Deep-learning-assisted portable microscopy for disease diagnosis in low-resource settings

Higgins, Oliver (2023) Deep-learning-assisted portable microscopy for disease diagnosis in low-resource settings. PhD thesis, University of Glasgow.

Abstract

Diagnostics play a pivotal role in the global fight against communicable disease, allowing us to track the spread of outbreaks and target interventions effectively. Microscopy remains the gold-standard diagnostic for many diseases, particularly neglected tropical diseases (NTDs) such as schistosomiasis and soil transmitted helminth (STH) infections. STH infections affect more than a billion people worldwide, but primarily occur in areas with limited diagnostic coverage. Unfortunately, the application of microscopy as a diagnostic presents practical and logistical problems to health workers in low-resource settings because of the expertise, time and equipment required for processing samples. The World Health Organization (WHO) recognises the utility of microscopy as a tool to monitor and evaluate efforts towards STH elimination programmes and has highlighted the automation of the Kato-Katz (KK) faecal smear test as an achievable target for an improved STH diagnostic in the short term.

This thesis describes the development of two types of 3D-printed microscopes. The first design is a smartphone-controlled portable microscope, which is used to scan KK slides. The smartphone microscope system includes a deep-learning-based object detection algorithm which automatically detects helminth eggs in images, reducing the time and expertise required to achieve diagnosis. As part of this work, field-trials were undertaken in collaboration with the Ministry of Health Uganda, to assess the feasibility of using automated microscopy for the diagnosis of helminth infections.

The second microscope device is a lens-less microscope which uses partially coherent illumination to achieve holographic imaging. The device has the potential to offer large field-of-view (FOV) phase imaging at high resolutions, while retaining a small form factor which lends itself to transportation and use in low-resource settings. Experimental characterisation of the lens-less microscope is supplemented by a description of the algorithms used to reconstruct images and a discussion of the feasibility of image reconstruction on low-powered portable hardware.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: Q Science > QR Microbiology; R Medicine > RB Pathology; T Technology > T Technology (General); T Technology > TA Engineering (General). Civil engineering (General)
• Colleges/Schools: College of Science and Engineering > School of Engineering > Biomedical Engineering
• Supervisor's Name: Cooper, Professor Jonathan and Reboud, Professor Julien
• Date of Award: 2023
• Depositing User: Theses Team
• Unique ID: glathesis:2023-83947
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 15 Nov 2023 16:46
• Last Modified: 28 Nov 2025 16:13
• Thesis DOI: 10.5525/gla.thesis.83947
• URI: https://theses.gla.ac.uk/id/eprint/83947