Automated Individualization of Size-varying and Touching Neurons in Macaque Cerebral Microscopic Images

In biomedical research, cell analysis is important to assess physiological and pathophysiological information. In this context, virtual microscopy offers the unique possibility to study the tissues at a cellular scale, which requires a high spatial resolution and produces tremendous as well as complex data making the automated analysis challenging. In this article, we address the problem of individualization of size-varying and touching neurons in optical microscopy 2D images. We propose a general protocol: i) after a step of segmentation of neuron class using Random Forest technique, a novel min-max filter is proposed to enhance neuron centroids and boundaries information, enabling the use of region growing process based on a contour-based model to drive it to neuron boundary and achieve individualization of touching neurons; ii) taking into account size-varying neurons, an adaptive multi-scale procedure aiming at individualizing touching neurons is proposed. This protocol has been evaluated in 17 major anatomical regions concerning three NeuN-stained macaque brain sections, presenting various and representative neuron densities. Qualitative and quantitative analyses have demonstrated that the proposed method provides satisfactory results in most regions (ex. caudate, cortex, subiculum and putamen) and outperforms a classical Watershed algorithm. Neuron counting using the proposed method has shown a high correlation with an adapted stereology counting technique performed by two experts (resp. 0.983 and 0.975 for the two experts). Applying the proposed method, neuron size is estimated between 2 and 28.6 μm in diameter which corresponds to the values estimated in the literature. Further works will aim to evaluate the effects of staining variations as well as inter-subject effects on our protocol. The proposed method is planned to be used in large-scale biological studies to improve analysis with supplementary neuron morphology and distribution information.