Alzheimer’s Disease Diagnosis via SVM-Driven Classification of Structural MRI Data

Abstract

In this study we introduce an automated system designed to classify Alzheimers disease (AD) using MRI brain scans. Our technique employs support vector machine (SVM) classification to effectively differentiate between AD patients and older adults without the disease by analyzing brain MRI images. The process includes dividing 2D T1 weighted MR images into regions of interest and extracting gray matter characteristics, from each region. By leveraging SVMs reliability and proven success in classification tasks along with the principle of minimizing risks our research showcases performance in generalizing results. This SVM based approach shows promise in distinguishing AD affected brains from ones presenting a valuable opportunity, for enhancing medical analysis and interpretation of MRI findings.

Introduction

I. INTRODUCTION

Dementia is becoming especially Alzheimers disease, which accounts for half to sixty percent of dementia cases, in older individuals. It is crucial to diagnose Alzheimers accurately as it allows for treatment with cholinesterase inhibitors. Research, into the symptoms of Alzheimers has led to the recognition of mild cognitive impairment (MCI). The use of imaging has become crucial in diagnosing Alzheimers disease. The process of AD related brain degeneration typically begins in the lobe gradually affecting the hippocampus, neocortex and other related areas. In particular there is focus, on assessing the shrinkage of the temporal lobe (MTA) targeting key regions such, as the hippocampus, entorhinal cortex and amygdala. However assessing MTA faces challenges during stages of cognitive decline and Alzheimers disease onset. In light of these difficulties our research introduces a technique, for AD classification using MR image analysis. We chose to utilize SVMs due to their effectiveness in managing data and identifying intricate decision boundaries, which makes them ideal for AD classification tasks. Moreover SVMs come with the benefit of regularization which helps address issues related to overfitting commonly seen in noisy datasets. Additionally we suggest a preprocessing process aimed at standardizing and improving the quality of the input data to ensure trustworthy classification outcomes. Our goal is to offer an early approach, for AD by leveraging the complete potential of whole brain MR image data.

II. LITERATURE REVIEW

The literature review discussed the use of Support Vector Machines (SVM), in Alzheimers Disease (AD) classification with MRI data. Various methods like Sparse Logistic Regression (SLR) Regularized SLR (SRSLR) ReliefF based feature selection Feature Based Morphometry (FBM) with Bag of Visual Words (BOVW) modeling and shape analysis using P type Fourier descriptors have been explored for AD detection and classification. While each method has strengths and weaknesses SVM is highlighted as a choice for AD classification. In the realm of Alzheimers Disease classification using MRI data different methods have been studied, each offering approaches with pros and cons. Sparse Logistic Regression is noted for its ability to improve AD classification accuracy by selecting features through sparsity penalties. However it may face challenges related to imbalances between features and samples that could impact its effectiveness. On the side ReliefF based feature selection stands out for its compatibility, with Support Vector Machines enhancing AD classification accuracy by pinpointing white matter voxels. This method addresses constraints linked with SVMs. On the hand Feature Based Morphometry (FBM) combined with the Bag of Visual Words (BOVW) model and shape analysis using P type Fourier descriptors present viewpoints by focusing on the structural aspects of MRI data. FBM utilizes BOVW modeling for feature extraction making it particularly relevant, for studying brain changes related to AD. Shape analysis, demonstrated through P type Fourier descriptors takes an approach by examining changes in brain structures such as the ventricle shape showing potential benefits over conventional volume analysis. While these techniques offer insights Support Vector Machines (SVMs) consistently emerge as an versatile option for AD classification. This is due, to their capability to handle data recognize complex decision boundaries and provide regularization methods to address overfitting concerns. In conclusion this review highlights how SVM based approaches consistently outperform techniques in terms of dependability and precision.

III. METHODOLOGY

In this part we describe the approach used in our research including gathering and preparing data extracting features and applying Support Vector Machine (SVM) for classification purposes.

