Deep learning-enabled segmentation of ambiguous bioimages with deepflash2.

Bioimages frequently exhibit low signal-to-noise ratios due to experimental conditions, specimen characteristics, and imaging trade-offs. Reliable segmentation of such ambiguous images is difficult and laborious. Here we introduce deepflash2, a deep learning-enabled segmentation tool for bioimage analysis. The tool addresses typical challenges that may arise during the training, evaluation, and application of deep learning models on ambiguous data. The tool's training and evaluation pipeline uses multiple expert annotations and deep model ensembles to achieve accurate results. The application pipeline supports various use-cases for expert annotations and includes a quality assurance mechanism in the form of uncertainty measures. Benchmarked against other tools, deepflash2 offers both high predictive accuracy and efficient computational resource usage. The tool is built upon established deep learning libraries and enables sharing of trained model ensembles with the research community. deepflash2 aims to simplify the integration of deep learning into bioimage analysis projects while improving accuracy and reliability.

Integrated cardio-behavioral responses to threat define defensive states.

Fear and anxiety are brain states that evolved to mediate defensive responses to threats. The defense reaction includes multiple interacting behavioral, autonomic and endocrine adjustments, but their integrative nature is poorly understood. In particular, although threat has been associated with various cardiac changes, there is no clear consensus regarding the relevance of these changes for the integrated defense reaction. Here we identify rapid microstates that are associated with specific behaviors and heart rate dynamics, which are affected by long-lasting macrostates and reflect context-dependent threat levels. In addition, we demonstrate that one of the most commonly used defensive behavioral responses-freezing as measured by immobility-is part of an integrated cardio-behavioral microstate mediated by Chx10+ neurons in the periaqueductal gray. Our framework for systematic integration of cardiac and behavioral readouts presents the basis for a better understanding of complex neural defensive states and their associated systemic functions.

Unbiased analysis of the dorsal root ganglion after peripheral nerve injury: no neuronal loss, no gliosis, but satellite glial cell plasticity.

ABSTRACT: Pain syndromes are often accompanied by complex molecular and cellular changes in dorsal root ganglia (DRG). However, the evaluation of cellular plasticity in the DRG is often performed by heuristic manual analysis of a small number of representative microscopy image fields. In this study, we introduce a deep learning-based strategy for objective and unbiased analysis of neurons and satellite glial cells (SGCs) in the DRG. To validate the approach experimentally, we examined serial sections of the rat DRG after spared nerve injury (SNI) or sham surgery. Sections were stained for neurofilament, glial fibrillary acidic protein (GFAP), and glutamine synthetase (GS) and imaged using high-resolution large-field (tile) microscopy. After training of deep learning models on consensus information of different experts, thousands of image features in DRG sections were analyzed. We used known (GFAP upregulation), controversial (neuronal loss), and novel (SGC phenotype switch) changes to evaluate the method. In our data, the number of DRG neurons was similar 14 d after SNI vs sham. In GFAP-positive subareas, the percentage of neurons in proximity to GFAP-positive cells increased after SNI. In contrast, GS-positive signals, and the percentage of neurons in proximity to GS-positive SGCs decreased after SNI. Changes in GS and GFAP levels could be linked to specific DRG neuron subgroups of different size. Hence, we could not detect gliosis but plasticity changes in the SGC marker expression. Our objective analysis of DRG tissue after peripheral nerve injury shows cellular plasticity responses of SGCs in the whole DRG but neither injury-induced neuronal death nor gliosis.

On the objectivity, reliability, and validity of deep learning enabled bioimage analyses.

Bioimage analysis of fluorescent labels is widely used in the life sciences. Recent advances in deep learning (DL) allow automating time-consuming manual image analysis processes based on annotated training data. However, manual annotation of fluorescent features with a low signal-to-noise ratio is somewhat subjective. Training DL models on subjective annotations may be instable or yield biased models. In turn, these models may be unable to reliably detect biological effects. An analysis pipeline integrating data annotation, ground truth estimation, and model training can mitigate this risk. To evaluate this integrated process, we compared different DL-based analysis approaches. With data from two model organisms (mice, zebrafish) and five laboratories, we show that ground truth estimation from multiple human annotators helps to establish objectivity in fluorescent feature annotations. Furthermore, ensembles of multiple models trained on the estimated ground truth establish reliability and validity. Our research provides guidelines for reproducible DL-based bioimage analyses.

