Automated detection of GFAP-labeled astrocytes in micrographs using YOLOv5.

Astrocytes, a subtype of glial cells with a complex morphological structure, are active players in many aspects of the physiology of the central nervous system (CNS). However, due to their highly involved interaction with other cells in the CNS, made possible by their morphological complexity, the precise mechanisms regulating astrocyte function within the CNS are still poorly understood. This knowledge gap is also due to the current limitations of existing quantitative image analysis tools that are unable to detect and analyze images of astrocyte with sufficient accuracy and efficiency. To address this need, we introduce a new deep learning framework for the automated detection of GFAP-immunolabeled astrocytes in brightfield or fluorescent micrographs. A major novelty of our approach is the applications of YOLOv5, a sophisticated deep learning platform designed for object detection, that we customized to derive optimized classification models for the task of astrocyte detection. Extensive numerical experiments using multiple image datasets show that our method performs very competitively against both conventional and state-of-the-art methods, including the case of images where astrocytes are very dense. In the spirit of reproducible research, our numerical code and annotated data are released open source and freely available to the scientific community.

Predicting the Effects of Repetitive Transcranial Magnetic Stimulation on Cognitive Functions in Patients With Alzheimer's Disease by Automated EEG Analysis.

Alzheimer's disease (AD) is a progressive, neurodegenerative brain disorder that generally affects the elderly. Today, after the limited benefit of the pharmacological treatment strategies, numerous noninvasive brain stimulation techniques have been developed. Transcranial magnetic stimulation (TMS), based on electromagnetic stimulation, is one of the most widely used methods. The main problem in the use of TMS is the existence of large individual variability in the results. This causes a waste of money, time, and more importantly, a burden for delicate patients. Hence, it is a necessity to form an efficient and personalized TMS application protocol. In this paper, we performed a machine-learning analysis to see whether it is possible to predict the responses of patients with AD to TMS by analyzing their electroencephalography (EEG) signals. For that purpose, we analyzed both the EEG signals collected before and after the TMS application (EEG1 and EEG2, respectively). Through correlating EEG1 and repetitive transcranial magnetic stimulation (rTMS) outcomes, we tried to see whether it is possible to predict patients' responses before the treatment application. On the other hand, by EEG2 analysis, we investigated TMS impacts on EEG, more importantly if this impact is correlated with patients' response to the treatment. We used the support vector machine (SVM) classifier due to its multiple advantages for the current task with feature selection processes by stepwise linear discriminant analysis (SWLDA) and SVM. However, to justify our numerical analysis framework, we examined and compared the performances of different feature selection and classification techniques. Since we have a limited sample number, we used the leave-one-out method for the validation with the Monte Carlo technique to eliminate bias by a small sample size. In the conclusion, we observed that the correlation between rTMS outcomes and EEG2 is stronger than EEG1, since we observed, respectively, 93 and 79% of accuracies during our data analysis. Besides the informative features of EEG2 are focused on theta band, it indicates that TMS is characterizing the theta band signals in patients with AD in direct relation to patients' response to rTMS. This shows that it is more possible to determine patients' benefit from the TMS at the early stages of the treatment, which would increase the efficiency of rTMS applications in patients with Alzheimer's disease.

A multistep deep learning framework for the automated detection and segmentation of astrocytes in fluorescent images of brain tissue.

While astrocytes have been traditionally described as passive supportive cells, studies during the last decade have shown they are active players in many aspects of CNS physiology and function both in normal and disease states. However, the precise mechanisms regulating astrocytes function and interactions within the CNS are still poorly understood. This knowledge gap is due in large part to the limitations of current image analysis tools that cannot process astrocyte images efficiently and to the lack of methods capable of quantifying their complex morphological characteristics. To provide an unbiased and accurate framework for the quantitative analysis of fluorescent images of astrocytes, we introduce a new automated image processing pipeline whose main novelties include an innovative module for cell detection based on multiscale directional filters and a segmentation routine that leverages deep learning and sparse representations to reduce the need of training data and improve performance. Extensive numerical tests show that our method performs very competitively with respect to state-of-the-art methods also in challenging images where astrocytes are clustered together. Our code is released open source and freely available to the scientific community.

Improved detection of soma location and morphology in fluorescence microscopy images of neurons.

BACKGROUND: Automated detection and segmentation of somas in fluorescent images of neurons is a major goal in quantitative studies of neuronal networks, including applications of high-content-screenings where it is required to quantify multiple morphological properties of neurons. Despite recent advances in image processing targeted to neurobiological applications, existing algorithms of soma detection are often unreliable, especially when processing fluorescence image stacks of neuronal cultures. NEW METHOD: In this paper, we introduce an innovative algorithm for the detection and extraction of somas in fluorescent images of networks of cultured neurons where somas and other structures exist in the same fluorescent channel. Our method relies on a new geometrical descriptor called Directional Ratio and a collection of multiscale orientable filters to quantify the level of local isotropy in an image. To optimize the application of this approach, we introduce a new construction of multiscale anisotropic filters that is implemented by separable convolution. RESULTS: Extensive numerical experiments using 2D and 3D confocal images show that our automated algorithm reliably detects somas, accurately segments them, and separates contiguous ones. COMPARISON WITH EXISTING METHODS: We include a detailed comparison with state-of-the-art existing methods to demonstrate that our algorithm is extremely competitive in terms of accuracy, reliability and computational efficiency. CONCLUSIONS: Our algorithm will facilitate the development of automated platforms for high content neuron image processing. A Matlab code is released open-source and freely available to the scientific community.