Centres for Doctoral Training

AI-Driven Adaptive Multiscale Imaging for Next-Generation Smart Diagnostic Microscopes

Supervisors:

Ke Yuan, MVLS / School of Cancer Studies

John Le Quesne, MVLS, School of Cancer Sciences 

Richard Bowman, College of Science and Engineering / School of Physics and Astronomy

 

PhD Project Summary

Digital pathology is revolutionising cancer diagnosis, but it faces a massive data crisis. Current microscopes ‘blindly’ scan entire tissue slides at maximum resolution, creating gigantic files that cost hospitals millions to store and transmit. However, human pathologists don't work this way; they use low magnification to survey the tissue, and only zoom in on specific, important or highly suspicious areas.

This project will build a "smart" microscope that mimics human intelligence. By integrating cutting-edge self-supervised Artificial Intelligence (ssAI) with automated optical hardware, we will develop a prototype scanner that rapidly takes a low-power overview, algorithmically identifies the most important regions, and selectively scans only those specific areas at high resolution.

Based in a highly interdisciplinary team spanning Computing Science, Physics, and Clinical Pathology, you will design the AI algorithms, test them virtually using massive datasets of lung cancer tissue imagery, and ultimately deploy them onto prototype microscope hardware. This is an opportunity to invent the next generation of medical imaging hardware, drastically reducing the carbon footprint and financial cost of digital pathology while accelerating AIdriven cancer diagnosis.