Postdoctoral researcher in Robust Reinforcement Learning for High Throughput Computational Haematology (University College London)
Monday 3rd May 2021
The UCL Department of Computer Science is the leading computer science research department in the UK, ranked #1 in the most recent Research Excellence Framework. It carries out research, undergraduate, and postgraduate teaching across all areas of theoretical and applied computer science. Together with our partners at the Great Ormond Street Hospital (GOSH) we are building the next generation of Artificial Intelligence (AI) assisted optical microscopy which will accelerate haematology research, clinical diagnosis and treatment selection. Analysis of blood films under a light microscope at high magnification is essential to diagnosis, especially in assessing leukemias and their sub types. The level of detail needed to distinguish cell features specific to such blood pathologies obtained with high magnification objective lenses comes at the expense of a limited Field of View (FoV) and a limited Depth of Focus (DOF). Specimens mostly present agglomerations of overlapping cells and thin areas with distorted cells. As a result, only specific regions of the film, containing spatially separated blood cells at a suitable number density, are useful for morphological and textural analysis. Finding these clinically relevant regions requires visual inspection and manual operation of both the microscope x-y stage and focus, a process which is time consuming and requires the availability of a trained microscopist.
The candidate will develop a robust and generalizable state-of-the art Deep Reinforcement Learning (DRL) approach to automatically control a motorised microscope system in a way which mimics the cognitive steps of a human microscopist. More precisely, they will build a DRL able to automatically position the optimal regions of the samples under the objective, control the focus of the microscope and trigger stack acquisition. This will enable high throughput imaging essential to expedite patient diagnosis and treatment selection. The post-holder will have the opportunity to use our next generation programable microscopy platform for data acquisition and validation. The final objective is to build an “online” DRL prototype embedded in the control of the motorized stage of the digital microscope. The research fellow will equally interact with our clinical partners to validate the approach and UCL Business to explore patenting and commercialization opportunities. The position is available for one year in a first instance.
If you have any queries regarding the vacancy or the application process, please contact Professor Delmiro Fernandez-Reyes (firstname.lastname@example.org) or Dr. Petru Manescu (email@example.com).
Closing date: 19/05/2021
|Organization||University College London|
|Title||Postdoctoral researcher in Robust Reinforcement Learning for High Throughput Computational Haematology|