Impact of functional electrical stimulation on nerve-damaged muscles by quantifying fat infiltration using deep learning.

Quantitative imaging in life sciences has evolved into a powerful approach combining advanced microscopy acquisition and automated analysis of image data. The focus of the present study is on the imaging-based evaluation of the posterior cricoarytenoid muscle (PCA) influenced by long-term functional electrical stimulation (FES), which may assist the inspiration of patients with bilateral vocal fold paresis. To this end, muscle cross-sections of the PCA of sheep were examined by quantitative image analysis. Previous investigations of the muscle fibers and the collagen amount have not revealed signs of atrophy and fibrosis due to FES by a laryngeal pacemaker. It was therefore hypothesized that regardless of the stimulation parameters the fat in the muscle cross-sections would not be significantly altered. We here extending our previous investigations using quantitative imaging of intramuscular fat in cross-sections. In order to perform this analysis both reliably and faster than a qualitative evaluation and time-consuming manual annotation, the selection of the automated method was of crucial importance. To this end, our recently established deep neural network IMFSegNet, which provides more accurate results compared to standard machine learning approaches, was applied to more than 300 H&E stained muscle cross-sections from 22 sheep. It was found that there were no significant differences in the amount of intramuscular fat between the PCA with and without long-term FES, nor were any significant differences found between the low and high duty cycle stimulated groups. This study on a human-like animal model not only confirms the hypothesis that FES with the selected parameters has no negative impact on the PCA, but also demonstrates that objective and automated deep learning-based quantitative imaging is a powerful tool for such a challenging analysis.

Deep learning-based characterization of neutrophil activation phenotypes in ex vivo human Candida blood infections.

Early identification of human pathogens is crucial for the effective treatment of bloodstream infections to prevent sepsis. Since pathogens that are present in small numbers are usually difficult to detect directly, we hypothesize that the behavior of the immune cells that are present in large numbers may provide indirect evidence about the causative pathogen of the infection. We previously applied time-lapse microscopy to observe that neutrophils isolated from human whole-blood samples, which had been infected with the human-pathogenic fungus Candida albicans or C. glabrata, indeed exhibited a characteristic morphodynamic behavior. Tracking the neutrophil movement and shape dynamics over time, combined with machine learning approach, the accuracy for the differentiation between the two Candida species was about 75%. In this study, the focus is on improving the classification accuracy of the Candida species using advanced deep learning methods. We implemented (i) gated recurrent unit (GRU) networks and transformer-based networks for video data, and (ii) convolutional neural networks (CNNs) for individual frames of the time-lapse microscopy data. While the GRU and transformer-based approaches yielded promising results with 96% and 100% accuracy, respectively, the classification based on videos proved to be very time-consuming and required several hours. In contrast, the CNN model for individual microscopy frames yielded results within minutes, and, utilizing a majority-vote technique, achieved 100% accuracy both in identifying the pathogen-free blood samples and in distinguishing between the Candida species. The applied CNN demonstrates the potential for automatically differentiating bloodstream Candida infections with high accuracy and efficiency. We further analysed the results of the CNN using explainable artificial intelligence (XAI) techniques to understand the critical features and patterns, thereby shedding light on potential key morphodynamic characteristics of neutrophils in response to different Candida species. This approach could provide new insights into host-pathogen interactions and may facilitate the development of rapid, automated diagnostic tools for differentiating fungal species in blood samples.

IMFSegNet: Cost-effective and objective quantification of intramuscular fat in histological sections by deep learning.

The assessment of muscle condition is of great importance in various research areas. In particular, evaluating the degree of intramuscular fat (IMF) in tissue sections is a challenging task, which today is still mostly performed qualitatively or quantitatively by a highly subjective and error-prone manual analysis. We here realize the mission to make automated IMF analysis possible that (i) minimizes subjectivity, (ii) provides accurate and quantitative results quickly, and (iii) is cost-effective using standard hematoxylin and eosin (H&E) stained tissue sections. To address all these needs in a deep learning approach, we utilized the convolutional encoder-decoder network SegNet to train the specialized network IMFSegNet allowing to accurately quantify the spatial distribution of IMF in histological sections. Our fully automated analysis was validated on 17 H&E-stained muscle sections from individual sheep and compared to various state-of-the-art approaches. Not only does IMFSegNet outperform all other approaches, but this neural network also provides fully automated and highly accurate results utilizing the most cost-effective procedures of sample preparation and imaging. Furthermore, we shed light on the opacity of black-box approaches such as neural networks by applying an explainable artificial intelligence technique to clarify that the success of IMFSegNet actually lies in identifying the hard-to-detect IMF structures. Embedded in our open-source visual programming language JIPipe that does not require programming skills, it can be expected that IMFSegNet advances muscle condition assessment in basic research across multiple areas as well as in research fields focusing on translational clinical applications.

