Applying machine learning algorithms on multi-sensor applications

Abstract

IMO-IMOMEC (standing for: Institute for Materials Research - Institute for Materials Research in MicroElectronics) at Hasselt researches the possibility to implement electronics in the medical sector. These electronics are accompanied by logic and machine learning. The goal of this thesis is to do a feasibility study on the implementation of machine learning in medical applications, with the necessary steps to achieve this.

This thesis studies the machine learning algorithms by using prediction- and classification algorithms. The prediction projects include a study on the growth of yeast cells and the transition between two different liquids in a pipeline, both are accomplished with the help of impedance measuring techniques. The classification project implements tomography on wounds and classifies these results afterwards based on their intensity.

The prediction projects are both realised with a double exponential-, logarithmic- and power law regression. The matlab library EIDORS made tomography possible with help of image processing, many electrode models have been tested to determine the most efficient set-up. The classification exists out of a neural network with three hidden layers that classifies the tomographic images of the wounds.