Deep sparse multi-task learning for feature selection in Alzheimer’s disease diagnosis

Abstract

Recently, neuroimaging-based Alzheimer’s disease (AD) or mild cognitive impairment (MCI) diagnosis has attracted researchers in the field, due to the increasing prevalence of the diseases. Unfortunately, the unfavorable high-dimensional nature of neuroimaging data, but a limited small number of samples available, makes it challenging to build a robust computer-aided diagnosis system. Machine learning techniques have been considered as a useful tool in this respect and, among various methods, sparse regression has shown its validity in the literature. However, to our best knowledge, the existing sparse regression methods mostly try to select features based on the optimal regression coefficients in one step. We argue that since the training feature vectors are composed of both informative and uninformative or less informative features, the resulting optimal regression coefficients are inevidently affected by the uninformative or less informative features. To this end, we first propose a novel deep architecture to recursively discard uninformative features by performing sparse multi-task learning in a hierarchical fashion. We further hypothesize that the optimal regression coefficients reflect the relative importance of features in representing the target response variables. In this regard, we use the optimal regression co-efficients learned in one hierarchy as feature weighting factors in the following hierarchy, and formulate a weighted sparse multi-task learning method. Lastly, we also take into account the distributional characteristics of samples per class and use clustering-induced subclass label vectors as target response values in our sparse regression model. In our experiments on the ADNI cohort, we performed both binary and multi-class classification tasks in AD/MCI diagnosis and showed the superiority of the proposed method by comparing with the state-of-the-art methods.

Deep Residual Learning for Neuroimaging: An application to Predict Progression to Alzheimer’s Disease

“…In this specific analysis, for estimating the performance of the SVM classifier, the classical univariate feature selection procedure using F-test (ANOVA) was implemented for dimension reduction following which the optimal value of the penalty (cost) parameter in the linear SVM was estimated. For the SAE method, we considered three hidden layers and employed a grid search to select the number of units in the intermediate layers based on the results in H. Il Suk et al, 2015. The boxplots for the accuracies for the different cross-validation folds using the Resnet, SAE and SVM models are shown in Figure 6C.…”

Deep ensemble learning of sparse regression models for brain disease diagnosis

“…Based on our earlier work (Suk et al, 2015b; 2016a), where we observed that parameter values of higher than 0.3 were not useful and never chosen in cross-validation, we thus defined the parameter space with 10 values equally spaced between 0.01 and 0.3. As for the size of parameter space ( M = 10), it is determined empirically.…”

Pulmonary nodule diagnosis using dual‐modal supervised autoencoder based on extreme learning machine

“…Accordingly, the issues of feature computing and selection can potentially be addressed by this deep learning framework without a complicated pipeline of image processing steps. Recently, the deep learning technology has been widely used in medical diagnosis (Rodger, ; Suk, Lee, Shen, et al, ; Suk & Shen, ). Hua et al (Hua, Hsu, Hidayati, Cheng, & Chen, ) introduced models of a deep belief network and a convolutional neural network (CNN) in the context of nodule classification in CT images.…”

“…Recently, the deep learning technology has been widely used in medical diagnosis (Rodger, 2015;Suk, Lee, Shen, et al, 2015;Suk & Shen, 2013). Hua et al (Hua, Hsu, Hidayati, Cheng, & Chen, 2015) introduced models of a deep belief network and a convolutional neural network (CNN) in the context of nodule classification in CT images.…”