Enhancing Alzheimer's Disease Staging through Multi-Modal Neuroimaging: Integrating MRI and PET for Improved Classification
Abstract (300 words maximum)
Alzheimer's disease (AD) is the prevalent neurodegenerative disease and the sixth leading cause of death in the US. Proper staging of AD is critical to enable early diagnosis and treatment. structural Magnetic Resonance Imaging (MRI) is employed intensively to detect brain atrophy in AD, while Positron Emission Tomography (PET) scans provide metabolic and amyloid deposition data, crucial for disease definition. Single-modality conventional techniques are plagued by misalignment and weak discriminatory power between Mild Cognitive Impairment (MCI) and Cognitively Normal (CN), and MCI and AD. Multi-modal learning promises improved predictive performance but is plagued by feature heterogeneity and modality misalignment. In this work, AD classification is attempted to be enhanced by using multi-modal data, i.e., from the OASIS-3 database. We employ novel feature integration algorithms to address feature space misalignment and enhance the process of fusion in this research. The freesurfer is employed to extract the brain anatomy features from structured MRI and the PET unified pipeline is employed to extract the features from PET images. We performed supervised learning using the concatenated features from MRI and PET images. The performance will be done using classification metrics such as accuracy, precision, recall, and F1-score to establish the effectiveness of the proposed method. Through the use of multi-modal neuroimaging data, we can envision a significant improvement in the accuracy of AD staging predictions. This study contributes to the development of efficient diagnostic tools that assist clinicians in early diagnosis and tracking of the disease. The findings from this study will assist in providing valuable insights into the synergistic benefits of integrating MRI and PET imaging, ultimately enriching research in neurodegenerative diseases and patient outcomes.
Academic department under which the project should be listed
CCSE - Computer Science
Primary Investigator (PI) Name
Chen Zhao
Enhancing Alzheimer's Disease Staging through Multi-Modal Neuroimaging: Integrating MRI and PET for Improved Classification
Alzheimer's disease (AD) is the prevalent neurodegenerative disease and the sixth leading cause of death in the US. Proper staging of AD is critical to enable early diagnosis and treatment. structural Magnetic Resonance Imaging (MRI) is employed intensively to detect brain atrophy in AD, while Positron Emission Tomography (PET) scans provide metabolic and amyloid deposition data, crucial for disease definition. Single-modality conventional techniques are plagued by misalignment and weak discriminatory power between Mild Cognitive Impairment (MCI) and Cognitively Normal (CN), and MCI and AD. Multi-modal learning promises improved predictive performance but is plagued by feature heterogeneity and modality misalignment. In this work, AD classification is attempted to be enhanced by using multi-modal data, i.e., from the OASIS-3 database. We employ novel feature integration algorithms to address feature space misalignment and enhance the process of fusion in this research. The freesurfer is employed to extract the brain anatomy features from structured MRI and the PET unified pipeline is employed to extract the features from PET images. We performed supervised learning using the concatenated features from MRI and PET images. The performance will be done using classification metrics such as accuracy, precision, recall, and F1-score to establish the effectiveness of the proposed method. Through the use of multi-modal neuroimaging data, we can envision a significant improvement in the accuracy of AD staging predictions. This study contributes to the development of efficient diagnostic tools that assist clinicians in early diagnosis and tracking of the disease. The findings from this study will assist in providing valuable insights into the synergistic benefits of integrating MRI and PET imaging, ultimately enriching research in neurodegenerative diseases and patient outcomes.