Abstract:Alzheimer's Disease (AD) is an irreversible neurodegenerative disorder that often progresses from Mild Cognitive Impairment (MCI), leading to memory loss and significantly impacting patients' lives. Clinical trials indicate that early targeted interventions for MCI patients can potentially slow or halt the development and progression of AD. Previous research has shown that accurate medical classification requires the inclusion of extensive multimodal data, such as assessment scales and various neuroimaging techniques like Magnetic Resonance Imaging (MRI) and Positron Emission Tomography (PET). However, consistently tracking the diagnosis of the same individual over time and simultaneously collecting multimodal data poses significant challenges. To address this issue, we introduce GFE-Mamba, a classifier based on Generative Feature Extraction (GFE). This classifier effectively integrates data from assessment scales, MRI, and PET, enabling deeper multimodal fusion. It efficiently extracts both long and short sequence information and incorporates additional information beyond the pixel space. This approach not only improves classification accuracy but also enhances the interpretability and stability of the model. We constructed datasets of over 3000 samples based on the Alzheimer's Disease Neuroimaging Initiative (ADNI) for a two-step training process. Our experimental results demonstrate that the GFE-Mamba model is effective in predicting the conversion from MCI to AD and outperforms several state-of-the-art methods. Our source code and ADNI dataset processing code are available at <a class="link-external link-https" href="https://github.com/Tinysqua/GFE-Mamba" rel="external noopener nofollow">this https URL</a>.

What problem does this paper attempt to address?

The problem that this paper attempts to solve is the prediction of the development of Alzheimer's disease (AD) from the mild cognitive impairment (MCI) stage to the full - dementia stage. Specifically, the research aims to improve the prediction accuracy of whether MCI patients will progress to AD, so as to achieve early intervention and thus delay or prevent the development of the disease. ### Main problems and challenges 1. **Data acquisition and fusion**: - In order to accurately predict the conversion from MCI to AD, it is necessary to combine multi - modal data, such as assessment scales, magnetic resonance imaging (MRI) and positron emission tomography (PET). However, in clinical practice, it is very challenging to continuously track the diagnosis of the same patient and simultaneously collect multi - modal data. - Although PET imaging can provide information on brain metabolic activities with high sensitivity and specificity, its high cost and complex technology limit its wide application. 2. **Limitations of existing methods**: - Current prediction methods rely on neuroimaging, cognitive scale scores and biomarker tests, but these methods have significant challenges in terms of accuracy and reliability, especially when dealing with individual differences. ### Solution: GFE - Mamba model To solve the above problems, the authors propose a new model named GFE - Mamba, which is based on Generative Feature Extraction (GFE) and improves the prediction performance in the following ways: 1. **3D GAN - ViT architecture**: - Combine 3D Generative Adversarial Network (GAN) with Vision Transformer (ViT) to generate PET images from MRI, so that its features can be utilized even in the absence of PET data. - The loss function includes Mean Squared Error (MSE), adversarial loss and perceptual loss to ensure the quality of the generated PET images. 2. **Multi - modal Mamba classifier**: - Integrate assessment scales, MRI and PET data, and process long - sequence information through six Mamba blocks to enhance the classification ability of the model. - Introduce time - interval extraction and assessment scale pre - processing to better capture the changes in the patient's state. 3. **Pixel - level dual - cross - attention mechanism**: - Integrate MRI and PET data at the pixel level, which solves the limitations of shallow spatial image information processing and improves the interpretability and stability of the model. ### Experimental verification - Researchers constructed three datasets using more than 3,000 samples from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database for a two - step training process. - The experimental results show that the GFE - Mamba model performs excellently in predicting the conversion from MCI to AD and is superior to several existing advanced methods. Through these improvements, the GFE - Mamba model not only improves the classification accuracy, but also enhances the interpretability and stability of the model, providing a more reliable tool for the early identification and intervention of Alzheimer's disease.

GFE-Mamba: Mamba-based AD Multi-modal Progression Assessment via Generative Feature Extraction from MCI

Attention-based and Micro Designed EfficientNetB2 for Diagnosis of Alzheimer’s Disease

Hybrid Multimodality Fusion with Cross-Domain Knowledge Transfer to Forecast Progression Trajectories in Cognitive Decline

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

Deep learning and multimodal feature fusion for the aided diagnosis of Alzheimer's disease

MMTFN: Multi‐modal multi‐scale transformer fusion network for Alzheimer's disease diagnosis

Alzheimer’s Disease Diagnosis from Multi-Modal Data Via Feature Inductive Learning and Dual Multilevel Graph Neural Network

Enhancing Early Alzheimer's Disease Classification Accuracy Through the Fusion of Smri and Rsmeg Data: a Deep Learning Approach

Alzheimer's disease diagnosis from multi-modal data via feature inductive learning and dual multilevel graph neural network

A deep feature fusion network with global context and cross-dimensional dependencies for classification of mild cognitive impairment from brain MRI

Prediction of Progressive Mild Cognitive Impairment by Multi-Modal Neuroimaging Biomarkers

Identification of mild cognitive impairment using multimodal 3D imaging data and graph convolutional networks

An Effective Multimodal Image Fusion Method Using MRI and PET for Alzheimer's Disease Diagnosis

Multicenter and Multichannel Pooling GCN for Early AD Diagnosis Based on Dual-Modality Fused Brain Network

MLC-GCN: Multi-Level Generated Connectome Based GCN for AD Analysis

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

A coupled-GAN architecture to fuse MRI and PET image features for multi-stage classification of Alzheimer's disease

Deep joint learning diagnosis of Alzheimer's disease based on multimodal feature fusion

Dual attention based fusion network for MCI Conversion Prediction

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

Hierarchical feature representation and multimodal fusion with deep learning for AD/MCI diagnosis