Automatic alzheimer’s disease classification using 3D deep learning model

Automatic classification of Alzheimer’s disease (AD) from 3D medical images is a challenging research problem. To retract brain circuitry and architecture, computerized reconstruction of 3-D neuron structure from microscope images is required. Despite numerous attempts, this process is still challenging, especially when images have noise or erratic neurite patterns. 3D Image segmentation can detect neural voxels before tracing or reconstructing neurons. As a result, we propose to automate the process of reconstructing neurons utilizing a 3D deep learning model for accurate AD detection in this paper. The proposed model consists of automatic neuron reconstruction from the 3D microscopic image, features engineering, and AD classification. The automatic neuron reconstruction using a 3D deep learning model consists of steps like segmentation on 3D microscopy images using Deep Learning (DL) models, the discovery of segmented structures, and the reconstruction of identified structures using an intelligent backtracking algorithm. The proposed 3D deep learning model is designed based on existing models to enhance the automatic neuron reconstruction performances from the 3D microscopic images. After neuron reconstruction, we design the deep learning-based feature engineering approach to extract structural and functional features. Finally, automatic AD classification is performed using a deep learning classifier model. The Convolutional Neural Network (CNN) model is designed for autonomous neuron feature engineering and AD classification. The performance of the proposed model is evaluated using different publicly available datasets. The outcomes reveal that the proposed model significantly improves AD classification performances using an automatic 3D neuron reconstruction approach compared to the state-of-the-art solutions.