Research in biology generates many image datasets, mostly from microscopy. These images have to be analyzed, and much of this analysis relies on a human expert looking at the images and manually annotating features. Image datasets are often large, and human annotation can be subjective, so automating image analysis is highly desirable. This is where machine learning algorithms, such as deep learning, have proven to be useful. In order for deep learning algorithms to work first they have to be 'trained'. Deep learning algorithms are trained by being given a training dataset that has been annotated by human experts. The algorithms extract the relevant features to look out for from this training dataset and can then look for these features in other image data. However, it is also worth noting that because these models try to mimic the annotation behavior presented to them during training as well as possible, they can sometimes also mimic an expert's subjectivity when annotating data. Segebarth, Griebel et al. asked whether this was the case, whether it had an impact on the outcome of the image data analysis, and whether it was possible to avoid this problem when using deep learning for imaging dataset analysis. For this research, Segebarth, Griebel et al. used microscopy images of mouse brain sections, where a protein called cFOS had been labeled with a fluorescent tag. This protein typically controls the rate at which DNA information is copied into RNA, leading to the production of proteins. Its activity can be influenced experimentally by testing the behaviors of mice. Thus, this experimental manipulation can be used to evaluate the results of deep learning-based image analyses. First, the fluorescent images were interpreted manually by a group of human experts. Then, their results were used to train a large variety of deep learning models. Models were trained either on the results of an individual expert or on the results pooled from all experts to come up with a consensus model, a deep learning model that learned from the personal annotation preferences of all experts. This made it possible to test whether training a model on multiple experts reduces the risk of subjectivity. As the training of deep learning models is random, Segebarth, Griebel et al. also tested whether combining the predictions from multiple models in a so-called model ensemble improves the consistency of the analyses. For evaluation, the annotations of the deep learning models were compared to those of the human experts, to ensure that the results were not influenced by the subjective behavior of one person. The results of all bioimage annotations were finally compared to the experimental results from analyzing the mice's behaviors in order to check whether the models were able to find the behavioral effect on cFOS. Segebarth, Griebel et al. concluded that combining the expert knowledge of multiple experts reduces the subjectivity of bioimage annotation by deep learning algorithms. Combining such consensus information in a group of deep learning models improves the quality of bioimage analysis, so that the results are reliable, transparent and less subjective.

A novel missense variant in MYO3A is associated with autosomal dominant high-frequency hearing loss in a German family.

BACKGROUND: MYO3A, encoding the myosin IIIA protein, is associated with autosomal recessive and autosomal dominant nonsyndromic hearing loss. To date, only two missense variants located in the motor-head domain of MYO3A have been described in autosomal dominant families with progressive, mild-to-profound sensorineural hearing loss. These variants alter the ATPase activity of myosin IIIA. METHODS: Exome sequencing of a proband from a three-generation German family with prelingual, moderate-to-profound, high-frequency hearing loss was performed. Segregation analysis confirmed a dominant inheritance pattern. Regression analysis of mean hearing level thresholds per individual and ear was performed at high-, mid-, and low-frequencies. RESULTS: A novel heterozygous missense variant c.716T>C, p.(Leu239Pro) in the kinase domain of MYO3A was identified that is predicted in silico as disease causing. High-frequency, progressive hearing loss was identified. CONCLUSION: Correlation analysis of pure-tone hearing thresholds revealed progressive hearing loss, especially in the high-frequencies. In the present study, we report the first dominant likely pathogenic variant in MYO3A in a European family and further support MYO3A as an autosomal dominant hearing loss gene.

Reward signaling in a recurrent circuit of dopaminergic neurons and peptidergic Kenyon cells.

Dopaminergic neurons in the brain of the Drosophila larva play a key role in mediating reward information to the mushroom bodies during appetitive olfactory learning and memory. Using optogenetic activation of Kenyon cells we provide evidence that recurrent signaling exists between Kenyon cells and dopaminergic neurons of the primary protocerebral anterior (pPAM) cluster. Optogenetic activation of Kenyon cells paired with odor stimulation is sufficient to induce appetitive memory. Simultaneous impairment of the dopaminergic pPAM neurons abolishes appetitive memory expression. Thus, we argue that dopaminergic pPAM neurons mediate reward information to the Kenyon cells, and in turn receive feedback from Kenyon cells. We further show that this feedback signaling is dependent on short neuropeptide F, but not on acetylcholine known to be important for odor-shock memories in adult flies. Our data suggest that recurrent signaling routes within the larval mushroom body circuitry may represent a mechanism subserving memory stabilization.