Quantitative neuronal morphometry by supervised and unsupervised learning

Abstract

Summary We present a protocol to characterize the morphological properties of individual neurons reconstructed from microscopic imaging. We first describe a simple procedure to extract relevant morphological features from digital tracings of neural arbors. Then, we provide detailed steps on classification, clustering, and statistical analysis of the traced cells based on morphological features. We illustrate the pipeline design using specific examples from zebrafish anatomy. Our approach can be readily applied and generalized to the characterization of axonal, dendritic, or glial geometry. For complete context and scientific motivation for the studies and datasets used here, refer to Valera et al. (2021) .

A tool for mapping microglial morphology, morphOMICs, reveals brain-region and sex-dependent phenotypes

“…1a and 10a ). Due to interdependency of these parameters, we expanded the list to 27 diverse classifiers (Supplementary Table 3 ) 51 and applied bootstrap and dimensionality reduction, but even these parameters were insufficient to resolve differences in microglia morphology in CK-p25 (Fig. 6a ), 5xFAD (Extended Data Fig.…”

Large scale similarity search across digital reconstructions of neural morphology

“…This allows the programmatic ranking of similarities for thousands of reconstructions simultaneously with results in fractions of a second. Such a performance, which to our knowledge has not been previously achieved, may facilitate ever more powerful unbiased classification of neural morphology [27].…”

Assisted neuroscience knowledge extraction via machine learning applied to neural reconstruction metadata on NeuroMorpho.Org

“…Unrestricted access to mined metadata on publicly shared repositories is vital to enable reproducibility, replicability, further scientific exploration, and data-driven computational modeling [ 5 , 14 , 33 ]. Within the domain of neural morphology, recent developments include detailed statistical analyses enabled by machine learning [ 8 ] and tools for organizing large amounts of data based on arbitrary combinations of user-selected metadata [ 1 ]. More broadly, the prominence of such endeavors is continuously growing in neuroscience [ 3 ].…”