Abstract:Alzheimer's disease (AD) is a severe irreversible neurodegenerative disease that has great sufferings on patients and eventually leads to death. Early detection of AD and its prodromal stage, mild cognitive impairment (MCI) which can be either stable (sMCI) or progressive (pMCI), is highly desirable for effective treatment planning and tailoring therapy. Recent studies recommended using multimodal data fusion of genetic (single nucleotide polymorphisms, SNPs) and neuroimaging data (magnetic resonance imaging (MRI) and positron emission tomography (PET)) to discriminate AD/MCI from normal control (NC) subjects. However, missing multimodal data in the cohort under study is inevitable. In addition, data heterogeneity between phenotypes and genotypes biomarkers makes learning capability of the models more challenging. Also, the current studies mainly focus on identifying brain disease classification and ignoring the regression task. Furthermore, they utilize multistage for predicting the brain disease progression. To address these issues, we propose a novel multimodal neuroimaging and genetic data fusion for joint classification and clinical score regression tasks using the maximum number of available samples in one unified framework using convolutional neural network (CNN). Specifically, we initially perform a technique based on linear interpolation to fill the missing features for each incomplete sample. Then, we learn the neuroimaging features from MRI, PET, and SNPs using CNN to alleviate the heterogeneity among genotype and phenotype data. Meanwhile, the high learned features from each modality are combined for jointly identifying brain diseases and predicting clinical scores. To validate the performance of the proposed method, we test our method on 805 subjects from Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset. Also, we verify the similarity between the synthetic and real data using statistical analysis. Moreover, the experimental results demonstrate that the proposed method can yield better performance in both classification and regression tasks. Specifically, our proposed method achieves accuracy of 98.22%, 93.11%, and 97.35% for NC vs. AD, NC vs. sMCI, and NC vs. pMCI, respectively. On the other hand, our method attains the lowest root mean square error and the highest correlation coefficient for different clinical scores regression tasks compared with the state-of-the-art methods.

A Multi-classification Accessment Framework for Reproducible Evaluation of Multimodal Learning in Alzheimer's Disease

A Multimodal Learning Machine Framework for Alzheimer’s Disease Diagnosis Based on Neuropsychological and Neuroimaging Data

A Progressive Single-Modality to Multi-Modality Classification Framework for Alzheimer's Disease Sub-type Diagnosis

Multi-modal Self-paced Locality Preserving Learning for Diagnosis of Alzheimer’s Disease

Multimodal Identification of Alzheimer's Disease: A Review

Multimodal deep learning for Alzheimer’s disease dementia assessment

A feature-aware multimodal framework with auto-fusion for Alzheimer's disease diagnosis

A Feature-Aware Multimodal Framework with Auto-Fusion for Alzheimer’s Disease Diagnosis

Patch-based Deep Multi-Modal Learning Framework for Alzheimer’s Disease Diagnosis Using Multi-View Neuroimaging

Patch-based Interpretable Deep Learning Framework for Alzheimer’s Disease Diagnosis Using Multimodal Data

Toward Robust Early Detection of Alzheimer's Disease via an Integrated Multimodal Learning Approach

Multi-modal cross-attention network for Alzheimer's disease diagnosis with multi-modality data

Inter-modality Relationship Constrained Multi-Modality Multi-Task Feature Selection for Alzheimer's Disease and Mild Cognitive Impairment Identification.

Alzheimer’s Disease Diagnosis Via Multimodal Feature Fusion

Label-aligned multi-task feature learning for multimodal classification of Alzheimer's disease and mild cognitive impairment

MDMA: Multimodal Data and Multi-attention Based Deep Learning Model for Alzheimer’s Disease Diagnosis

Joint learning framework of cross-modal synthesis and diagnosis for Alzheimer’s disease by mining underlying shared modality information

Joint Learning Framework of Cross-Modal Synthesis and Diagnosis for Alzheimer's Disease by Mining Underlying Shared Modality Information.

Subclass-based multi-task learning for Alzheimer's disease diagnosis

Alzheimer's disease diagnosis framework from incomplete multimodal data using convolutional neural networks

Multi-Modal Data Analysis for Alzheimer’s Disease Diagnosis: An Ensemble Model Using Imagery and Genetic Features