A. Data Collection and Preprocessing

1. Data Acquisition: We acquired a set of 2D T1 weighted MR images that includes scans, from individuals with Alzheimers disease (AD) as scans from older control participants. This collection of data is diverse. Has an amount of samples, for conducting insightful analyses.
2. Preprocessing Pipeline: We used a preprocessing process to standardize and improve the quality of the MRI data we received. This process involved the steps:
3. Binary Mask Generation: Applied the Otsu method to create a binary mask to separate brain tissue from background noise.
4. Skull Stripping: Eroded the binary mask to remove the skull and other unwanted features, leaving only the brain tissue.
5. Connected Component Labeling: Labeled the connected regions within the mask to identify distinct brain regions.
6. Largest Connected Region: Identified the largest connected region within the mask, which corresponds to the brain region of interest, and removed all other regions.
7. Mask Dilation: Dilated the mask to include more features while preserving the primary brain region.
8. Application of Mask: Applied the final mask obtained from the preprocessing pipeline to the original MRI images, effectively removing unwanted features and retaining the brain region for further analysis.

B. Feature Extraction

1. Region of Interest (ROI) Parcellation: We divided the processed 2D T1 weighted MR images into areas of interest (AOIs) for diagnosing Alzheimers disease. These areas included brain structures linked to AD like the hippocampus, entorhinal cortex and amygdala.
2. Gray Matter Feature Extraction: We analyzed characteristics of the matter, in each region of interest (ROI). This analysis encompassed data based on intensity, texture characteristics and geometric properties to ensure that the extracted features effectively represented the information, in the MR images.

C. Support Vector Machine (SVM) Classification

1. Dataset Split: We divided the data into sets, for training and testing making sure to keep a balance to train and assess the model.
2. Feature Scaling: We adjusted the gray matter characteristics we gathered to have a mean and variance to aid in training the SVM model.
3. Linear SVM Model: We utilized a linear kernel SVM classifier with standard hyperparameters.
4. Model Training: We trained the linear SVM classifier on the training dataset.
5. Model Evaluation: We tested the SVM classifier on the test data to see how well it classified information. We looked at metrics, like accuracy, precision, recall and F1 score to evaluate its performance.

IV. RESULTS

In this part we share the outcomes of our SVM driven approach, for categorizing Alzheimers Disease (AD) by analyzing MRI brain scans. The research involved employing a linear SVM classifier, with settings and a thorough preprocessing procedure detailed in Section III.

A. Classification Performance

Our approach showed classification results on the dataset proving its ability to differentiate between individuals, with Alzheimers disease and older control participants. We calculated the following performance measures:

1. Accuracy: Our model achieved an accuracy of 98.36%, indicating a high level of correctness in classifying AD patients and control subjects.
2. Precision: The precision of our model was 98.28%, indicating the ability to correctly classify AD patients without many false positives.
3. Recall: Our model demonstrated a recall rate of 98.43%, signifying its capacity to accurately identify AD patients while minimizing false negatives.
4. F1-Score: The F1-score, a harmonic mean of precision and recall, reached 98.36%, further confirming the robustness of our classification approach.

B. Comparison with Existing Methods

In order to provide context, for our findings we conducted a comparison of the effectiveness of our SVM approach, with methods reported in previous studies. The following table outlines the results of this evaluation:

TABLE I: Comparison of Classification Results [1]

Our new technique consistently performed better, than methods showing accuracy, precision, recall and F1 score. These findings highlight the effectiveness and potential real world impact of our approach, in diagnosing AD.

V. FUTURE SCOPE

In the field of Alzheimers disease (AD) categorization our study sets the foundation, for exploring possibilities by utilizing a blend of Convolutional Neural Networks (CNN) to extract features and Support Vector Machines (SVM) for classification:

A. Custom SVM Kernels

Our current model uses a SVM kernel. Future studies could delve into creating and refining more customized kernels that suit the unique traits of each dataset. This approach could enhance the accuracy of classification. Make the model easier to interpret.

B. Advanced CNN-Based Preprocessing

Improving the processing stage by integrating cutting edge CNN designs, for enhancing brain images and extracting features shows promise. Innovative methods, for cleaning up images aligning them and standardizing them can enhance the quality of feature maps. Boost classification accuracy.

C. Domain Adaptation for Universal Models

Researchers can explore domain adaptation techniques to adjust a model, for datasets, with varying acquisition methods and image qualities. This adjustment may enhance the models generalization and usefulness across scenarios.

