The project aims to improve the accuracy, consistency, efficiency, and ergonomics of asbestos analysis while maintaining the critical role of trained human analysts in the decision-making process.

The first strand focuses on image digitization, optimization, and annotation storage. Traditionally, asbestos fibre counting is undertaken through direct observation using approved light microscopes, requiring analysts to spend extended periods viewing samples through microscope eyepieces. This can result in fatigue, eye strain, and musculoskeletal discomfort, all of which may influence counting performance over time. The project therefore seeks to capture microscope images digitally and present them on-screen, providing analysts with a larger and more comfortable viewing area while expanding the effective field of analysis.

To achieve this, suitable hardware will be identified to capture high-quality images from approved microscopes. Software solutions will then be evaluated for their ability to convert these images into digital formats suitable for future image comparison, automation, and machine learning applications. Annotation tools will also be assessed to enable efficient recording and marking of fibres, ideally through simple user actions such as mouse clicks. During this phase, comparisons between individual analysts and automated systems will be undertaken to understand variations in fibre identification and create a robust dataset for future AI training. A secure storage and sharing platform will be established to build an image repository that can support ongoing machine learning development. Finally, laboratory pilots will be designed and implemented to evaluate the practicality and effectiveness of the proposed solutions within operational environments.

The second strand focuses on fibre recognition supported by human validation. Building on the image library developed in the first strand, artificial intelligence models will be trained to recognize asbestos fibres and compare live microscope images against known examples. The objective is not to replace analysts but to provide intelligent decision support that accelerates and enhances the counting process. In practice, the system will present a real-time digital image to the analyst and automatically highlight potential fibres for review. Analysts will then be able to validate groups of identified fibres or assess individual candidates through straightforward interactions. Performance comparisons between AI systems and human analysts will provide valuable insights into reliability, consistency, and areas requiring further development. A key outcome of this strand will be the creation of a human-in-the-loop protocol that maintains professional oversight while enabling increasing levels of trusted automation.

The third strand addresses microscope automation. Fibre counting currently requires analysts to make repeated manual adjustments to microscope settings, which can be time-consuming and can introduce variability. The project will investigate automation technologies that can be integrated with approved microscopes to control image capture and microscope functions through software commands. These automated controls will work in conjunction with the AI fibre recognition system and respond to operator inputs when required. By linking automation triggers to specific imaging requirements, the system will automatically optimize microscope settings, capture images at the appropriate resolution, and support efficient fibre analysis. This will reduce repetitive manual tasks, minimize the risk of human error, improve consistency, and further reduce physical strain on laboratory staff.

Together, these three strands provide a phased pathway toward a future asbestos microscopy system that combines expert human judgment with advanced digital technologies to deliver safer, faster, and more reliable fibre counting.

If you are interested in contributing to this work or have experience or  expertise that is relevant, please contact membership@bohs.org