A Deep Learning-based Pipeline for Segmenting the Cerebral Cortex Laminar Structure in Histology Images

• DOI: 10.1007/s12021-024-09688-0
• Publication Date: 2024-10-17T06:02:58Z
• Authors: Jiaxuan Wang; Rui Gong; Shahrokh Heidari; Mitchell Rogers; Toshiki Tani; Hiroshi Abe; Noritaka Ichinohe; Alexander Woodward; Patrice J. Delmas
• Keywords: Cytoarchitecture

Abstract Characterizing the anatomical structure and connectivity between cortical regions is a critical step towards understanding information processing properties of brain will help provide insight into nature neurological disorders. A key feature mammalian cerebral cortex its laminar structure. Identifying these layers in neuroimaging data important for their global to understand patterns neurons brain. We studied Nissl-stained myelin-stained slice images common marmoset (Callithrix jacchus), which new world monkey that becoming increasingly popular neuroscience community as an object study. present novel computational framework first acquired labels using AI-based tools followed by trained deep learning model segment layers. obtained Euclidean distance $$\varvec{1274.750 \pm 156.400}$$ <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML"> <mml:mrow> <mml:mn>1274.750</mml:mn> <mml:mo>±</mml:mo> <mml:mn>156.400</mml:mn> </mml:mrow> </mml:math> $$\varvec{\mu m}$$ <mml:mi>μ</mml:mi> <mml:mi>m</mml:mi> acquisition, was acceptable range computing half average thickness ( $$\varvec{1800.630~ \mu <mml:mn>1800.630</mml:mn> <mml:mspace/> ). compared our layer segmentation pipeline with proposed Wagstyl et al. PLoS biology, 18 (4), e3000678 2020) adapted 2D data. better mean $$\varvec{95^{th}}$$ <mml:msup> <mml:mn>95</mml:mn> <mml:mi>th</mml:mi> </mml:msup> percentile Hausdorff (95HD) $$\varvec{92.150~ <mml:mn>92.150</mml:mn> . Whereas 95HD $$\varvec{94.170~ <mml:mn>94.170</mml:mn> from also pipeline’s performance against theirs dataset (the BigBrain dataset). The results showed quality, $$\varvec{85.318 \%}$$ <mml:mn>85.318</mml:mn> <mml:mo>%</mml:mo> Jaccard Index pipeline, while $$\varvec{83.000 <mml:mn>83.000</mml:mn> stated paper.