Deep Feature Fusion Framework for Alzheimer’s Disease Staging using Neuroimaging Modalities
Keywords: Alzheimer’s Disease, Neuroimaging Features, 3D Image Classification.
TL;DR: An automated and accurate multimodal system integrating MRI and PET images for early Alzheimer’s diagnosis.
Abstract: Alzheimer’s Disease (AD) is a significant neurodegenerative disorder. Detecting AD early is essential for effective management and improving the quality of life for both patients and their families. Recent advancements in medical imaging technology have introduced neuroimaging-based methods for early AD diagnosis. However, the challenges in early AD detection suggest that using a single modality dataset in deep learning (DL) studies, particularly neuroimaging, might not yield precise predictions of AD progression compared to integrating data from multiple imaging modalities. Utilizing information from multi-modal data fusion can enhance the detection of subtle changes and biomarkers, leading to more reliable and accurate diagnosis. In our study, we develop an automated multimodal system that integrates MRI and PET images at an intermediate fusion level, facilitating the early diagnosis of Alzheimer’s disease. This fusion approach does not require extensive preprocessing steps typically associated with the image fusion method. Our proposed methodology outperforms previous studies in differentiating between individuals with Alzheimer’s disease and cognitively normal (CN) individuals, achieving an AUC score of 97.67% and an accuracy (ACC) of 95.24%.
Submission Number: 8
Loading