Examine bacterial cell images using Machine Learning techniques

December 31, 2021

Objective

The goal of this assignment was to examine a set of bacterial cell images using machine learning techniques, including feature extraction, features selection and clustering, in order to help the biologists organize similar images.


We were given a zip file containing a set of 563 PNG images taken from a super-resolution fluorescence microscopy photograph of Staphylococcus aureus, a common cause of hospital infections and often resistant to multiple antibiotics.


The images provided for this assignment were obtained by automatic segmentation and include cells in different stages of their life cicle as well as segmentation errors, not corresponding to real cells. The image below shows a sample of the images provided.
All images have the same dimensions, 50 by 50 pixels, with a black background and the segmented region centered in the image.

In addition to the images, we were given a text file that had information on the identification of the cell cycle phase of some cells.
These cells were manually labelled by biologists. The figure below illustrates examples from the 3 phases. The first phase before the cell starts to divide, the second covers the first part of the division, with the formation of a membrane ring where the cells will divide, and the third phase corresponds to the final stage of cell division. However, only some images are labelled.
In this assignment, our objective was to load all images, extract features, examine them and select a subset for clustering with the goal of reaching some conclusion about the best way of grouping these images.

Implementation

TODO