Cellular 3D-reconstruction and analysis in the human cerebral cortex using automatic serial sections

Nick Larsen et al. developed a pipeline to collect and image serial sections from fixed human cortex, then apply deep learning to detect pyramidal cells from 3D reconstructions of these sections. Their results reiterate that cortical pyramidal cells are organized in a columnar structure and highlight the potential of this method, which is universally applicable to characterize cells for various tissues. Techniques involving three-dimensional (3D) tissue structure reconstruction and analysis provide a better understanding of changes in molecules and function. We have developed AutoCUTS-LM, an automated system that allows the latest advances in 3D tissue reconstruction and cellular analysis developments using light microscopy on various tissues, including archived tissue. The workflow in this paper involved advanced tissue sampling methods of the human cerebral cortex, an automated serial section collection system, digital tissue library, cell detection using convolution neural network, 3D cell reconstruction, and advanced analysis. Our results demonstrated the detailed structure of pyramidal cells (number, volume, diameter, sphericity and orientation) and their 3D spatial organization are arranged in a columnar structure. The pipeline of these combined techniques provides a detailed analysis of tissues and cells in biology and pathology.