Automatic tool for Alzheimer's disease diagnosis using PCA and Bayesian classification rules

Abstract

An automatic tool to assist the interpretation of single photon emission computed tomography (SPECT) and positron emission tomography (PET) images for the diagnosis of the Alzheimer's disease (AD) is demonstrated. The main problem to be handled is the so-called small size sample, which consists of having a small number of available images compared to the large number of features. This problem is faced by intensively reducing the dimension of the feature space by means of principal component analysis (PCA). Our approach is based on Bayesian classifiers, which uses a posteriori information to determine in which class the subject belongs, yielding 88.6 and 98.3% accuracy values for SPECT and PET images, respectively. These results mean an improvement over the accuracy values reached by other existing techniques.

Feature selection using factor analysis for Alzheimer's diagnosis using PET images

“…In Ref. 35, a CAD system for an automatic evaluation of the neuroimages is presented. In that work, principal component analysis ͑PCA͒ based methods are proposed as feature extraction techniques, enhanced by other linear approaches such as linear discriminant analysis or the measure of the Fisher discriminant ratio for feature selection.…”

On the early diagnosis of Alzheimer's Disease from multimodal signals: A survey

“…Dehghan [83] improved these results combining both FDG and PiB PET scans, and using PCA and SVM algorithms for feature extraction and classification, they achieved 94.12% of accuracy distinguishing AD from healthy controls and 82.05% in the case of MCI and controls. A group of investigators of the University of Granada has published several important works proposing automatic PET based AD diagnosis tools [84][85][86], reporting high accuracies of up to 98.3% distinguishing AD patients and healthy controls, 77.47% separating CTLs from both AD and MCI patients and 68.79% in classifying MCI patients and controls.…”

“…CAD systems have been developed using SPECT images and machine-learning techniques [84,86,94,97,98]. Lopez et al [84] have been able to distinguish AD patients of Alzheimer's Disease neuroimaging initiative (ADNI) database [99] from CTLs with 96.7% accuracy, using PCA based features of preselected slices of interest and an SVM classifier with a quadratic kernel.…”

18F-FDG PET imaging analysis for computer aided Alzheimer’s diagnosis

“…Positive and negative likelihoods (PL/NL) are also displayed as a measure of the positive and negative predictive value of the method, given its prevalence independence. Results using PCA and ICA are compared to VAF [68,67] and PCA-f [35,42,43], selecting the best performance scenarios for kernel SVM and knn classifiers. 1 The training process is showed in detail in Figs.…”