###Optimal Microscopy Slide Navigation with Reinforcement Learning for High Resolution Cell Morphology Analysis
Project ID: 2228bd1164 (You will need this ID for your application)
Research Theme: Artificial Intelligence and Robotics
UCL Lead department: Computer Science
Lead Supervisor: Petru Manescu
Project Summary:
Together with molecular biology, image-based phenotyping of cells is at the heart of biomedicine and healthcare. Capturing the most informative regions of a biological sample on a glass slide under a microscope at high magnifications remains a bottleneck for high-throughput quantitative cell morphology analysis. This project aims to overcome this challenge through an artificial intelligent microscope scanning system for optimal sampling of biospecimens. More precisely, you will investigate the use of Deep Reinforcement Learning (DRL) techniques to control our digital microscope platform and locate areas of interest on a glass microscopy slide in an efficient and robust manner amenable to be translated to the clinic. In this way, manual operation and calibration of the microscope or expensive impractical solutions can be avoided paving the way to high-throughput high-resolution cell morphology analysis as well as paving the way to current research on mapping genetic mutation to a wide range of clinically relevant cell-morphologies. We are looking for an enthusiastic candidate with knowledge in reinforcement learning ready to take on a multi-disciplinary project. In a first instance, you will design, train, and evaluate a series of DRL models able to navigate virtual whole slides of digitized peripheral blood smears and rapidly locate image fields containing monolayers of separated cells suitable for morphological analysis. The goal here is to design and identify the best RL navigation strategy that best mimics an expert human microscopist. Next, the candidate will deploy, synchronise, and fine-tune these DRL models in a real digital microscope scanning system. The system will be evaluated against a human agent. Based in the UCL Computer Science department, you will be engaging with experts in robotics, optical microscopy, artificial intelligence, and clinical decision making.