Computational Approaches For The Automatic Quantification of Cells From Brain Images

Microglia are glial cells residing in the central nervous system (CNS). They represent the first line of immune defence within the CNS and are responsible for fundamental physiological and pathological processes. Given their importance, the quantification of these cells is fundamental in a clinical context. However, this process is a major challenge, as conventional cell counting involves a specific set of tools and devices, being extremely costly. Currently, most cell-counting processes are manual. Such processes are time-consuming, tedious, and imprecise, being heavily dependent on the operator. To address this, new approaches have been developed to improve the quantification process. Indeed, an automated solution can greatly increase the standardised process as it shows better accuracy and efficiency. In this work, we compare and demonstrate that, on the one hand, we have classical computational approaches, where software and assistants for automatic cell counting, such as ImageJ, are applied in images that contain scattered cells. On the other hand, deep learning approaches show similar accuracy to the manual counting process but present a significant enhancement in reproducibility and efficiency.