Analyzing white blood cells in blood samples using deep learning techniques

From ISLAB/CAISR
Title Analyzing white blood cells in blood samples using deep learning techniques
Summary To analyze white blood cell content in blood samples using deep learning techniques.
Keywords
TimeFrame Fall 2019
References
Prerequisites Good knowledge of machine learning, convolutional neural networks and programming skills for implementing machine learning algorithms
Author
Supervisor Mattias Ohlsson
Level Master
Status Open

Generate PDF template

To measure the amount of white blood cells in blood samples a combination of image analysis and machine learning can be used. This is the case for the instrument WBC-Diff manufactured by HemoCue. It uses a neural network classifier based on manually calculated features from images of blood cells. The WBC-Diff instrument can also detect subtypes of white blood cells.

The aim of this project is to investigate if other machine learning approaches, such as convolutional neural networks, can improve their current state-of-the-art when analyzing blood cells in blood samples. Specific questions to study include:

1. Prediction of blood cell or not a blood cell

2. White blood cell type prediction (5 different types)

The training material consists of 2000-5000 images per class with a resolution of 48x48 pixels in approximately 40 images slices.

This master project is a collaboration with the company HemoCue. HemoCue will provide domain knowledge and all the necessary image material.