Automatic Quantification and Labeling of Cerebral Microbleeds, Oxygen Saturation and Sources of Abnormal Susceptibility

摘要

ABSTRACT The detection, localization and quantification of cerebral microbleeds (CMBs) plays an important role in diagnosing and establishing appropriate treatment plans in neurodegenerative diseases specifically in vascular dementia (VaD). To date, evaluating CMBs is time consuming, inaccurate and sometimes not possible. We propose to mitigate these problems by developing our software, “qSPIN”, that will provide fast and easy-to-use methods for: 1) automatic identification of CMBs and veins, 2) automatic quantification of CMBs, 3) automatic quantification of oxygen saturation in veins, and 4) creation of a user-friendly software for the practicing radiologist. Recent developments in MRI have provided a new means by which to study the role of CMBs and venous abnormalities in neurological diseases such as Alzheimer’s Disease (AD), VaD, stroke and traumatic brain injury (TBI). Susceptibility weighted imaging (SWI) has proven to be a powerful tool by which to detect CMBs and quantitative susceptibility mapping (QSM) can be used to measure changes in oxygen saturation. Knowing how many CMBs there are can predict the onset of VaD, determine whether anti-platelet therapy in stroke should be used, and correlate with neuropsychological outcome for patients with TBI. Our recent version of multi-echo SWI makes it possible to obtain both an arteriogram and a venogram simultaneously. Oxygen saturation can also be used to monitor perfusion changes and extend the window of treatment in stroke. Currently, most radiologists and technologists do not have time to perform such detailed quantitative processing and thus it is not being done clinically. Our qSPIN software will provide this quantitative data. With the number, size, and location of CMBs or venous abnormalities, a better diagnosis would be possible. Our group is uniquely positioned to address this problem having developed many of these techniques. The novelty of our approach is the marriage of SWI, QSM, STAGE and deep learning techniques to detect these vascular and functional abnormalities. To accomplish the goals of this proposal, we will develop user friendly software that incorporates all imaging information from SWI and QSM to label CMBs. We will also provide the location of the CMBs in Talairach coordinates using a template dataset. In the end, we will have a complete picture of the prevalence of CMBs, abnormal oxygen saturation and their locations in patients with neurodegenerative disease that will improve diagnosis and potentially change their treatment.