MIT CINCS (Communications Information Networks Circuits and Signals) / Hamilton Institute Seminar
Speaker: Dr George Adams, Imperial College London
Title: "Application of deep learning and spatial statistics within 3D microscopy"
Abstract: The bone marrow is semi-solid organ present within the majority of larger bones within the human body. As the source of all blood within mammals it plays a key role in human health. Despite making up less than 5% of the human body by weight, the bone marrow produces 10 times more cells than all the other organs of body combined. In a healthy adult, this corresponds to approximately a trillion cells each day.
Compared to many other organs, our understanding of the physiology and cellular organization of the bone marrow remains poor. One of the main reasons for this is because historically imaging of the bone marrow has been extremely challenging. With recent advances in microscopy, immuno-histochemistry and tissue processing it is now possible to clarify bone marrow structures, turning them ‘invisible’, allowing light to pass through dense structures and excite the fluorophores of choice, allowing cells and structures of interest to be detected. Visualization is further improved using transgenic technology which allows researchers to breed mice which express fluorescent proteins within specific bone marrow cell and structures. With these developments, we can now image large portions of the bone marrow, and in some cases the entire bone marrow structure. These images containing millions of cells and provide a unique insight into bone marrow biology.
With these developments in imaging technology comes a new set of problems. The extraction, classification and analysis of the vast amount of cellular data present within these large image files is a huge challenge. This talk will describe how deep learning algorithms are being utilized to identify and classify cells within these microscopy images. Modern techniques for 3D image segmentation will also be discussed along with their advances and limitations. Finally, it will be described how spatial statistical models can be adapted for use within a 3D bone marrow space and how these models can be used to explore the changes that take place within the bone marrow during infection.
Bio: Dr George Adams is a clinical haematologist and research fellow in the department of Life Sciences at Imperial College London, UK. Prior to studying medicine, he completed an undergraduate and master’s degree in biochemistry at Imperial College London, specializing in the area of biophysics. He received his medical degree from the University of Bristol and undertook specialist physician and haematology training in London. As a haematologist he has dual certification in both medicine and pathology and is member of Royal College of Physicians (UK) and of the Royal College of Pathologists. He completed 2 additional master’s degrees in Epidemiology (Imperial College London) and Statistics (University of Glasgow). He was awarded the SAS prize from the University of Glasgow as the top statistics graduate within the department of mathematics. He was awarded an academic fellowship at Imperial College London in 2015 and in 2018 was awarded a prestigious Wellcome Trust 4i (Immunity, Inflammation, Infection and Informatics) research clinical fellowship which he undertook within the laboratory of Professor Cristina Lo Celso (Imperial College London). He has received research awards from both the British Society of Haematology (BSH) and the American Society of Hematology (ASH). The focus of his research in Professor Lo Celso’s lab is the development of 3D bone marrow imaging and the use of deep-learning neural networks within 3D microscopy imaging. He has published in journals such as Science Translational Medicine, Bone Marrow Transplantation and Journal of Thrombosis and Thrombolysis.