Bell Jar: A Semi-Automated Registration and Cell Counting Tool for Mouse Neurohistology Analysis

Abstract

To investigate the anatomical organization of neural circuits across the whole brain, it is essential to register the experimental brain tissues to a reference atlas accurately. This procedure is also a prerequisite to quantify the locations and numbers of cells of interest in specific regions. However, it remains challenging to do registration on experimental tissue due to the intrinsic variation among the specimens, tissue deformation introduced by histological processing, and the potential inconsistency in the judgment of the experimenter during manual annotation. Here, we introduce Bell Jar, a multi-platform analysis tool with semi-automated affine warping of atlas maps onto microscopic images of brain slices and machine learning-based cell detection. Bell Jars intuitive GUI and internal dependency management enable users of all skill levels to obtain accurate results without programming expertise. To compare the performance of Bell Jar with previously published methods1-4, we labeled neurons in the mouse visual cortex with either an engineered rabies virus or an adeno-associated virus (AAV) for neural circuit tracing5, and quantified Bell Jars performance at each step of the pipeline for image alignment, segmentation, and cell counting. We demonstrated that Bell Jars output is as reliable as manual counting by an expert; it is more accurate than currently available techniques even with noisy data and takes less time with fewer user interventions. Bell Jar is an easy-to-navigate application that provides a reproducible, automated analysis workflow to facilitate the precise mapping of histological images of the mouse brain to the reference atlas and the quantification of cellular signals it is trained to recognize.