Hippocampus Segmentation-Based Alzheimer’s Disease Diagnosis and Classification of MRI Images

Arabian Journal for Science and Engineering https://doi.org/10.1007/s13369-022-07538-2 RESEARCH ARTICLE-COMPUTER ENGINEERING AND COMPUTER SCIENCE Hippocampus Segmentation-Based Alzheimer’s Disease Diagnosis and Classification of MRI Images A. Balasundaram1 · Sruthi Srinivasan2 · A. Prasad2 · Jahan Malik2 · Ayush Kumar2 Received: 4 June 2022 / Accepted: 12 December 2022 © King Fahd University of Petroleum & Minerals 2023 Abstract Alzheimer’s disease represents a neurological condition characterized by steady cognitive decline and eventual memory loss due to the death of brain cells. It is one of the most prominent dementia types observed in patients and which hence underlines the imminent need for potential methods to diagnose the disease early on. This work considers a novel approach by utilizing a reduced version of one of the datasets used in this work to achieve a considerably accurate prediction while also enabling quicker training. It leverages image segmentation to isolate the hippocampus region from brain MRI images and then strikes a comparison between models trained on the segmented portions and models trained on complete images. This research uses two datasets—4 classes of images from Kaggle and a popular OASIS 2 MRI and demographic dataset. A deep learning-based approach was adopted to train the Kaggle dataset to perform severity classification, and the hippocampus region segmented from a reduced version of the OASIS dataset was trained on supervised and ensemble learning algorithms to detect Alzheimer’s disease. The metric used for the assessment of model performance is classification accuracy. A comparative analysis between the proposed approach and existing work was also performed, and it was observed that the proposed approach is effective in the early diagnosis of Alzheimer’s disease. Keywords Hippocampus segmentation · Alzheimer’s · Deep learning · Computer vision · Medical image processing 1 Introduction Alzheimer’s disease has been identified as the most prevalent variety of dementia. Starting with a seemingly mild memory loss, this disease gradually escalates to a loss of the ability to hold conversations and respond to the environment. Alzheimer’s disease affects parts of the brain that control language, thought, and memory. It can adversely impact a person’s ability to run daily errands. Statistics show that 1 in 9 people above the age of 65 has Alzheimer’s disease. This comprises 11.4% of the world’s population. Research shows that Alzheimer’s disease cases have increased by 16% due to the Covid-19 pandemic. Another research shows that 1 in 3 Alzheimer patients die, which results in a higher number of B A. Balasundaram 1 School of Computer Science and Engineering, Center for Cyber Physical Systems, Vellore Institute of Technology, Chennai, Tamil Nadu, India 2 School of Computer Science and Engineering, Vellore Institute of Technology (VIT), Chennai, Tamil Nadu, India deaths than other chronic diseases such as breast cancer and prostate cancer combined [1]. Hence, early diagnosis of the disease is of vital importance. Alzheimer’s disease is classified into different classes concerning the rate of affection in the brain, namely mild, moderate, and severe. Patients with mild Alzheimer’s experience greater memory loss. Patients are often found wandering and getting lost, having trouble handling money, and struggling to do everyday tasks. Patients with moderate Alzheimer’s disease have trouble reasoning, have poor hearing, and lose the ability to smell. At this stage patients also have delusions, hallucinations, and paranoia. Patients with severe Alzheimer’s are completely dependent on others for day-to-day tasks, the brain tissue shrinks significantly, and the patient ends up in bed for the rest of their life [2]. The objective of this research is to use computer vision to analyze brain MRI images of 2 different datasets, namely, the four classes of images from Kaggle and the OASIS dataset to classify and diagnose Alzheimer’s disease. Subsequently, deep learning techniques are applied to the Kaggle dataset after necessary pre-processing, to determine the most suitable model that yields the highest accuracy. On the other hand, 123 Arabian Journal for Science and Engineering the brain MRI images in the OASIS dataset are segmented to extract the hippocampus region in the brain, the shrinkage of which is known to aggravate Alzheimer’s disease and dementia. These segmented images are trained alongside complete images, and the best model is once again determined based on the accuracy score. Moreover, to reduce the training time, the dataset is reduced to half its original size and the training results of this reduced dataset are analyzed alongside the complete dataset training results. The main highlights of this paper are as follows: • This work considers a novel approach by utilizing a reduced version of the OASIS-2 dataset to speed up the training process without compromising on accuracy. • Image segmentation was used to isolate the hippocampus region from brain MRI images and then strike a comparison between models trained on the segmented portions and models trained on complete images. • Deep learning models such as the CNN model, multilayer model, Resnet50, and more were utilized for the classification of severity. • Machine learning and ensemble learning algorithms were used for the detection of Alzheimer’s disease by segmenting the hippocampus region of the brain from MRI images. This work mainly introduces an approach to obtain significantly accurate results using a reduced version of the OASIS-2 dataset to train models. In place of using all the images from each.nifti file in the dataset, this work uses only one image per.nifti file in the dataset. Furthermore, only one slice of each brain image was used to perform training, compared to conventional approaches that use the entire.nifti file for training. Additionally, this work performs a striking comparison using only the cropped portion of the brain for training. Followed by that it also compares training with only half the dataset to training with the entire dataset. This approach is especially useful in training computationally heavy deep learning models to achieve faster training times while also obtaining fairly accurate predictions. The remainder of this paper is organized as follows—the next section is a literature survey of existing approaches for Alzheimer’s disease detection and classification. This is followed by a proposed approach section detailing the methodology used in this work. The next section is the experimental setup which describes the software used for implementation. Following this, the results obtained in this work are discussed and explained and the paper is wrapped up with a conclusion section. 2 Literature Survey Several studies pertaining to the early detection of Alzheimer’s disease have been carried out. While some of the works used comma-separate (...truncated)