Meeting Banner
Abstract #2047

Application of support vector machines to multi-modal hemo-metabolic data for classification of disease severity in patients with extreme arterial steno-occlusive diseases

Spencer L. Waddle1, Sarah K. Lants2, Larry T. Davis2, Meher R. Juttukonda2, Matthew R. Fusco3, Lori C. Jordan4, and Manus J. Donahue2

1Chemical and Physical Biology Program, Vanderbilt, Nashville, TN, United States, 2Radiology and Radiological Sciences, Vanderbilt, Nashville, TN, United States, 3Neurosurgery, Vanderbilt, Nashville, TN, United States, 4Pediatrics - Division of Pediatric Neurology, Vanderbilt, Nashville, TN, United States

Traditional hemodynamic imaging approaches such as arterial spin labeling (ASL) and hypercapnic blood oxygenation level-dependent (BOLD) reactivity provide contrasts that are frequently difficult to interpret using conventional analyses in arterial steno-occlusive disease patients with extreme blood arrival and vascular reactivity delay times. We investigated applying a supervised learning procedure to exploit endovascular and vascular compliance artifacts as potential indicators of disease severity; results show that less-conventional variables which report on endovascular blood signal and delayed vascular compliance outperform conventional variables, such as mean ASL signal and BOLD signal change.

This abstract and the presentation materials are available to members only; a login is required.

Join Here