Alzheimer's disease (AD) is a neurodegenerative disorder that affects the life quality of millions of people worldwide. To diagnose new cases in a timely manner, we propose a new novelty detection technique that combines Autoencoder and Minimum Covariance Determinant (MCD). The technique consists of two steps: first, we use an Autoencoder to extract low-dimensional and discriminative features from the publicly available ADNI dataset, where we only train the Autoencoder with normal data, making the abnormal data more distinguishable in the feature space; second, based on the features of normal data, we use MCD to construct a decision boundary, and judge the degree of abnormality by the distance of the test point to the boundary. Compared with traditional methods without using Autoencoder, our technique has significant advantages in terms of accuracy and sensitivity, and can effectively deal with data imbalance problem. Experimental results show that our method can efficiently detect novel AD cases, and has a wide range of application prospects.
Artificial intelligence techniques have been deeply involved in the heterogeneous data aspects of biomedical applications. However, the high dimensionality and computational complexity of data can make classification, pattern recognition and data visualization difficult. Choosing appropriate dimensionality reduction techniques can help increase processing speed, reduce the time and effort required to extract valuable information, and ensure high accuracy. In this study, Alzheimer's disease data were taken as an example. Individual cases with missing values were removed, and non-digital data were converted to digital data using Min-Max normalization. Then principal component analysis (PCA) was applied to map the original feature space to 1 dimension and the variance of the validation set was calculated by 5-fold cross-validation to find the appropriate K value. The results showed that when PCA was applied to reduce the data to 1 dimension, the AUC (95% confidence interval) of the KNN classifier reached 0.898 ± 0.014, which was 30.4%higher than the case without PCA. Our current findings suggest that in many busy clinics and hospitals, it is quite worthwhile to use dimensionality reduction methods to save model computing time and to use KNN models to obtain better accuracy.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.