AI and Blockchain-assisted diagnostics in resource-limited setting

Guo, Xin (2024) AI and Blockchain-assisted diagnostics in resource-limited setting. PhD thesis, University of Glasgow.

Abstract

Diseases, including communicable and noncommunicable diseases, have been one of the major causes of human morbidity and mortality since the beginning of our history. Although many diseases have become treatable or preventable, thanks to interventions including pharmaceutical and technological advances, many people die each year in developing countries and remote rural areas due to limited (or even no) access to medical facilities and expertise. An accurate, rapid, and reliable diagnostic test is vital to improved disease treatment and prevention. However, running diagnostic tests usually requires complex, expensive instruments, professionally trained operators, and a stable power supply. Unfortunately, these resources are generally limited or unavailable in many low-resource settings.

Although there are countless limitations in running diagnostic tests in low-resource settings, various endeavours have been made to overcome the existing obstacles. One of the most important advances has been the development of point-of-care or point-of-need tests. These diagnostic assays can be delivered in convenient formats and have successfully reduced the cost of running diagnostics, so playing an essential role in disease management and lifesaving in low-income countries.

One key aspect of diagnosis may be the interpretation of the test, which can either be done by an expert in the field or by communicating that data to a remote expert or a “smart” system to interpret the data. Accurately interpreting the test outcome can help the patients receive appropriate treatment timely. However, issues presented in data management during such communication, such as tampered and counterfeited test results and unsecured data sharing between end users (patients) and professionals (doctors, healthcare workers, researchers, etc.). Also, problems like unreliable electricity supply and internet connection were found during the field study conducted by our group previously, and those issues can also delay the diagnosis of the disease.

In this PhD study, an AI-assisted platform for DNA-based malaria diagnostic tests was developed and tested in the field. This platform allows users to run a test with a low-cost portable heater and record the test information with an Android phone. It can be used to run LAMP-based malaria tests with a portable heater and read the test results automatically with 97.8% accuracy. And it only takes around 20 milliseconds to classify one image on an inexpensive (~£100) Android phone. When the internet connection is available, the test information can be safely kept in a Blockchain network for future use to inform treatment or surveillance activities.

Expertise developed in the deep neural network was also used to train algorithms for the diagnosis of retinopathies, involving developing methods for retina vessel segmentation and classification, which explores the possibility of applying AI to diagnostics in low-resource settings. In such settings, accessing medical expertise can be challenging. It has been found that using only a convolutional neural network is not sufficient in identifying arteries and veins. Models were trained for performing vessel segmentation and classification tasks; for segmenting vessels from the background achieved over 95% accuracy and over 0.8 mean average over the union score (MIoU) on the DRIVE dataset, while for A/V classification tasks, the MIoU decreased to less than 0.7. However, combining it with the traditional approach has the potential to achieve good performance.

In addition, research was conducted on the utilisation of digital technologies to assist other researchers and engage with the public. To assist researchers in determining the minimum required sample size, a web-based calculator was developed during the COVID-19 pandemic. Furthermore, a website was created containing 360-degree images to help individuals comprehend the challenges of diagnostics and healthcare in developing regions and to raise awareness about how infectious diseases spread.

• Item Type: Thesis (PhD)
• Qualification Level: Doctoral
• Subjects: Q Science > QA Mathematics > QA76 Computer software; R Medicine > RA Public aspects of medicine > RA0421 Public health. Hygiene. Preventive Medicine; R Medicine > RB Pathology; T Technology > T Technology (General)
• Colleges/Schools: College of Science and Engineering > School of Engineering
• Supervisor's Name: Cooper, Professor Jonathan and Reboud, Professor Julien
• Date of Award: 2024
• Depositing User: Theses Team
• Unique ID: glathesis:2024-84174
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 28 Mar 2024 14:20
• Last Modified: 28 Mar 2024 14:40
• Thesis DOI: 10.5525/gla.thesis.84174
• URI: https://theses.gla.ac.uk/id/eprint/84174

Related URLs

• Data DOI
• Article DOI