James Pipe

Thomas Chenevert

Meike Vernooij

Many neurological diseases, especially those occurring at older age, have a long subclinical phase during which a person is asymptomatic and does not seek medical attention. As a consequence, once symptoms manifest, in many instances the pathologic changes caused by the disease process are already advanced and mostly irreversible. To study disease in the asymptomatic stage, population-based studies are of great importance. Medical imaging applied in these studies, or ‘population imaging’ can, non- or minimally-invasively, show the changes that occur in the human body that may reflect either early disease, intermediate factors or risk indicators of disease.

Susumu Mori

In this presentation, the basic concept of anatomical segmentation is explained. This presentation first discusses about the importance of tissue segmentation for modern medical data analysis, in which difficult neurological conditions are often the target of the research. Then different types of segmentation approaches are explained. In the last section, an interesting paradox of the tissue segmentation, namely the lack of ground truth, is discussed in detail. This presentation should be informative for both method developers and users.

Shingo Kakeda

The role of brain MRI in diagnosis of multiple sclerosis (MS) and vascular lesions is well established, and the recently developed MR techniques, including synthetic MRI, myelin map with q-Space diffusion MRI, susceptibility-weighted imaging (SWI) and quantitative susceptibility mapping (QSM), further improve the diagnostic value in a research and clinical routine setting. This course will introduce the recent data pertaining to the use of new MR techniques in assessing MS lesion and small vascular lesions (cerebral microinfarcts).

Yong He

Masayuki Ohzeki

We introduce a concept of the generative adversarial network and consider its possibility of medical application.

 

Bradley Erickson

This session will describe the basic concepts of machine learning, both traditional machine learning and deep learning. Particular emphasis will be placed on how ML methods can give results that appear good, but may not be correct or representative of real world performance.