D. Integration of Additional Patient Information

By including details, about the patient like their gender, age, genetic indicators and cognitive tests, during the models training phase we can develop a thorough diagnostic tool. Further studies could investigate how these elements influence the classification of Alzheimers disease and support tailored treatments.

E. Multi-Class Classification

Expanding the model to deal with class classification that includes Normal Control (NC) Mild Cognitive Impairment (MCI) and AD could help in detecting MCI at an early stage. By utilizing algorithms and larger datasets we can achieve identification of these different stages.

F. Real-Time Prediction and Deployment

Enhancing the speed of predictions is vital, for uses. Studies should emphasize refining models. Utilizing hardware methods to support instant or nearly instant Alzheimers disease classification, which can help prompt interventions and aid, in clinical decision making.

G. Ethical and Privacy Considerations

As the capabilities of AD classification models advance it’s crucial to prioritize privacy concerns. Moving forward research should focus on developing strategies to safeguard data privacy and promote AI standards, in healthcare settings.

Conclusion

We have come up with a method that can automatically differentiate between people, in the stages of Alzheimers disease and those who are healthy. This new approach shows potential for detecting Alzheimers disease. Our next steps involve testing its accuracy in diagnosing Alzheimers in individuals with Mild Cognitive Impairment (MCI) and potentially using it for identifying conditions. Furthermore we aim to assess the reliability of this technique by trying it out on patients, with MRI scans taken from MR machines with varying settings.

References

[1] A. Hussain and A. Khunteta, \"Semantic Segmentation of Brain Tumor from MRI Images and SVM Classification using GLCM Features,\" 2020 Second International Conference on Inventive Research in Computing Applications (ICIRCA), Coimbatore, India, 2020, pp. 38-43, doi: 10.1109/ICIRCA48905.2020.9183385. [2] T. Li and W. Zhang, \"Classification of brain disease from magnetic resonance images based on multi-level brain partitions,\" 2016 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC), Orlando, FL, USA, 2016, pp. 5933-5936, doi: 10.1109/EMBC.2016.7592079. [3] H. Fuse, K. Oishi, N. Maikusa, T. Fukami and J. A. D. N. Initiative, \"Detection of Alzheimer’s Disease with Shape Analysis of MRI Images,\" 2018 Joint 10th International Conference on Soft Computing and Intelligent Systems (SCIS) and 19th International Symposium on Advanced Intelligent Systems (ISIS), Toyama, Japan, 2018, pp. 1031-1034, doi: 10.1109/SCIS-ISIS.2018.00171. [4] S. U. Sadat, H. H. Shomee, A. Awwal, S. N. Amin, M. T. Reza and M. Z. Parvez, \"Alzheimer’s Disease Detection and Classification using Transfer Learning Technique and Ensemble on Convolutional Neural Networks,\" 2021 IEEE International Conference on Systems, Man, and Cybernetics (SMC), Melbourne, Australia, 2021, pp. 1478-1481, doi: 10.1109/SMC52423.2021.9659179. [5] A. Rao, Y. Lee, A. Gass and A. Monsch, \"Classification of Alzheimer’s Disease from structural MRI using sparse logistic regression with optional spatial regularization,\" 2011 Annual International Conference of the IEEE Engineering in Medicine and Biology Society, Boston, MA, USA, 2011, pp. 4499-4502, doi: 10.1109/IEMBS.2011.6091115. [6] Kruthika, K. R., et al. \"Classification of alzheimer and MCI phenotypes on MRI data using SVM.\" Advances in Signal Processing and Intelligent Recognition Systems: Proceedings of Third International Symposium on Signal Processing and Intelligent Recognition Systems (SIRS-2017), September 13-16, 2017, Manipal, India. Springer International Publishing, 2018. [7] Jongkreangkrai, C., et al. \"Computer-aided classification of Alzheimer’s disease based on support vector machine with combination of cerebral image features in MRI.\" Journal of physics: conference series. Vol. 694. No. 1. IOP Publishing, 2016.

Copyright

Copyright © 2024 Samarveer Moray, Aditya Sakhare, , Prathamesh Salokhe, Kedar Sathe. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.