This proposal builds upon our previous work on computational neuroanatomy of aging, and investigates the use of advanced pattern analysis methods in early detection of imaging phenotypes of Alzheimer's Disease (AD). AD is a disease of rapidly increasing prevalence and clinical significance, in an aging population. Currently available diagnostic tests detect the disease at extremely late stages, perhaps more than a decade after it first begins to develop. A variety of imaging methods offers great promise for the detection of early signs of brain change, which can be indicative of future clinical progression to AD. However structural and functional imaging patterns characteristic of AD are subtle, spatially complex, evolving dynamically with age/time, and are overlaid on changes induced by normal aging. This raises the need for sophisticated methods for image analysis and pattern recognition. Towards this goal, Aim 3 integrates high-dimensional pattern analysis and recognition methods aiming to quantitatively describe imaging patterns that have diagnostic and predictive value. Since these patterns can be spatially complex and subtle, conventional image analysis methods can be challenged. We therefore build on the state of the art of machine learning and pattern recognition, focusing on the major challenge encountered in medical imaging: high dimensionality of imaging data relative to the available sample sizes. Optimal multi-parametric feature construction and dimensionality reduction methods are necessary to meet this challenge and are considered in Aim 2 along with multi-parametric classification and clustering methods. The foundation of the pattern analysis of the work in Aim 3 is also set in Aim 1, which deals with more low-level fundamental image analysis problem: registration, which is a process that places image data to a standardized coordinate system, thereby enabling statistical analysis of spatio-temporal imaging patterns across healthy and patient populations: an optimal mathematical representation of morphological data is developed via a group-wise registration process. We apply this method to structural, functional and amyloid deposition image data from two of the largest studies of normal aging (BLSA) and AD (ADNI).

Public Health Relevance

Alzheimer's Disease poses a major medical and socioeconomic problem, which becomes increasingly prevalent as the population lives longer. This project develops and applies computer methods for multi-parametric analysis of MRI and PET scans, aiming to detect very early markers of the disease. As patterns of brain change are complex, subtle, and occur in tandem with normal changes brought by healthy aging, they are extremely difficult to detect at early stages, during which effective treatments are likely to exist in the near future. Advanced methods for pattern analysis and recognition have shown high promise in terms of measuring and detecting such early changes. This project will contribute such methods to the field. It will also apply them to two of the largest studies of normal aging and Alzheimer's disease: BLSA and ADNI.

Agency
National Institute of Health (NIH)
Institute
National Institute on Aging (NIA)
Type
Research Project (R01)
Project #
5R01AG014971-14
Application #
8881034
Study Section
Special Emphasis Panel (ZRG1)
Program Officer
Hsiao, John
Project Start
1998-08-01
Project End
2017-05-31
Budget Start
2015-06-01
Budget End
2017-05-31
Support Year
14
Fiscal Year
2015
Total Cost
Indirect Cost
Name
University of Pennsylvania
Department
Radiation-Diagnostic/Oncology
Type
Schools of Medicine
DUNS #
042250712
City
Philadelphia
State
PA
Country
United States
Zip Code
19104
Eavani, Harini; Habes, Mohamad; Satterthwaite, Theodore D et al. (2018) Heterogeneity of structural and functional imaging patterns of advanced brain aging revealed via machine learning methods. Neurobiol Aging 71:41-50
Varol, Erdem; Sotiras, Aristeidis; Davatzikos, Christos (2018) MIDAS: Regionally linear multivariate discriminative statistical mapping. Neuroimage 174:111-126
Pehlivanova, Marieta; Wolf, Daniel H; Sotiras, Aristeidis et al. (2018) Diminished Cortical Thickness is Associated with Impulsive Choice in Adolescence. J Neurosci :
Varikuti, Deepthi P; Genon, Sarah; Sotiras, Aristeidis et al. (2018) Evaluation of non-negative matrix factorization of grey matter in age prediction. Neuroimage 173:394-410
Erus, Guray; Doshi, Jimit; An, Yang et al. (2018) Longitudinally and inter-site consistent multi-atlas based parcellation of brain anatomy using harmonized atlases. Neuroimage 166:71-78
Dong, Aoyan; Toledo, Jon B; Honnorat, Nicolas et al. (2017) Heterogeneity of neuroanatomical patterns in prodromal Alzheimer's disease: links to cognition, progression and biomarkers. Brain 140:735-747
Zhu, Hancan; Cheng, Hewei; Yang, Xuesong et al. (2017) Metric Learning for Multi-atlas based Segmentation of Hippocampus. Neuroinformatics 15:41-50
Varol, Erdem; Sotiras, Aristeidis; Davatzikos, Christos et al. (2017) HYDRA: Revealing heterogeneity of imaging and genetic patterns through a multiple max-margin discriminative analysis framework. Neuroimage 145:346-364
Honnorat, Nicolas; Parker, Drew; Tunç, Birkan et al. (2017) Subject-Specific Structural Parcellations Based on Randomized AB-divergences. Med Image Comput Comput Assist Interv 10433:407-415
Rathore, Saima; Habes, Mohamad; Iftikhar, Muhammad Aksam et al. (2017) A review on neuroimaging-based classification studies and associated feature extraction methods for Alzheimer's disease and its prodromal stages. Neuroimage 155:530-548

Showing the most recent 10 out of 130 publications