Smart Region-Growing: a novel algorithm for the segmentation of 3D clarified confocal image stacks

MotivationAccurately mapping the brain at the micro-scale is still a challenge in cellular neuroscience. While notable success has been reached in the field of tissue clarification and confocal imaging to obtain high-fidelity maps of three-dimensional neuron organization, neuron segmentation is still far away of ground-truth and manual segmentation performed by experts may be needed. The need of an expert is in part related to the limited success of the algorithms and tools performing single-neuron segmentation from 3D microscopic image data available in the State of Art, in part to the non-complete information given by these methods, which typically perform neuron tracing and thus limit the interpret-ability of results.ResultsIn this paper, a novel algorithm for segmenting single neurons in their own arrangement within the brain is presented. The algorithm performs a region growing procedure with local thresholds based on the pixel intensity statistics typical of confocal acquisitions of biological samples and described by a mixture model. The algorithm is developed and tested on 3D confocal datasets obtained from clarified tissues. We compare the result of our algorithm with those obtained by manual segmentation performed by 6 different experts in terms of neuron surface area, volume and Sholl profiles. Statistical analysis performed on morphologic features extracted from the segmented structures confirms the feasibility of our approach.AvailabilityThe Smart Region Growing (SmRG) algorithm used in the analysis along with test confocal image stacks is available on request to the [email protected] informationSupplementary data are available on request to the authors.

Improved deconvolution of very weak confocal signals

Deconvolution is typically used to sharpen fluorescence images, but when the signal-to-noise ratio is low, the primary benefit is reduced noise and a smoother appearance of the fluorescent structures. 3D time-lapse (4D) confocal image sets can be improved by deconvolution. However, when the confocal signals are very weak, the popular Huygens deconvolution software erases fluorescent structures that are clearly visible in the raw data. We find that this problem can be avoided by prefiltering the optical sections with a Gaussian blur. Analysis of real and simulated data indicates that the Gaussian blur prefilter preserves meaningful signals while enabling removal of background noise. This approach is very simple, and it allows Huygens to be used with 4D imaging conditions that minimize photodamage.