Self-improved LeNet and modified SegNet for Alzheimer's disease classification using MRI

Alzheimer's disease is the leading cause of dementia, typically affecting the elderly. It results in cognitive and memory loss with progression that can lead to death. Although the exact cause remains unclear, it is believed to involve genetics, diet and environment. One key sign of Alzheimer's disease is the shrinkage of the hippocampus and frontal lobe cortex. Magnetic resonance imaging is frequently used to diagnose Alzheimer's disease due to its ability to capture detailed images of soft tissues. This study proposes an innovative Multi-head Parallel LeNet5-based Alzheimer's Disease Classification framework. The proposed Multi-head Parallel LeNet5-based Alzheimer's Disease Classification system processes magnetic resonance imaging images through several key steps. First, the Modified Wiener Filter is applied for noise reduction during preprocessing. Next, Pyramid Convolutional Kernels-based SegNet performs segmentation to isolate relevant brain regions. In the feature extraction phase, Improved Local Gabor Binary Pattern Histogram Sequence, shape features and deep features from VGG16 and ResNet are used to capture both texture and structural details. These features are augmented and then input into a Multi-head Parallel LeNet-5 (MPL5) classifier, which produces the final classified output for accurate and early detection of Alzheimer's disease. The Multi-head Parallel LeNet5-based Alzheimer's Disease Classification model predicts a higher accuracy score of 98% to confirm that it more consistently classifies Alzheimer's disease using magnetic resonance imaging.