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

Abstract

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.The human brain is a complex network of over 2 × 10 11 neural cells comprising neurons and glial cells. Neuroglia comprise astrocytes, oligodendrocytes, NG2 glia, microglia and all peripheral glia 1 . Astrocytes, a subtype of glial cells with a complex star-shaped morphology, are the most abundant cells in most part of the human brain 2 . Although they were long characterized as supportive cells only involved in maintaining neuron homeostasis and survival, a number of studies during the last decade have revealed that astrocytes play an active role in fundamental brain processes underlying neuronal development and function. Several studies have implicated astrocytes in controlling the development of the nervous system through axon guidance and synaptogenesis. During neural circuit development, astrocytes are responsible for the regularization of the neuronal network by pruning abnormal or dysfunctioning synapses 3 . In addition, together with the pre-and post-synaptic parts of two neuronal synapses, astrocytes can form a so-called "tripartite" synapse that helps modulate synaptic transmission via the release of neurotransmitters such as glutamate and ATP 4,5 . The critical roles of astrocytes in neuronal development and connectivity make them among the most promising targets for innovative therapies designed to treat a range of brain disorders or neurological injuries 6 . As a result, the attention on astrocytes has dramatically increased in recent years.Astrocytes have been shown to reflect their diverse abilities and functions on their special structural design, and alterations in astrocyte morphology are known to correlate to traumatic brain injury, infection, ischemia, autoimmune responses, and neurodegenerative diseases 5,7,8 . For instance, their intricate ar...

Astrocyte regional heterogeneity revealed through machine learning‐based glial neuroanatomical assays

“…We suggest that using Otsu threshold, or other computationally basic segmentation methods depending on the type of cell being studied, as a bridging solution to time consuming manual ground truth segmentation. The deep learning astrocyte segmentation neural net developed by Kayasandik et al had a high accuracy based on only 118 128 × 128 pixel images, however it is impractical to increase the training data set due to how time consuming it is to manually generate ground truth images (Kayasandik et al, 2020). We propose as a solution, it takes significantly less time to apply thresholding methods to automatically generate ground truths.…”

“…In addition, this technique is successful in generating data for arborization quantification, it fails to include the entire astrocyte data for pixel segmentation. In recent years, astrocyte analysis has become more advanced using convolutional neural networks for counting the number of astrocytes and segmenting (Kayasandik et al, 2020;Suleymanova et al, 2018). However, cell counting fails to elucidate the morphological changes astrocytes undergo following different disease states (Kayasandik et al, 2020 (Kayasandik et al, 2020).…”

Novel Tools and Investigative Approaches for the Study of Oligodendrocyte Precursor Cells (NG2-Glia) in CNS Development and Disease

“…Of course, histological sectioning results in the loss of important volumetric information, which can otherwise more accurately inform on important parameters such as cell size, polarity, branch number, branch length, and the number of synapses and interacting cells (Hillman, 2000 ). Additionally, the categorization and measurement of relevant OPC parameters could be further enhanced by the use of computational techniques, such as “deep learning” which, in the case of studies concerning astrocytes, have already been used to characterize morphological parameters in a relatively unbiased manner (Kayasandik et al, 2020 ). It follows that investigations regarding OPC morphology may benefit from such computational techniques (Kayasandik et al, 2020 ), possibly highlighting previously unseen characteristics of OPCs which may provide functional insights.…”

“…Additionally, the categorization and measurement of relevant OPC parameters could be further enhanced by the use of computational techniques, such as “deep learning” which, in the case of studies concerning astrocytes, have already been used to characterize morphological parameters in a relatively unbiased manner (Kayasandik et al, 2020 ). It follows that investigations regarding OPC morphology may benefit from such computational techniques (Kayasandik et al, 2020 ), possibly highlighting previously unseen characteristics of OPCs which may provide functional insights.…”

Automated Morphometric Analysis Reveals Plasticity Induced by Chronic Antidepressant Treatment in Hippocampal Astrocytes

“…These steps were performed manually using ImageJ. However, these steps can also be automated using machine learning tools published elsewhere 36,37 . Images of cells from 2 different experimental groups were saved into 2 separate folders named appropriately.…”