Identification of Mitosis Stages Using Artificial Neural Networks for 3D Time Lapse Cell Sequences

Abstract

Cells, the fundamental units of life, are central to medical research, particularly in cancer studies due to their rapid, uncontrolled division. Understanding cell behavior is crucial, with a focus on mitosis, which has distinct cell division stages. However, precise detection of these phases, especially mitosis initiation in 3D, remains an underexplored research area. Our work explores 3D cell behavior, leveraging the increasing computational capabilities and prevalence of 3D imaging techniques. We introduce diverse 3D Convolutional Neural Network (CNN) architectures such as a base 3D CNN model, 3D CNN binary model, and 3D CNN pairwise model. An ensemble model based on the 3D CNN architectures shows higher classification accuracy on two time-series datasets. This research gives better insights into understanding cell behaviour in a multidimensional manner, contributing to medical research. To the best of our understanding, we are the first to delve into the utilization of Convolutional Neural Network architectures for the 3D classification of mitosis stages.