Leung-Malik Features and Adaboost Perform Classification of Alzheimer’s Disease Stages

Abstract

This paper discusses the categorization of brain MRIs as CN, MCI and AD. In all, 200 T2W MRI volumes and 200 T1w MRI volumes from the ADNI dataset are being used for this study. We collected a centre slice of axial orientation from all the respective MRI volumes. The Leung-Malik Filter bank was the source for the extraction of relative texton attributes that are not affected by rotation. The Bag of Dictionary was constructed from LM Filtered images by aggregation and k-means. The collected texton characteristics were used to carry binary classification; in addition to this model the features are used to carry multi-class classification. These are performed using a variety of machine learning classifiers. We trained the models using skull-stripped T1w and T2w MRI images. In relation to other classifiers, skull-removed T2w MRI is used to train with Adaboost Classifier, using decision tree as a meta-classifier-enhanced binary and multi-class classifier performance. The performance of the proposed framework is evaluated in comparison to the various existing works. Different statistical parameters such as sensitivity, specificity, and accuracy are used to evaluate the model performance.

KEYWORDS:

• Adaboost
• Classification
• Feature Extraction
• Machine learning
• Metaclassifier

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Notes on contributors

Shaik Basheera

Shaik Basheera is working as associate professor in the Department of ECE at Chirala Engineering College. He completed his PhD from Acharya Nagarjuna University, master's from Jawaharlal Nehru Technological University, Anantapur, and bachelor's from JNTUH, Hyderabad. His research areas are image processing, machine learning, and artificial intelligence.

M. Satya Sai Ram

M Satya Sai Ram is working as professor in the Department of ECE, RVR & JC College of Engineering. He acts as PhD guide. His research areas are speech processing, signal processing and image processing. He completed his PhD from Jawaharlal Nehru Technological University Hyderabad, India. Email: [email protected]