Multi-scale features extraction from baseline structure MRI for MCI patient classification and AD early diagnosis

Abstract

In this study, we investigate multi-scale features extracted from baseline structural magnetic resonance imaging (MRI) for classifying patients with mild cognitive impairment (MCI), who have either converted or not converted to Alzheimer's disease (AD) three years after their baseline visit. Total 549 subjects from the Alzheimer's disease Neuroimaging Initiative (ADNI) database are included, and there are 228 Normal controls (NC), 133 MCI patients (71 of them converted to AD within 3 years, referred as MCI converters, or MCIc) and 188 AD patients. The images are preprocessed with the standard voxel-based morphometry method with segmentation of grey matter, white matter and cerebrospinal fluid. Wavelet frame, a multi-scale image representation approach, is applied to extract features of different scales and directions from the processed grey matter image data. The features are extracted for both whole grey matter images and grey matter images of the hippocampus region. The support vector machine is adopted to construct classifiers for MCIc and MCI non-converters (MCInc). The accuracy using a leave-one-out procedure for classification of AD vs NC and MCIc vs MCInc is 84.13% and 76.69% respectively, both achieved by local hippocampus data. Our study shows that the proposed multi-scale method is capable of discriminating MCI converters and non-converters, and it can be potentially useful for MCI prognosis in clinical applications.

Temporal Correlation Structure Learning for MCI Conversion Prediction

“…To evaluate our Temporal-GAN model, we compare with the following methods: SVM-Linear (support vector machine with linear kernel), which has been widely applied in MCI conversion prediction [6,15]; SVM-RBF (SVM with RBF kernel), as employed in [10,21]; and SVM-Polynomial (SVM with polynomial kernel) as used in [10]. Also, to validate the improvement by learning the temporal correlation structure, we compare with the Neural Network with exactly the same structure in our classification network (network C in Fig.…”

Multi-Domain Transfer Learning for Early Diagnosis of Alzheimer’s Disease

“…To further evaluate the efficacy of our proposed multi-domain transfer learning (MDTL) method for early diagnosis of AD, we list a comparison between the MDTL and some representative state-of-the-art methods in the recent 5 years (Cho et al 2012; Coupé et al 2012; Cuingnet et al 2011; Duchesne and Mouiha 2011; Eskildsen et al 2013; Hu et al 2016; Khedher et al 2015; Liu et al 2014; Moradi et al 2015; Ota et al 2015; Westman et al 2013; Wolz et al 2011; Zhu et al 2014), and show them in Table 8. Here, we provide two performance measurements (i.e., ACC: Accuracy; and AUC: Area Under the receiver operating characteristic Curve) in Table 8.…”

Interpretable Deep Temporal Structure Learning Model for Early Detection of Alzheimer’s Disease

“…Recent advances in machine learning have enabled several new models for MCI conversion prediction. In [17], linear SVM method was employed to distinguish between MCI converters and non-converters, where the extraction of multi-scale features from baseline MRI data strengthened the classification. In [47], the authors used a non-linear classification model to flexibly describe the complex relationship between neuroimaging data and disease status.…”

“…To evaluate our ITGAN model, we compare with the following methods: SVM-Linear(support vector machine with linear kernel), which has been widely applied in MCI conversion prediction [17,34]; SVM-RBF (SVM with RBF kernel), as employed in [25,47]; and SVM-Polynomial (SVM with polynomial kernel) as used in [25]. Also, to validate the improvement by learning the temporal correlation structure, we compare ).…”

“…Neuroimaging provides an effective tool to characterize the structure and functionality of nervous system, thus has greatly contributed to Alzheimer's study [48]. Extensive work has been proposed to predict MCI conversion using neuroimaging data [34,17]. Previous methods usually formulate MCI conversion prediction as a binary classification (distinguishing MCI converters from non-converters) [34,47] or multi-class classification problem (when considering other classes such as AD or health control (HC)) [17,44], where the methods take the neuroimaging data at baseline time as the input and classify if the MCI samples will convert to AD in years.…”