Multimodal fusion diagnosis of the Alzheimer's disease via lightweight CNN-LSTM model using magnetic resonance imaging (MRI)

Alzheimer's disease is categorized as a primary neurodegenerative ailment that mostly affects individuals in the elderly age and those reaching later stages of life. The recognition of this conditions may be ascribed to the persistent cognitive processes that display similarities to the state of anxiety. The timely identification of Alzheimer's disease (AD) and its precursor, mild cognitive impairment (MCI), occurring potentially up to 8 years prior to the manifestation of dementia symptoms, has great significance. The incorporation of Neuroimaging methods are now indispensable in the assessment and timely identification of Alzheimer's disease (AD). The use of artificial intelligence in the medical domain has become prevalent, and deep learning has emerged as a promising approach for computer-assisted Alzheimer's disease diagnosis using magnetic resonance imaging (MRI).This paper presents a unique, efficient, and precise multimodal fusion deep learning architecture that utilizes Convolutional Neural Network (CNN) and Long Short-Term Memory (LSTM). In this study, we used a well-established dataset called Alzheimer's Disease Neuroimaging Initiative-1 (ADNI-1) as the gold-standard dataset for detecting Alzheimer's disease (AD). ADNI-1 included 3D characteristics, which were transformed into 2D features and then employed in our innovative model. Subsequently, the innovative model was subjected to training and evaluation using the aforementioned dataset. Based on the empirical findings, the newly introduced model achieved a remarkable accuracy of 92.3 %, surpassing the accuracies achieved by the current leading baseline models while using fewer parameters and requiring less processing complexity.