiographs (CXR), and compare results to those of physicians working alone, or with AI support. MATERIALS AND METHODS: An AI system was fine-tuned to discriminate confirmed COVID-19 pneumonia, from other viral and bacterial pneumonia and non-pneumonia patients and used to review 302 CXR images from adult patients retrospectively sourced from nine different databases. Fifty-four physicians blind to diagnosis, were invited to interpret images under identical conditions in a test set, and randomly assigned either to receive or not receive support from the AI system. Comparisons were then made between diagnostic performance of physicians working with and without AI support. AI system performance was evaluated using the area under the receiver operating characteristic (AUROC), and sensitivity and specificity of physician performance compared to that of the AI system. RESULTS: Discrimination by the AI system of COVID-19 pneumonia showed an AUROC curve of 0.96 in the validation and 0.83 in the external test set, respectively. The AI system outperformed physicians in the AUROC overall (70% increase in sensitivity and 1% increase in specificity, p < 0.0001). When working with AI support, physicians increased their diagnostic sensitivity from 47% to 61% (p < 0.001), although specificity decreased from 79% to 75% (p = 0.007). CONCLUSIONS: Our results suggest interpreting chest radiographs (CXR) supported by AI, increases physician diagnostic sensitivity for COVID-19 detection. This approach involving a human-machine partnership may help expedite triaging efforts and improve resource allocation in the current crisis.

Cell cycle dynamics of mouse embryonic stem cells in the ground state and during transition to formative pluripotency.

Mouse embryonic stem cells (mESCs) can be maintained as homogeneous populations in the ground state of pluripotency. Release from this state in minimal conditions allows to obtain cells that resemble those of the early post-implantation epiblast, providing an important developmental model to study cell identity transitions. However, the cell cycle dynamics of mESCs in the ground state and during its dissolution have not been extensively studied. By performing live imaging experiments of mESCs bearing cell cycle reporters, we show here that cells in the pluripotent ground state display a cell cycle structure comparable to the reported for mESCs in serum-based media. Upon release from self-renewal, the cell cycle is rapidly accelerated by a reduction in the length of the G1 phase and of the S/G2/M phases, causing an increased proliferation rate. Analysis of cell lineages indicates that cell cycle variables of sister cells are highly correlated, suggesting the existence of inherited cell cycle regulators from the parental cell. Together with a major morphological reconfiguration upon differentiation, our findings support a correlation between this in vitro model and early embryonic events.

Rats Synchronize Locomotion with Ultrasonic Vocalizations at the Subsecond Time Scale.

Acoustic signals have the potential for transmitting information fast across distances. Rats emit ultrasonic vocalizations of two distinct classes: "22-kHz" or "alarm" calls and "50-kHz" calls. The latter comprises brief sounds in the 30-80-kHz range, whose ethological role is not fully understood. We recorded ultrasonic vocalizations from pairs of rats freely behaving in neighboring but separated arenas. 50-kHz vocalizations in this condition were tightly linked to the locomotion of the emitter at the subsecond time scale, their rate sharply increasing and decreasing prior to the onset and offset of movement respectively. This locomotion-linked vocalization behavior showed a clear "audience effect," as rats recorded alone displayed lower vocal production than rats in social settings for equivalent speeds of locomotion. Furthermore, calls from different categories across the 50 and 22-kHz families displayed markedly different correlations with locomotor activity. Our results show that rat vocalizations in the high ultrasonic range are social signals carrying spatial information about the emitter and highlight the possibility that they may play a role in the social coordination of spatial behaviors.

Rodent ultrasonic vocalizations are bound to active sniffing behavior.

During rodent active behavior, multiple orofacial sensorimotor behaviors, including sniffing and whisking, display rhythmicity in the theta range (~5-10 Hz). During specific behaviors, these rhythmic patterns interlock, such that execution of individual motor programs becomes dependent on the state of the others. Here we performed simultaneous recordings of the respiratory cycle and ultrasonic vocalization emission by adult rats and mice in social settings. We used automated analysis to examine the relationship between breathing patterns and vocalization over long time periods. Rat ultrasonic vocalizations (USVs, "50 kHz") were emitted within stretches of active sniffing (5-10 Hz) and were largely absent during periods of passive breathing (1-4 Hz). Because ultrasound was tightly linked to the exhalation phase, the sniffing cycle segmented vocal production into discrete calls and imposed its theta rhythmicity on their timing. In turn, calls briefly prolonged exhalations, causing an immediate drop in sniffing rate. Similar results were obtained in mice. Our results show that ultrasonic vocalizations are an integral part of the rhythmic orofacial behavioral ensemble. This complex behavioral program is thus involved not only in active sensing but also in the temporal structuring of social communication signals. Many other social signals of mammals, including monkey calls and human speech, show structure in the theta range. Our work points to a mechanism for such structuring in rodent ultrasonic vocalizations.

Far transfer to language and math of a short software-based gaming intervention.

Executive functions (EF) in children can be trained, but it remains unknown whether training-related benefits elicit far transfer to real-life situations. Here, we investigate whether a set of computerized games might yield near and far transfer on an experimental and an active control group of low-SES otherwise typically developing 6-y-olds in a 3-mo pretest-training-posttest design that was ecologically deployed (at school). The intervention elicits transfer to some (but not all) facets of executive function. These changes cascade to real-world measures of school performance. The intervention equalizes academic outcomes across children who regularly attend school and those who do not because of social and familiar circumstances.

Scale-invariant transition probabilities in free word association trajectories.

Free-word association has been used as a vehicle to understand the organization of human thoughts. The original studies relied mainly on qualitative assertions, yielding the widely intuitive notion that trajectories of word associations are structured, yet considerably more random than organized linguistic text. Here we set to determine a precise characterization of this space, generating a large number of word association trajectories in a web implemented game. We embedded the trajectories in the graph of word co-occurrences from a linguistic corpus. To constrain possible transport models we measured the memory loss and the cycling probability. These two measures could not be reconciled by a bounded diffusive model since the cycling probability was very high (16% of order-2 cycles) implying a majority of short-range associations whereas the memory loss was very rapid (converging to the asymptotic value in approximately 7 steps) which, in turn, forced a high fraction of long-range associations. We show that memory loss and cycling probabilities of free word association trajectories can be simultaneously accounted by a model in which transitions are determined by a scale invariant probability distribution.

Folic acid supplementation in dementia: a preliminary report.

Recent work on high plasma homocysteine levels in patients at risk for developing Alzheimer's disease has led to the hypothesis that folic acid supplementation might reduce risk in such patients. The authors report on the effects of folic acid 10 mg/day versus placebo on 11 patients (only 7 completers) with dementia and low-normal folic acid levels. This is the first study evaluating folic acid or placebo in patients with dementia. Subjects had low-normal baseline folic acid levels. The magnitude of change between baseline and second testing was not statistically significant between the 2 groups. However, there was a trend for the folate group to perform worse on two specific cognitive measures, suggesting a possible trend toward worsening of some cognitive abilities after the folic acid. The folic acid in very high doses was well tolerated. Larger studies are necessary before empirically administering folic acid to patients already suffering from dementia.