Deep-Learning Methods for the Long-Term Prognosis of Alzheimer's Disease Integrating Cerebrospinal Fluid Biomarkers, Magnetic Resonance Imaging, Positron Emission Tomography, and Cognitive Test Scores

Authors

Ulyana P. Martin, Florida International UniversityFollow

Document Type

Dissertation

Degree

Doctor of Philosophy (PhD)

Major/Program

Electrical and Computer Engineering

First Advisor's Name

Malek Adjouadi

First Advisor's Committee Title

Committee chair

Second Advisor's Name

David Loewenstein

Second Advisor's Committee Title

Committee member

Third Advisor's Name

Armando Barreto

Third Advisor's Committee Title

Committee member

Fourth Advisor's Name

Mercedes Cabrerizo

Fourth Advisor's Committee Title

Committee member

Fifth Advisor's Name

Naphtali Rishe

Fifth Advisor's Committee Title

Committee member

Keywords

Longitudinal analysis, Deep learning, multitask, multimodal, missing data, LSTM, Cerebrospinal fluid, magnetic resonance imaging (MRI), Alzheimer's disease, Alzheimer's Disease Neuroimaging Initiative (ADNI)

Date of Defense

3-17-2023

Abstract

Alzheimer’s disease (AD) is a neurodegenerative condition characterized by sharp cognitive decline with no confirmed effective treatment or cure. This makes it critically important to identify the symptoms of Alzheimer’s disease in its early stages before significant cognitive deterioration has taken hold and even before any brain morphology and neuropathology are noticeable. Therefore, the central aim of this research is the development of novel multitask multimodal deep learning models for the prediction of cognitive test scores.
The association of CSF Amyloid-β, pTau, and specific MRI regions with disease
progression is first assessed to determine how closely such biomarkers relate to the AD pathology, adding insight into the potential differences that exist between stable subjects and those who progress to AD.
Thereafter, five different multimodal deep neural networks are developed with different architectures and underlying characteristics, in search of an optimal model for predicting the cognitive test scores of the Mini Mental State Examination (MMSE) and the modified Alzheimer’s Disease Assessment Scale (ADAS-CoG13) over a span of 60 months (5 years). The longitudinal multimodal data utilized to train and test the models was obtained from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) study and includes cerebrospinal fluid (CSF) levels of tau and beta-amyloid, structural measures from magnetic resonance imaging (MRI), functional and metabolic measures from positron emission tomography (PET), and cognitive scores from the aforementioned neuropsychological tests (Cog). To improve the results of said models, a data augmentation technique is created to generate more training and testing samples from the available data. The models developed herein delve into two main issues: 1) exploring the merits of single task (chapter 5 and 6.2.1) vs. multitask (chapters 4 and 6.2.2) for predicting future cognitive scores; and 2) determining whether time-varying input data (chapters 6.2.1 and 6.2.2) are better suited than specific time-points (chapters 4 and 5) for optimizing these prediction results.
The best model (6.2.1) yields 90.27% (SD=1.36) prediction accuracy for 6 months
after the last visit, 87.42% (SD =3.42) for 12 months, 87.30% (SD=1.94) for 18 months, 81.79% (SD=7.78) for 24 months, 82.48% (SD=3.21) for 30 months, 80.49% (SD=5.87) for 36 months, 80.58% (SD=3.95) for 42 months, 78.68% (SD=6.66) for 48 months, 70.12%(SD=15.05) for 54 months, and 79.91% (SD=8.84) for 60 months. The analysis provided is comprehensive as it includes converters in the CN and MCI groups (CNc, MCIc), and does not reject the unstable groups in the CN and MCI groups (CNun and MCIun), groups that revert back to CN from MCI and to MCI from AD, so as not to bias the results.

Identifier

FIDC011073

ORCID

0000-0001-8283-7055

Previously Published In

Morar, U., Martin, H., Izquierdo, W., Forouzannezhad, P., Zarafshan,Curiel R.E., Roselli M., Cabrerizo, M., Loewenstein, D., Duara, R., and Adjouadi, M. A Deep-Learning Approach for the Prediction of Mini-Mental State Examination Scores in a Multimodal Longitudinal StudyThe 2020 International Conference on Computational Science and Computational Intelligence (CSCI’20), Las Vegas, Nevada, USA, Dec. 2020

Morar, U., Izquierdo, W., Martin, H., Forouzannezhad, P., Zarafshan, E., Unger, E., Bursac, Z., Cabrerizo, M., Barreto, A., Vaillancourt, D. E., DeKosky, S. T., Loewenstein, D., Duara, R., and Adjouadi, M. (2022). A study of the longitudinal changes in multiple cerebrospinal fluid and volumetric magnetic resonance imaging biomarkers on converter and non-converter Alzheimer’s disease subjects with consideration for their amyloid beta status. Alzheimer’s and dementia, 14(1), e12258. https://doi.org/10.1002/dad2.12258

Recommended Citation

Martin, Ulyana P., "Deep-Learning Methods for the Long-Term Prognosis of Alzheimer's Disease Integrating Cerebrospinal Fluid Biomarkers, Magnetic Resonance Imaging, Positron Emission Tomography, and Cognitive Test Scores" (2023). FIU Electronic Theses and Dissertations. 5259.
https://digitalcommons.fiu.edu/etd/5259