Plenary Lecture
Deep Learning Can Detect Important Genomic Properties of Tumors from Routine MR Imaging
Dr. Bradley Erickson
Mayo Clinic
USA
E-mail: bje@mayo.edu
Abstract: There has been a long-standing interest in applying computers to make diagnoses from medical images. Despite substantial investment and effort, the products have largely not been successful, achieving performance that is at or below that of experts. Those approaches involved computing features thought to be important, and then applying machine learning methods to find which were valuable, and the appropriate weighting. Deep Learning has become quite popular in the non-medical image interpretation field, and has produced dramatically improved results on public competitions. These do not require computation of suspected important features, but simply take images and ‘answers’ to train. We have begun to apply this approach to medical images and found that in very short time, deep learning can find tumor genomic properties, tumor response, and even image segmentation tasks that is superior to traditional approaches. In this talk, these results, and the basis for them will be explained, and a discussion of why deep learning will likely be valuable for many challenges in medicine.
Brief Biography of the Speaker: Bradley J Erickson, MD PhD, received his MD and PhD degrees from Mayo Clinic. He went on to be trained in radiology, and then a Neuroradiology fellowship at Mayo, and has been on staff at Mayo for 20 years. He does clinical Neuroradiology, has chaired of the Radiology Informatics Division and is Associate Chair for Research. He has been vice chair of Information Technology for Mayo Clinic. He has been awarded multiple external grants, including NIH grants on MS, brain tumors, and medical image processing. He is a former president of the Society of Imaging Informatics in Medicine and serves on the Board of Directors for the American Board of Imaging Informatics and the Board of the IHE USA. He is organizer of a new meeting: Conference on Machine Intelligence in Medical Imaging. His particular focus in on the use of image processing and machine learning to improve of understanding and ability to diagnose these diseases.