Analysis of Deep Learning Methods for In-vitro Cell Tracking

Tracking the behaviour of cells over time can be used as a ground basis for developing models and methods for estimating the behaviour of a cell given different external conditions. In this thesis, we consider automated Cell Tracking in 2D in vitro cultures. In this framework, we utilized Computer Vision methods to perform segmentation of the cell images. We design a modification of a state-of-the-art algorithm for Cell Tracking, and implement state-of-the-art Deep Learning methods for automated Cell Tracking. We test the effectiveness of the proposed methods by using timelapses provided by Rizzoli Orthopedic Institute. Notably, we show that performance sensibly depends on the dataset used to train the algorithms.