TABLE 2.

Main soma quantification methods for mesoscale connectivity.

Soma quantification method	Sample preparation/microscopy	Principle
Kim et al. (2015)	DAPI stained nuclei	Intensity threshold
Frasconi et al. (2014)	GFP transgenic mice	Adaptation of intensity threshold: mean shift clustering to detect soma centers, supervised semantic deconvolution by means of neural networks for image enhancement and manifold learning for filtering false positives
ClearMap	YFP	Nuclei detection with background subtraction, filters, morphological operations, and 3D peak detection, followed by watershed segmentation and volume-based filtering to identify cells.
MIRACL Pipeline	YFP + DTI registering	Segmentation workflow in ImageJ, utilizing optimized pre-processing, morphological analysis algorithms, and a parallelized feature extraction algorithm for 3D cellular features.
CellPose (Stringer et al., 2021; Oltmer et al., 2023)	Light microscopy, HE stained histopathological images	A simulated diffusion process generates spatial gradients pointing toward the center of a cell, and a neural network trained on these gradients, along with pixel categorization, forms a gradient vector field used to predict masks by constructing a dynamical system with fixed points.
Suite2p (Pachitariu et al., 2017)	Two-photon calcium images	Greedy segmentation of nearby pixels
Hu et al. (2021)	Nissl stained	Modified 3D fully connected Unet
Wei et al. (2023)	fMOST	Lightweight neural network for quick soma detection in low resolution, followed by a network with multi-scale context and a module for precise soma localization.

DAPI, 4′,6-diamidino-2-phenylindole; GFP, green fluorescent protein; YFP, yellow fluorescent protein; DTI, diffusion tensor imaging; fMOST, fluorescence micro-optical sectioning tomography; 3D, three-dimensional; HE, Hematoxylin and eosin stain.