MRI를 이용한 치매 진단을 위한 패턴인식 기법의 비교평가에 대한 연구

Abstract

As the trend of aging in Korea grows rapidly, age-related diseases are also becoming more serious. There is a type of dementia, Alzheimer’s disease, Frontotemporal Dementia, Parkinson’s Dementia, Lewy Body Dementia, Vascular Dementia, Chronic Tranumatic Encephalopathy. In particular, dementia, especially, Alzheimer’s disease (AD), is one of the increasing concerns because the incidence rate is increasing with age. Depending on the recent researches that AD can be cured if it can be early diagnosed, the importance of early diagnosis of AD are getting more attention. There are several neuroimaging methods such as Positron Emission Tomography(PET) and Magnetic Resonance Imaging(MRI) for diagnosis of AD. The main aim of this research is to distinguish patients using MRI data which are easily affordable to us these days. For this purpose, we compared several classification algorithms in the field of pattern recognition using MRI-specific information. More specifically, for the classification of cognitively normal (CN), mild cognitive impairment (MCI), and AD subjects, we used Principal Component Analysis(PCA), Linear Discriminant Analysis(LDA), Support Vector Machine(SVM), Neural Networks (NNs), and Deep Learning. We briefly analyzed the characteristics of each algorithm, and compared them in the classification problemto find the suitable algorithm for assisting the diagnosis of AD progression. A result indicates the degree to distinguish the cognitively normal(CN), mild cognitive impairment(MCI), and Alzheimer’s disease(AD). The results of the principal component analysis and linear discriminant showed the picture, The result of support vector machine is about 63%, Neural Networks is about 39%, Convolution Neural Networks is about 33%. We still use support vector machine algorithms. But more than 150 features, more than 1000 data is expected that, if I get a good result in Deep Learning.