Accurate diagnosis and progression modeling of Alzheimer’s Disease (AD) are critical for effective intervention, disease monitoring, and patient care. Traditional approaches rely on a single modality, such as clinical assessments, neuroimaging, or genetic markers, but may fail in capturing the complex, multifaceted nature of AD. Therefore, multimodal learning addresses this limitation by integrating complementary information across sources, but conventional fusion strategies, such as early feature concatenation and late decision-level fusion, often model modalities independently and fail to capture high-order cross-modal interactions that are essential for reliable diagnosis and progression modeling. To address these limitations, we propose two complementary multimodal frameworks that together advance AD diagnosis and longitudinal risk assessment. For diagnosis, we propose a Multimodal Tensor Fusion Network (MTFN) that integrates heterogeneous data sources, including visual imagery, demographics, and longitudinal time-series data. This approach leverages tensor representations to model cross-modal interactions while preserving structural dependencies within each modality. For progression modeling, we propose a Multimodal Supervised Temporal Encoder Joint Modeling network (MSTE-JM) that uses a supervised temporal encoder to learn compact multimodal representations, which are subsequently incorporated into a Bayesian joint model linking longitudinal cognitive measurements with time-to-event outcomes.
Experiments on publicly available AD datasets demonstrate that MTFN outperforms deep learning classifiers in diagnostic accuracy, while MSTE-JM produces robust progression estimates that capture the temporal dynamics of cognitive decline and the risk of conversion. Together, these proposed frameworks establish tensor-based multimodal learning as a powerful contribution for advancing both accurate detection and long-term risk stratification of neurodegenerative diseases.
","abstract_html":"<p>Accurate diagnosis and progression modeling of Alzheimer’s Disease (AD) are critical for effective intervention, disease monitoring, and patient care. Traditional approaches rely on a single modality, such as clinical assessments, neuroimaging, or genetic markers, but may fail in capturing the complex, multifaceted nature of AD. Therefore, multimodal learning addresses this limitation by integrating complementary information across sources, but conventional fusion strategies, such as early feature concatenation and late decision-level fusion, often model modalities independently and fail to capture high-order cross-modal interactions that are essential for reliable diagnosis and progression modeling. To address these limitations, we propose two complementary multimodal frameworks that together advance AD diagnosis and longitudinal risk assessment. For diagnosis, we propose a Multimodal Tensor Fusion Network (MTFN) that integrates heterogeneous data sources, including visual imagery, demographics, and longitudinal time-series data. This approach leverages tensor representations to model cross-modal interactions while preserving structural dependencies within each modality. For progression modeling, we propose a Multimodal Supervised Temporal Encoder Joint Modeling network (MSTE-JM) that uses a supervised temporal encoder to learn compact multimodal representations, which are subsequently incorporated into a Bayesian joint model linking longitudinal cognitive measurements with time-to-event outcomes.</p> <p>Experiments on publicly available AD datasets demonstrate that MTFN outperforms deep learning classifiers in diagnostic accuracy, while MSTE-JM produces robust progression estimates that capture the temporal dynamics of cognitive decline and the risk of conversion. Together, these proposed frameworks establish tensor-based multimodal learning as a powerful contribution for advancing both accurate detection and long-term risk stratification of neurodegenerative diseases.</p>","abstract_has_math":false,"creators":["Le, Tiffany Tien"],"institution":null,"degree_name":null,"degree_level":"Thesis","degree_discipline":"Electrical Engineering and Computer Science","degree_department":null,"school":null,"contributors":["Yuxin Wen, Ph.D.","Thomas C. Piechota, Ph.D.","Chelsea Parlett-Pelleriti, Ph.D.","Trudi Qi, Ph.D."],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2026,"date_issued":"2026-05-01T07:00:00Z","date_published":"2026-05-01T07:00:00Z","updated_at":"2026-07-24T01:38:47Z","subjects":["Alzheimer's Disease","Tensor Fusion","Multimodal Learning","Joint Modeling","Machine Learning","Disease Progression","Artificial Intelligence and Robotics","Data Science","Longitudinal Data Analysis and Time Series","Neurosciences","Survival Analysis"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://digitalcommons.chapman.edu/eecs_theses/9","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Yuxin Wen, Ph.D.","Thomas C. Piechota, Ph.D.","Chelsea Parlett-Pelleriti, Ph.D.","Trudi Qi, Ph.D."]},{"key":"dc:creator","label":"Author","values":["Le, Tiffany Tien"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2028-05-17T07:00:00Z"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Electrical Engineering and Computer Science"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Alzheimer's Disease","Tensor Fusion","Multimodal Learning","Joint Modeling","Machine Learning","Disease Progression","Artificial Intelligence and Robotics","Data Science","Longitudinal Data Analysis and Time Series","Neurosciences","Survival Analysis"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://digitalcommons.chapman.edu/eecs_theses/9"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Accurate diagnosis and progression modeling of Alzheimer’s Disease (AD) are critical for effective intervention, disease monitoring, and patient care. Traditional approaches rely on a single modality, such as clinical assessments, neuroimaging, or genetic markers, but may fail in capturing the complex, multifaceted nature of AD. Therefore, multimodal learning addresses this limitation by integrating complementary information across sources, but conventional fusion strategies, such as early feature concatenation and late decision-level fusion, often model modalities independently and fail to capture high-order cross-modal interactions that are essential for reliable diagnosis and progression modeling. To address these limitations, we propose two complementary multimodal frameworks that together advance AD diagnosis and longitudinal risk assessment. For diagnosis, we propose a Multimodal Tensor Fusion Network (MTFN) that integrates heterogeneous data sources, including visual imagery, demographics, and longitudinal time-series data. This approach leverages tensor representations to model cross-modal interactions while preserving structural dependencies within each modality. For progression modeling, we propose a Multimodal Supervised Temporal Encoder Joint Modeling network (MSTE-JM) that uses a supervised temporal encoder to learn compact multimodal representations, which are subsequently incorporated into a Bayesian joint model linking longitudinal cognitive measurements with time-to-event outcomes.
