Logo image
A Deep Learning Approach for Neuronal Cell Body Segmentation in Neurons Expressing GCaMP using a Swin Transformer
Journal article   Open access   Peer reviewed

A Deep Learning Approach for Neuronal Cell Body Segmentation in Neurons Expressing GCaMP using a Swin Transformer

Mohammad Shafkat Islam, Pratyush Suryavanshi, Samuel M. Baule, Joseph Glykys and Stephen Baek
eNeuro, Vol.10(9), p.ENEURO.0148-23.2023
09/2023
DOI: 10.1523/ENEURO.0148-23.2023
PMCID: PMC10523838
PMID: 37704367
url
https://doi.org/10.1523/ENEURO.0148-23.2023View
Published (Version of record) Open Access

Abstract

Neuronal cell body analysis is crucial for quantifying changes in neuronal sizes under different physiological and pathological conditions. Neuronal cell body detection and segmentation mainly rely on manual or pseudo-manual annotations. Manual annotation of neuronal boundaries is time-consuming, requires human expertise, and has intra-/inter observer variances. Also, determining where the neuron’s cell body ends and where the axons and dendrites begin is taxing. We developed a deep-learning-based approach that uses a state-of-the-art shifted windows (Swin) transformer for automated, reproducible, fast, and unbiased 2D detection and segmentation of neuronal somas imaged in mouse acute brain slices by multiphoton microscopy. We tested our Swin algorithm during different experimental conditions of low and high signal fluorescence. Our algorithm achieved a mean Dice score of 0.91, a precision of 0.83, and a recall of 0.86. Compared to two different convolutional neural networks, the Swin transformer outperformed them in detecting the cell boundaries of GCamP6s expressing neurons. Thus, our Swin transform algorithm can assist in the fast and accurate segmentation of fluorescently labeled neuronal cell bodies in thick acute brain slices. Using our flexible algorithm, researchers can better study the fluctuations in neuronal soma size during physiological and pathological conditions. Significance Statement Neuronal cell body partitioning is essential for evaluating the effects of physiological and pathological conditions. Neuronal segmentation is challenging due to the complex morphological structures of neuronal cell bodies and their surroundings. Most current approaches for detecting and segmenting neurons are based on manual or pseudo-manual annotations of the neuronal boundaries by human experts. These are time-consuming and have intra-/inter-observer variability. Leveraging the current success of vision transformers for general object detection and segmentation tasks, we developed a deep-learning-based approach for automated, fast, robust 2D neuronal cell body segmentation using a state-of-the-art vision transformer (Swin transformer). This approach for neuronal cell body segmentation can assist researchers in evaluating the changes in neuronal cell body sizes under different pathological conditions.

Details

Metrics

69 Record Views
Logo image