Automated characterisation of neutrophil activation phenotypes in ex vivo human Candida blood infections.

Rapid identification of pathogens is required for early diagnosis and treatment of life-threatening bloodstream infections in humans. This requirement is driving the current developments of molecular diagnostic tools identifying pathogens from human whole blood after successful isolation and cultivation. An alternative approach is to determine pathogen-specific signatures from human host immune cells that have been exposed to pathogens. We hypothesise that activated immune cells, such as neutrophils, may exhibit a characteristic behaviour - for instance in terms of their speed, dynamic cell morphology - that allows (i) identifying the type of pathogen indirectly and (ii) providing information on therapeutic efficacy. In this feasibility study, we propose a method for the quantitative assessment of static and morphodynamic features of neutrophils based on label-free time-lapse imaging data. We investigate neutrophil activation phenotypes after confrontation with fungal pathogens and isolation from a human whole-blood assay. In particular, we applied a machine learning supported approach to time-lapse microscopy data from different infection scenarios and were able to distinguish between Candida albicans and C. glabrata infection scenarios with test accuracies well above 75%, and to identify pathogen-free samples with accuracy reaching 100%. These results significantly exceed the test accuracies achieved using state-of-the-art deep neural networks to classify neutrophils by their morphodynamics.

Enhanced segmentation of label-free cells for automated migration and interaction tracking.

In biomedical research, the migration behavior of cells and interactions between various cell types are frequently studied subjects. An automated and quantitative analysis of time-lapse microscopy data is an essential component of these studies, especially when characteristic migration patterns need to be identified. Plenty of software tools have been developed to serve this need. However, the majority of algorithms is designed for fluorescently labeled cells, even though it is well-known that fluorescent labels can substantially interfere with the physiological behavior of interacting cells. We here present a fully revised version of our algorithm for migration and interaction tracking (AMIT), which includes a novel segmentation approach. This approach allows segmenting label-free cells with high accuracy and also enables detecting almost all cells within the field of view. With regard to cell tracking, we designed and implemented a new method for cluster detection and splitting. This method does not rely on any geometrical characteristics of individual objects inside a cluster but relies on monitoring the events of cell-cell fusion from and cluster fission into single cells forward and backward in time. We demonstrate that focusing on these events provides accurate splitting of transient clusters. Furthermore, the substantially improved quantitative analysis of cell migration by the revised version of AMIT is more than two orders of magnitude faster than the previous implementation, which makes it feasible to process video data at higher spatial and temporal resolutions.

Comparative assessment of immune evasion mechanisms in human whole-blood infection assays by a systems biology approach.

Computer simulations of mathematical models open up the possibility of assessing hypotheses generated by experiments on pathogen immune evasion in human whole-blood infection assays. We apply an interdisciplinary systems biology approach in which virtual infection models implemented for the dissection of specific immune mechanisms are combined with experimental studies to validate or falsify the respective hypotheses. Focusing on the assessment of mechanisms that enable pathogens to evade the immune response in the early time course of a whole-blood infection, the least-square error (LSE) as a measure for the quantitative agreement between the theoretical and experimental kinetics is combined with the Akaike information criterion (AIC) as a measure for the model quality depending on its complexity. In particular, we compare mathematical models with three different types of pathogen immune evasion as well as all their combinations: (i) spontaneous immune evasion, (ii) evasion mediated by immune cells, and (iii) pre-existence of an immune-evasive pathogen subpopulation. For example, by testing theoretical predictions in subsequent imaging experiments, we demonstrate that the simple hypothesis of having a subpopulation of pre-existing immune-evasive pathogens can be ruled out. Furthermore, in this study we extend our previous whole-blood infection assays for the two fungal pathogens Candida albicans and C. glabrata by the bacterial pathogen Staphylococcus aureus and calibrated the model predictions to the time-resolved experimental data for each pathogen. Our quantitative assessment generally reveals that models with a lower number of parameters are not only scored with better AIC values, but also exhibit lower values for the LSE. Furthermore, we describe in detail model-specific and pathogen-specific patterns in the kinetics of cell populations that may be measured in future experiments to distinguish and pinpoint the underlying immune mechanisms.