Experiments on publicly available AD datasets demonstrate that MTFN outperforms deep learning classifiers in diagnostic accuracy, while MSTE-JM produces robust progression estimates that capture the temporal dynamics of cognitive decline and the risk of conversion. Together, these proposed frameworks establish tensor-based multimodal learning as a powerful contribution for advancing both accurate detection and long-term risk stratification of neurodegenerative diseases.
"]},{"key":"dc:source","label":"Dc Source","values":["T. T. Le, \"Multimodal learning for disease diagnosis and progression modeling,\" M. S. thesis, Chapman University, Orange, CA, 2026. https://doi.org/10.36837/chapman.000733"]},{"key":"dc:title","label":"Title","values":["Multimodal Learning for Disease Diagnosis and Progression Modeling"]}]}],"canonical_facts":{"dc:contributor":["Yuxin Wen, Ph.D.","Thomas C. Piechota, Ph.D.","Chelsea Parlett-Pelleriti, Ph.D.","Trudi Qi, Ph.D."],"dc:creator":["Le, Tiffany Tien"],"dc:date.available":["2028-05-17T07:00:00Z"],"dc:description.abstract":["Accurate diagnosis and progression modeling of Alzheimer’s Disease (AD) are critical for effective intervention, disease monitoring, and patient care. Traditional approaches rely on a single modality, such as clinical assessments, neuroimaging, or genetic markers, but may fail in capturing the complex, multifaceted nature of AD. Therefore, multimodal learning addresses this limitation by integrating complementary information across sources, but conventional fusion strategies, such as early feature concatenation and late decision-level fusion, often model modalities independently and fail to capture high-order cross-modal interactions that are essential for reliable diagnosis and progression modeling. To address these limitations, we propose two complementary multimodal frameworks that together advance AD diagnosis and longitudinal risk assessment. For diagnosis, we propose a Multimodal Tensor Fusion Network (MTFN) that integrates heterogeneous data sources, including visual imagery, demographics, and longitudinal time-series data. This approach leverages tensor representations to model cross-modal interactions while preserving structural dependencies within each modality. For progression modeling, we propose a Multimodal Supervised Temporal Encoder Joint Modeling network (MSTE-JM) that uses a supervised temporal encoder to learn compact multimodal representations, which are subsequently incorporated into a Bayesian joint model linking longitudinal cognitive measurements with time-to-event outcomes.
Experiments on publicly available AD datasets demonstrate that MTFN outperforms deep learning classifiers in diagnostic accuracy, while MSTE-JM produces robust progression estimates that capture the temporal dynamics of cognitive decline and the risk of conversion. Together, these proposed frameworks establish tensor-based multimodal learning as a powerful contribution for advancing both accurate detection and long-term risk stratification of neurodegenerative diseases.
"],"dc:identifier":["https://digitalcommons.chapman.edu/eecs_theses/9"],"dc:source":["T. T. Le, \"Multimodal learning for disease diagnosis and progression modeling,\" M. S. thesis, Chapman University, Orange, CA, 2026. https://doi.org/10.36837/chapman.000733"],"dc:subject":["Alzheimer's Disease","Tensor Fusion","Multimodal Learning","Joint Modeling","Machine Learning","Disease Progression","Artificial Intelligence and Robotics","Data Science","Longitudinal Data Analysis and Time Series","Neurosciences","Survival Analysis"],"dc:title":["Multimodal Learning for Disease Diagnosis and Progression Modeling"],"thesis:degree_discipline":["Electrical Engineering and Computer Science"],"thesis:degree_level":["Thesis"]},"updated_at":"2026-07-24T01:38:47Z"}