Abstract:Accurate cerebral neuron segmentation is required before neuron counting and neuron morphological analysis. Numerous algorithms for neuron segmentation have been published, but they are mainly evaluated using limited subsets from a specific anatomical region, targeting neurons of clear contrast and/or neurons with similar staining intensity. It is thus unclear how these algorithms perform on cerebral neurons in diverse anatomical regions. In this article, we introduce and reliably evaluate existing machine learning algorithms using a data set of microscopy images of macaque brain. This data set highlights various anatomical regions (e.g., cortex, caudate, thalamus, claustrum, putamen, hippocampus, subiculum, lateral geniculate, globus pallidus, etc.), poor contrast, and staining intensity differences of neurons. The evaluation was performed using 10 architectures of six classic machine learning algorithms in terms of typical Recall, Precision, F‐score, aggregated Jaccard index (AJI), as well as a performance ranking of algorithms. F‐score of most of the algorithms is superior to 0.7. Deep learning algorithms facilitate generally higher F‐scores. U‐net with suitable layer depth has been evaluated to be excellent classifiers with F‐score of 0.846 and 0.837 when performing cross validation. The evaluation and analysis indicate the performance gap among algorithms in various anatomical regions and the strengths and limitations of each algorithm. The comparative result highlights at the same time the importance and difficulty of neuron segmentation and provides clues for future improvement. To the best of our knowledge, this work is the first comprehensive study for neuron segmentation in such large‐scale anatomical regions. Neuron segmentation plays a critical role in extracting cerebral information, such as neuron counting and neuron morphological analysis. Accurate automated cerebral neuron segmentation is a challenging task due to different kinds, poor contrast, staining intensity differences, and fuzzy boundaries of neurons. The comprehensive evaluation and analysis of performance among existing machine learning algorithms in diverse anatomical regions allows to make clear of the strengths and limitations of state‐of‐the‐art algorithm. The comprehensive study provides clues for future improvement and creation of automated methods.

Automated Detection of Highly Aggregated Neurons in Microscopic Images of Macaque Brain

Multiscale segmentation- and error-guided iterative convolutional neural network for cerebral neuron segmentation in microscopic images

Macaque Neuron Instance Segmentation Only with Point Annotations Based on Multiscale Fully Convolutional Regression Neural Network

Evaluation of Automated Segmentation Algorithms for Neurons in Macaque Cerebral Microscopic Images

Automated And Accurate Detection Of Soma Location And Surface Morphology In Large-Scale 3d Neuron Images

Automated identification of neurons and their locations

Accurate Automatic Segmentation of Amygdala Subnuclei and Modeling of Uncertainty via Bayesian Fully Convolutional Neural Network

DeepNFT: Towards Precise Neurofibrillary Tangle Detection Via Improving Multi-scale Feature Fusion and Adversary

DDeep3M-based Neuronal Cell Counting in 2D Large-scale Images

A Neurocomputational Method for Fully Automated 3D Dendritic Spine Detection and Segmentation of Medium-Sized Spiny Neurons

Evaluation of Deep Learning Topcoders Method for Neuron Individualization in Histological Macaque Brain Section

Automatic Neuron Detection in Calcium Imaging Data Using Convolutional Networks

Neurons Identification of Single-Photon Wide-Field Calcium Fluorescent Imaging Data

NeuroSeg-II: A Deep Learning Approach for Generalized Neuron Segmentation in Two-Photon Ca2+ Imaging

Graph-based cell pattern recognition for merging the multi-modal optical microscopic image of neurons

Automatically Segmenting and Reconstructing Neurons in SEM Images

Deep residual contextual and subpixel convolution network for automated neuronal structure segmentation in micro-connectomics

Deep learning based neuronal soma detection and counting for Alzheimer's disease analysis

Automated scalable segmentation of neurons from multispectral images

DENSELY CONNECTED LARGE KERNEL CONVOLUTIONAL NETWORK FOR SEMANTIC MEMBRANE SEGMENTATION IN MICROSCOPY IMAGES

An Efficient Method for Neuronal Tracking in Electron Microscopy Images