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

Dong Li,Hong Xie,Rene Werner,Yi Hu,Guangyu Wang,Ji-Song Guan,Claus C Hilgetag,Melissa Zavaglia

doi:10.1038/s41598-019-43432-y

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.

Highlights

The mammalian cerebral cortex is not thick, its laminar structure is remarkably complex[1,2,3,4,5]
Machine learning is a very promising approach in this regard, and existing findings of studies on conventional layer labelling and laminar properties of cortical neurons, such as morphological[31], physical[32,33,34,35,36], or functional[37] properties of the neurons, already provide a range of well-characterized features that can be employed for machine learning-based laminar pattern discrimination
These examples illustrate the potential of the proposed clustering approach to allow deeper insight into the layer structure and its hierarchical organization when compared to the classical approach of manual labelling

Summary

Introduction

The mammalian cerebral cortex is not thick, its laminar structure is remarkably complex[1,2,3,4,5]. The classical definition of laminar layers and their hierarchy contributed much to cortical research, by grouping neurons for comparable data analysis and computational modeling, it sometimes appears overly restrictive. It suggests that, in terms of hierarchical organization, Brodmann’s six layers can be considered to reside on the same hierarchy level, and further sub-layers on the levels below. Machine learning is a very promising approach in this regard, and existing findings of studies on conventional layer labelling and laminar properties of cortical neurons, such as morphological[31], physical[32,33,34,35,36], or functional[37] properties of the neurons, already provide a range of well-characterized features that can be employed for machine learning-based laminar pattern discrimination

Methods

Results

Discussion

Conclusion