Discrimination of the hierarchical structure of cortical layers in 2-photon microscopy data by combined unsupervised and supervised machine learning

Abstract

The laminar organization of the cerebral cortex is a fundamental characteristic of the brain, with essential implications for cortical function. Due to the rapidly growing amount of high-resolution brain imaging data, a great demand arises for automated and flexible methods for discriminating the laminar texture of the cortex. Here, we propose a combined approach of unsupervised and supervised machine learning to discriminate the hierarchical cortical laminar organization in high-resolution 2-photon microscopic neural image data of mouse brain without observer bias, that is, without the prerequisite of manually labeled training data. For local cortical foci, we modify an unsupervised clustering approach to identify and represent the laminar cortical structure. Subsequently, supervised machine learning is applied to transfer the resulting layer labels across different locations and image data, to ensure the existence of a consistent layer label system. By using neurobiologically meaningful features, the discrimination results are shown to be consistent with the layer classification of the classical Brodmann scheme, and provide additional insight into the structure of the cerebral cortex and its hierarchical organization. Thus, our work paves a new way for studying the anatomical organization of the cerebral cortex, and potentially its functional organization.

Multimodal memory components and their long-term dynamics identified in cortical layers II/III but not layer Vb

“…In addition, we exclude layer Va from the analysis because layer Va is not very easy to discriminate in some locations and more importantly it has been shown in previous studies that the response properties to external stimuli in layer Va can be significantly different from layers Vb [17], indicating their functional difference in the process of learning. Last but not least, layer II and layer III may also have functional difference, but it is not easy to distinguish between them basing on the cytoarchitecture architecture [18], so that we had better to analyse them as one layer compartment for a compromise.…”

Multimodal Memory Components and Their Long-Term Dynamics Identified in Cortical Layers II/III but Not Layer V

“…First, layer II and layer III are not easy to discriminate based on their cytoarchitecture as obtained in the protein expression data set (Li et al, 2019); thus, we had to analyze them as one joint laminar compartment. There may be some differences in terms of the long-term memory-dependent dynamics between these layers, but we have to leave this problem to future studies.…”

Engineering Tissue Fabrication With Machine Intelligence: Generating a Blueprint for Regeneration

“…Magnetic resonance imaging (MRI) is one of the most popular methods used for TE (Jackson et al, 2017). FTIR and Raman microspectroscopy are underexplored techniques in TE, presumably due to their low-resolution nature, but the resulting molecular vibrational or rotational modes can be used as a biochemical fingerprint to characterize tissues (Chen et al, 2012;LeCun et al, 2015;Schmidhuber, 2015;Albro et al, 2018;Gupta et al, 2019;Li et al, 2019;Marzi et al, 2019). Mass spectrometry is used for spatial localization of collagen and elastin in tissue samples (Angel et al, 2018).…”

“…Pattern discovery can be grouped into (i) ML methods directly targeting imaging data (Brent and Boucheron, 2018;Casiraghi et al, 2018;Gupta et al, 2019;Kistenev et al, 2019;Li et al, 2019;Rivenson et al, 2019;Vu et al, 2019), (ii) ML-based predictive modeling for TE scaffolds (Buggenthin et al, 2017;Tanaka et al, 2017;Chaudhury et al, 2018;Nitta et al, 2018;Marzi et al, 2019;Waisman et al, 2019), and (iii) a broad range of bioinformatics such as network analysis (Camacho et al, 2018). Specifically, several studies are (i) predicting tissue properties with DL from images or experimental observations (Liang et al, 2017;Brent and Boucheron, 2018;Kusumoto et al, 2018;Berisha et al, 2019;Gupta et al, 2019;Kistenev et al, 2019;Lutnick et al, 2019;Rivenson et al, 2019;Vu et al, 2019;Xie et al, 2019), (ii) classifying tissue type, state, and material properties with various ML methods (Casiraghi et al, 2018;Hailstone et al, 2018;Li et al, 2019), (iii) integrating multiple imaging platforms and experiments (Heredia-Juesas et al, 2018), (iv) modeling tissues for pattern discovery and predictive modeling (Bilgin et al, 2010;Yener, 2016;Kusumoto et al, 2018), and (v) extracting information from images for TE (Gholami et al, 2018).…”