AUTOMATED HIPPOCAMPAL SEGMENTATION AND 3D HIPPOCAMPAL ANALYSES FOR EPIDEMIOLOGICA Funded Grant uri icon

description

  • Recently major efforts in Alzheimer's disease (AD) research have concentrated on the search for disease-associated biomarkers that can identify people at risk for developing AD, help with early diagnosis, prognosis or therapeutic decision-making, and allow for an expedited evaluation of novel disease-modifying therapies. The current project will use an advanced automated hippocampal segmentation approach based on an adaptive machine-learning algorithm paired with the hippocampal radial atrophy mapping technique to establish the dynamic pattern of hippocampal changes that occur in Alzheimer's disease in 3D. AdaBoost is a widely recognized breakthrough computer vision approach that shows great promise as an objective, reproducible and reliable automated hippocampal segmentation technique. While implementing cutting edge mathematical and statistical concepts, the combined technique provides high throughput and sensitivity to disease-associated hippocampal changes. The main goal of this project is to test the performance, variability, robustness and reproducibility of the approach in very large 1.5T and 3T epidemiological and 1.5T clinical trial datasets and to ultimately establish the AdaBoost/radial atrophy methodology as a potential secondary/surrogate outcome measure for clinical trials in AD and MCI that will overcome many of the limitations imposed by the widely used cognitive batteries such as ceiling and floor effects, fluctuations and learning and practice effects. In addition we will identify the hippocampal morphological changes associated with imminent conversion from MCI to AD and investigate potential cognitive and laboratory biomarker correlations with hippocampal atrophy thus investigating potential disease biosignatures.