It is of great significance to apply deep learning for the early diagnosis of Alzheimer's Disease (AD). In this work, a novel tensorizing GAN with high-order pooling is proposed to assess Mild Cognitive Impairment (MCI) and AD. By tensorizing a three-player cooperative game based framework, the proposed model can benefit from the structural information of the brain. By incorporating the high-order pooling scheme into the classifier, the proposed model can make full use of the second-order statistics of the holistic Magnetic Resonance Imaging (MRI) images. To the best of our knowledge, the proposed Tensor-train, High-pooling and Semi-supervised learning based GAN (THS-GAN) is the first work to deal with classification on MRI images for AD diagnosis. Extensive experimental results on Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset are reported to demonstrate that the proposed THS-GAN achieves superior performance compared with existing methods, and to show that both tensor-train and high-order pooling can enhance classification performance. The visualization of generated samples also shows that the proposed model can generate plausible samples for semi-supervised learning purpose.

Tensorizing GAN with High-Order Pooling for Alzheimer's Disease Assessment

It is of great significance to apply deep learning for the early diagnosis of Alzheimer's disease (AD). In this work, a novel tensorizing GAN with high-order pooling is proposed to assess mild cognitive impairment (MCI) and AD. By tensorizing a three-player cooperative game-based framework, the proposed model can benefit from the structural information of the brain. By incorporating the high-order pooling scheme into the classifier, the proposed model can make full use of the second-order statistics of holistic magnetic resonance imaging (MRI). To the best of our knowledge, the proposed Tensor-train, High-order pooling and Semisupervised learning-based GAN (THS-GAN) is the first work to deal with classification on MR images for AD diagnosis. Extensive experimental results on Alzheimer's disease neuroimaging initiative (ADNI) data set are reported to demonstrate that the proposed THS-GAN achieves superior performance compared with existing methods, and to show that both tensor-train and high-order pooling can enhance classification performance. The visualization of generated samples also shows that the proposed model can generate plausible samples for semisupervised learning purpose. 

/pdf/tensorizing-gan-with-high-order-pooling-for-alzheimers-2uis8knz.pdf

Tensorizing GAN With High-Order Pooling for Alzheimer’s Disease Assessment

Computer-aided detection and diagnosis in ECG signals for heart diseases are gaining increasing attention. However, developing and selecting the highly performing diagnostic model suitable for clinical implications is still challenging. In this paper, we proposed a combined network of convolutional neural network (CNN) and Recurrent Neural Network (RNN), designed for the classification of ECG heart signals for diagnostic purpose. The proposed network consists of 2 convolutional layers with 5 × 5 kernels and ReLU activations, followed by 4 residual blocks, 2 bidirectional long short-term memory (biLSTM) layers, as well as 2 fully connected layers. Each residual block involved the structure of a Squeeze-and-Excitation Network (SENet) with lightweight property to recalibrate the feature map of the network. The last dense layer has 5 outputs, equivalent to the classes considered: non-ectopic beats, supraventricular ectopic beats, ventricular ectopic beats, fusion beats, and unknown beats. To train and evaluate the combined CNN and RNN, we transferred the knowledge acquired on beat classification tasks in 2017 PhysioNet/CinC Challenge to that in PhysionNet's MIT-BIH dataset. The developed network achieved a recognition sensitivity of 95.90%, accuracy = 95.90% and specificity = 96.34% with classification time of single sample = 6.23 s in detecting 5 ECG classes. A comparative analysis proved the high performance of the proposed combined CNN and RNN against previous methods, demonstrating the potential of our proposed network in the analysis of beat patterns. The proposed model can be applied in cloud computing or implemented in mobile devices to evaluate cardiac health with the highest precision.

/pdf/interpretation-of-electrocardiogram-ecg-rhythm-by-combined-2448olmg6t.pdf

Interpretation of Electrocardiogram (ECG) Rhythm by Combined CNN and BiLSTM

Drug-target interaction prediction is of great significance for narrowing down the scope of candidate medications, and thus is a vital step in drug discovery. Because of the particularity of biochemical experiments, the development of new drugs is not only costly, but also time-consuming. Therefore, the computational prediction of drug target interactions has become an essential way in the process of drug discovery, aiming to greatly reducing the experimental cost and time. We propose a learning-based method based on feature representation learning and deep neural network named DTI-CNN to predict the drug-target interactions. We first extract the relevant features of drugs and proteins from heterogeneous networks by using the Jaccard similarity coefficient and restart random walk model. Then, we adopt a denoising autoencoder model to reduce the dimension and identify the essential features. Third, based on the features obtained from last step, we constructed a convolutional neural network model to predict the interaction between drugs and proteins. The evaluation results show that the average AUROC score and AUPR score of DTI-CNN were 0.9416 and 0.9499, which obtains better performance than the other three existing state-of-the-art methods. All the experimental results show that the performance of DTI-CNN is better than that of the three existing methods and the proposed method is appropriately designed.

/pdf/a-learning-based-method-for-drug-target-interaction-4ltsxp0ilq.pdf

A learning-based method for drug-target interaction prediction based on feature representation learning and deep neural network.

