Tea Andersson and Selma Tomine Kjernsvik will present their master's thesis

Title: Seeing Clean: Automated Cleaning Assessment Using Image Analysis and Machine Learning

Author: Tea Andersson and Selma Tomine Kjernsvik

Supervisor: Anders Heyden

Examiner: Viktor Larsson

Abstract:

This thesis presents an image based method for objectively evaluating the cleaning effectiveness of
a hydrosound assisted bathing system, developed in collaboration with Arjo using their Symbliss
bath. Current assessment methods rely on visual inspection, which is subjective and difficult to
reproduce. The aim of this work was to develop a quantitative and standardised alternative based
on digital image analysis.
A silicone based skin model was developed to simulate human skin under controlled experimental
conditions. Images of Glo Germ fluorescent contamination were captured under UV illumination
before and after cleaning using a standardised camera setup. Colour and texture features were
extracted and used to train a Random Forest classifier to distinguish contaminated from cleaned
surfaces, and a continuous contamination score (0-100%) was computed to quantify the degree
of contamination. Model performance was evaluated using standard classification metrics, a
Mann-Whitney U-test, and independent expert visual assessment.
The classifier achieved a test accuracy of 81.2% and a perfect recall, with statistically significant
separation between contaminated and cleaned images (p < 0.001). Hydrosound assisted cleaning
produced a mean contamination score reduction of approximately 68.6 percentage points across
sessions, while water-only cleaning showed minimal or no reduction. The visual assessment
produced similar result, validating the models credibility.
The results suggest that image analysis can serve as an objective and reproducible tool for
evaluating cleaning performance in assisted bathing systems, with potential applications in both
product development and clinical quality assessment.