Effective diagnosis of Alzheimer's disease by means of large margin-based methodology

Abstract

Background: Functional brain images such as Single-Photon Emission Computed Tomography (SPECT) and Positron Emission Tomography (PET) have been widely used to guide the clinicians in the Alzheimer's Disease (AD) diagnosis. However, the subjectivity involved in their evaluation has favoured the development of Computer Aided Diagnosis (CAD) Systems.

Methods: It is proposed a novel combination of feature extraction techniques to improve the diagnosis of AD. Firstly, Regions of Interest (ROIs) are selected by means of a t-test carried out on 3D Normalised Mean Square Error (NMSE) features restricted to be located within a predefined brain activation mask. In order to address the small sample-size problem, the dimension of the feature space was further reduced by: Large Margin Nearest Neighbours using a rectangular matrix (LMNN-RECT), Principal Component Analysis (PCA) or Partial Least Squares (PLS) (the two latter also analysed with a LMNN transformation). Regarding the classifiers, kernel Support Vector Machines (SVMs) and LMNN using Euclidean, Mahalanobis and Energy-based metrics were compared.

Results: Several experiments were conducted in order to evaluate the proposed LMNN-based feature extraction algorithms and its benefits as: i) linear transformation of the PLS or PCA reduced data, ii) feature reduction technique, and iii) classifier (with Euclidean, Mahalanobis or Energy-based methodology). The system was evaluated by means of k-fold cross-validation yielding accuracy, sensitivity and specificity values of 92.78%, 91.07% and 95.12% (for SPECT) and 90.67%, 88% and 93.33% (for PET), respectively, when a NMSE-PLS-LMNN feature extraction method was used in combination with a SVM classifier, thus outperforming recently reported baseline methods.

Conclusions: All the proposed methods turned out to be a valid solution for the presented problem. One of the advances is the robustness of the LMNN algorithm that not only provides higher separation rate between the classes but it also makes (in combination with NMSE and PLS) this rate variation more stable. In addition, their generalization ability is another advance since several experiments were performed on two image modalities (SPECT and PET).

Figure 1

Feature extraction and classification diagram. Voxels and Features, combination of feature reduction techniques and classifiers evaluated with k-Fold cross validation.

Figure 2

Feature/Model Selection by means of Variance Explained. Variance Explained (%) versus PCA and PLS Components in bar diagram. Lines represent the accumulated Variance Explained (%) versus Principal Components and PLS Components.

Figure 3

SVM classification: Accuracy, Specificity and Sensitivity (%) versus number of reduced features for SPECT. Feature reduction techniques: a) PCA, b)PLS, c)LMNN-RECT.

Figure 4

LMNN classification (Euclidean, Mahalanobis and Energy-based models) for SPECT. Feature reduction techniques: a)PCA, b)PLS.

Figure 5

a) Kernel SVM for PLS features LMNN transformed b) linear SVM over PCA features directly (reduced to the half VAF) obtained for SPECT.

Figure 6

a) SVM classification: Accuracy, Specificity and Sensitivity (%) versus number of reduced features for PET database. Feature reduction techniques: PCA plus LMNN Transformation, PLS plus LMNN Transformation, LMNN-RECT, PCA and VAF b) LMNN classification (Euclidean, Mahalanobis and Energy-based models) for PCA and PLS features.

Figure 7

ROC Analysis. LMNN-based methods SVM classified (PCA+LMNN Transformation, PLS+LMNN Transformation, LMNN-RECT). Comparison to other recently reported methods represented by their operation points. a) SPECT database. The AUC obtained for each ROC is: PCA+LMNN (0.9411), PLS+LMNN (0.9424), LMNN-RECT (0.9076), VAF SVM (0.8993) and PCA SVM (0.9177). b)PET database. The AUC obtained for each ROC is: PCA+LMNN (0.9505), PLS+LMNN (0.9437), LMNN-RECT (0.9325), VAF SVM (0.8500) and PCA SVM (0.9006).