DeepD3, an Open Framework for Automated Quantification of Dendritic Spines

Abstract

Dendritic spines are the seat of most excitatory synapses in the brain, and a cellular structure considered central to learning, memory, and activity-dependent plasticity. The quantification of dendritic spines from light microscopy data is usually performed by humans in a painstaking and error-prone process. We found that human-to-human variability is substantial (inter-rater reliability 82.2 ± 6.4 %), raising concerns about the reproducibility of experiments and the validity of using human-annotated 'ground truth' as an evaluation method for computational approaches of spine identification. To address this, we present DeepD3, an open deep learning-based framework to robustly quantify dendritic spines in microscopy data in a fully automated fashion. DeepD3's neural networks have been trained on data from different sources and experimental conditions, annotated and segmented by multiple experts and they offer precise quantification of dendrites and dendritic spines. Importantly, these networks were validated in a number of datasets on varying acquisition modalities, species, anatomical locations and fluorescent indicators. The entire DeepD3 open framework, including the fully segmented training data, a benchmark that multiple experts have annotated, and the DeepD3 model zoo is fully available, addressing the lack of openly available datasets of dendritic spines while offering a ready-to-use, flexible, transparent, and reproducible spine quantification method.

Implicit neural representations in light microscopy

“…Dendritic spine quantification is, in general, an error-prone process with high inter-rater reliability (82.2 6.4%) [ 17 ]. We have determined, that the number of spines quantified in our final predictions depends on the noise level of the training data of our denoiser network, which then influences the number of identified spines.…”

“…We evaluated the spine recovery with the DeepD3 framework, a dedicated tool for the detection of dendritic spines and dendrites [ 17 ]. DeepD3 assigns a probability to every pixel for being a spine or dendrite respectively.…”

RESPAN: an accurate, unbiased and automated pipeline for analysis of dendritic morphology and dendritic spine mapping

“…This addresses challenges presented by other tools that require Python or MATLAB scripts, including tools that have been incorporated into our approach. Furthermore, unlike other methods that limit analysis outputs to dendritic spine coordinates 17 , or require the use of multiple programs not within the same environment 17 , our approach provides comprehensive 3D analysis within a single graphical user interface environment. RESPAN can also be readily adapted to new research questions using Fiji, a widely used image analysis software, and the tools within RESPAN (Fiji: doi:10.1038/nmeth.2019).…”

Spot Spine, a freely available ImageJ plugin for 3D detection and morphological analysis of dendritic spines

“…More recent applications include 3dSpAn (Das et al, 2022) and a spine detector plugin for the Vaa3D software (Iascone et al, 2020), as well as tools dedicated to two-photon microscopy (Singh et al, 2017;Rada et al, 2018;Argunşah et al, 2022;Vogel et al, 2023) and structured illumination microscopy (Kashiwagi et al, 2019). Machine learning based method has also been developed to identify dendritic spines (Blumer et al, 2015;Smirnov et al, 2018;Guerra et al, 2023), and deep learning approach was used to provide the automated spine segmentation softwares DeepSpineNet and DeepD3 (Vidaurre-Gallart et al, 2022;Fernholz et al, 2024). Herein, we present an ImageJ plugin that allows spine detection, spine heads segmentation and spine necks tracing.…”