Neuron cell count with deep learning in highly dense hippocampus images

Neural cell counting is one of the ways in which damage caused by neurodegenerative diseases can be assessed, but it is not an easy task when it comes to neuronal counting in the most densely populated areas of the hip-pocampus. In this regard, this work presents a leveraged deep learning (DL) model, an innovative way to treat histological images and their correspondent ground truth information, where highly dense cell population with fuzzy cell boundaries and low image quality exist. The proposed model achieves state-of-the-art results in the neuron cell count problem for the highly dense area of DG and CA hippocampus regions, by making use of better pixel characterization which in turn also delivers a more efficient model size and reduces training time. Furthermore, we show that the proposed image treatment can be applied to other DL models and help them to obtain a 12% performance increase. Also, we demonstrate that with the proposed methodology, an innovative and reliable way to count neural cells with poor image condition in histological analysis has been carried out.

Automated summary

This study proposes an innovative deep learning model for neural cell counting in densely populated areas, achieving state-of-the-art results, and also presents an improved image treatment method that can enhance the performance of other DL models.