A CAD system for diagnosing Alzheimer’s disease using 2D slices and an improved AlexNet-SVM method

Convolutional neural networks (CNNs)-based classifiers improve the accuracy of diagnosis and prediction for Alzheimer's disease (AD). However, exploiting specific brain regions with the AD is essential to understand pathological alteration in the AD and monitor its progression. This paper aims to construct novel AD classification models which have a good performance and interpretation on AD diagnosis. We propose the three classifiers including a simple broaden plain CNNs (SBPCNNs), a major slice-assemble CNNs (SACNNs) and a multi-slice CNNs (MSCNNs), which record the slice positions but have fewer parameters. Specifically, we integrate the ranking and the random forest methods to find the discriminative region that is consistent with domain knowledge about the AD. The results of the visualization explanation of pixel and slice level deliver a clearer understanding of the AD to specialists. The experimental results indicate that the proposed models are meaningful for AD classification.

/pdf/exploiting-discriminative-regions-of-brain-slices-based-on-4pskyedmci.pdf

Exploiting Discriminative Regions of Brain Slices Based on 2D CNNs for Alzheimer’s Disease Classification

The subjective cognitive decline (SCD) may last for decades prior to the onset of dementia and has been proposed as a risk population for development to amnestic mild cognitive impairment (aMCI) and Alzheimer disease (AD). Disruptions of functional connectivity and causal connectivity (CC) in the salience network (SN) are generally perceived as prominent hallmarks of the preclinical AD. Nevertheless, the alterations in anterior SN (aSN), and posterior SN (pSN) remain unclear. Here, we hypothesized that both the functional connectivity (FC) and CC of the SN subnetworks, comprising aSN and pSN, were distinct disruptive in the SCD and aMCI. We utilized resting-state functional magnetic resonance imaging to investigate the altered FC and CC of the SN subnetworks in 28 healthy controls, 23 SCD subjects, and 29 aMCI subjects. In terms of altered patterns of FC in SN subnetworks, aSN connected to the whole brain was significantly increased in the left orbital superior frontal gyrus, left insula lobule, right caudate lobule, and left rolandic operculum gyrus (ROG), whereas decreased FC was found in the left cerebellum superior lobule and left middle temporal gyrus when compared with the HC group. Notably, no prominent statistical differences were obtained in pSN. For altered patterns of CC in SN subnetworks, compared to the HC group, the aberrant connections in aMCI group were separately involved in the right cerebellum inferior lobule (CIL), right supplementary motor area (SMA), and left ROG, whereas the SCD group exhibited more regions of aberrant connection, comprising the right superior parietal lobule, right CIL, left inferior parietal lobule, left post-central gyrus (PG), and right angular gyrus. Especially, SCD group showed increased CC in the right CIL and left PG, whereas the aMCI group showed decreased CC in the left pre-cuneus, corpus callosum, and right SMA when compared to the SCD group. Collectively, our results suggest that analyzing the altered FC and CC observed in SN subnetworks, served as impressible neuroimaging biomarkers, may supply novel insights for designing preclinical interventions in the preclinical stages of AD.

Altered Patterns of Functional Connectivity and Causal Connectivity in Salience Subnetwork of Subjective Cognitive Decline and Amnestic Mild Cognitive Impairment

Alzheimer's disease (AD), which most commonly occurs in elder， is a chronic neurodegenerative disease with no agreed drugs or treatment protocols at present. Amnestic mild cognitive impairment (aMCI), earlier than AD onset and later than subjective cognitive decline (SCD) onset, has a serious probability of converting into AD. The SCD, which can last for decades, subjectively complains of decline impairment especially in memory. Distinct altered patterns of default mode network (DMN) subnetworks connected to the whole brain are perceived as prominent hallmarks of the early stages of AD. Nevertheless, the aberrant phase position connectivity (PPC) connected to the whole brain in DMN subnetworks still remains unknown. Here, We hypothesized that there exist distinct variations of PPC in DMN subnetworks connected to the whole brain for patients with SCD and aMCI, which might be acted as discriminatory neuroimaging biomarkers. We recruited 27 healthy controls (HC), 20 SCD and 28 aMCI subjects, respectively, to explore aberrant patterns of PPC in DMN subnetworks connected to the whole brain. In anterior DMN (aDMN), SCD group exhibited aberrant PPC in the regions of right superior cerebellum lobule (SCL), right superior frontal gyrus of medial part (SFGMP) and left fusiform gyrus (FG) in comparison of HC group, by contrast, no prominent difference was found in aMCI group. It is important to note that aMCI group showed increased PPC in the right SFGMP in comparison with SCD group. For posterior DMN (pDMN), SCD group showed decreased PPC in the left superior parietal lobule (SPL) and right superior frontal gyrus (SFG) compared to HC group. It is noteworthy that aMCI group showed decreased PPC in the left middle frontal gyrus of orbital part (MFGOP) and right SFG compared to HC group, yet increased PPC was found in the left superior temporal gyrus of temporal pole (STGTP). Additionally, aMCI group exhibited decreased PPC in the left MFGOP compared to SCD group. Collectively, our results have shown that the aberrant regions of PPC observed in DMN are related to cognitive function, and it might also be served as impressible neuroimaging biomarkers for timely intervention before AD occurs.

/pdf/altered-patterns-of-phase-position-connectivity-in-default-45vwfe61a5.pdf

Fujia Ren

Papers

Exploiting Discriminative Regions of Brain Slices Based on 2D CNNs for Alzheimer’s Disease Classification

Altered Patterns of Functional Connectivity and Causal Connectivity in Salience Subnetwork of Subjective Cognitive Decline and Amnestic Mild Cognitive Impairment

Altered Patterns of Phase Position Connectivity in Default Mode Subnetwork of Subjective Cognitive Decline and Amnestic Mild Cognitive Impairment.

MRI-based model for MCI conversion using deep zero-shot transfer learning