Background: Autism spectrum disorders (ASD) imply a spectrum of symptoms rather than a single phenotype. ASD could affect brain connectivity at different degree based on the severity of the symptom. Given their excellent learning capability, graph neural networks (GNN) methods have recently been used to uncover functional connectivity patterns and biological mechanisms in neuropsychiatric disorders, such as ASD. However, there remain challenges to develop an accurate GNN learning model and understand how specific decisions of these graph models are made in brain network analysis. Results: In this paper, we propose a graph attention network based learning and interpreting method, namely GAT-LI, which learns to classify functional brain networks of ASD individuals versus healthy controls (HC), and interprets the learned graph model with feature importance. Specifically, GAT-LI includes a graph learning stage and an interpreting stage. First, in the graph learning stage, a new graph attention network model, namely GAT2, uses graph attention layers to learn the node representation, and a novel attention pooling layer to obtain the graph representation for functional brain network classification. We experimentally compared GAT2 model’s performance on the ABIDE I database from 1035 subjects against the classification performances of other well-known models, and the results showed that the GAT2 model achieved the best classification performance. We experimentally compared the influence of different construction methods of brain networks in GAT2 model. We also used a larger synthetic graph dataset with 4000 samples to validate the utility and power of GAT2 model. Second, in the interpreting stage, we used GNNExplainer to interpret learned GAT2 model with feature importance. We experimentally compared GNNExplainer with two well-known interpretation methods including Saliency Map and DeepLIFT to interpret the learned model, and the results showed GNNExplainer achieved the best interpretation performance. We further used the interpretation method to identify the features that contributed most in classifying ASD versus HC. Conclusion: We propose a two-stage learning and interpreting method GAT-LI to classify functional brain networks and interpret the feature importance in the graph model. The method should also be useful in the classification and interpretation tasks for graph data from other biomedical scenarios.

Deep learning models have been successfully applied to the analysis of various functional MRI data. Convolutional neural networks (CNN), a class of deep neural networks, have been found to excel at extracting local meaningful features based on their shared-weights architecture and space invariance characteristics. In this study, we propose M2D CNN, a novel multichannel 2D CNN model, to classify 3D fMRI data. The model uses sliced 2D fMRI data as input and integrates multichannel information learned from 2D CNN networks. We experimentally compared the proposed M2D CNN against several widely used models including SVM, 1D CNN, 2D CNN, 3D CNN, and 3D separable CNN with respect to their performance in classifying task-based fMRI data. We tested M2D CNN against six models as benchmarks to classify a large number of time-series whole-brain imaging data based on a motor task in the Human Connectome Project (HCP). The results of our experiments demonstrate the following: (i) convolution operations in the CNN models are advantageous for high-dimensional whole-brain imaging data classification, as all CNN models outperform SVM; (ii) 3D CNN models achieve higher accuracy than 2D CNN and 1D CNN model, but 3D CNN models are computationally costly as any extra dimension is added in the input; (iii) the M2D CNN model proposed in this study achieves the highest accuracy and alleviates data overfitting given its smaller number of parameters as compared with 3D CNN.

The Named Entity Recognition (NER) task as a key step in the extraction of health information, has encountered many challenges in Chinese Electronic Medical Records (EMRs). Firstly, the casual use of Chinese abbreviations and doctors’ personal style may result in multiple expressions of the same entity, and we lack a common Chinese medical dictionary to perform accurate entity extraction. Secondly, the electronic medical record contains entities from a variety of categories of entities, and the length of those entities in different categories varies greatly, which increases the difficult in the extraction for the Chinese NER. Therefore, the entity boundary detection becomes the key to perform accurate entity extraction of Chinese EMRs, and we need to develop a model that supports multiple length entity recognition without relying on any medical dictionary. In this study, we incorporate part-of-speech (POS) information into the deep learning model to improve the accuracy of Chinese entity boundary detection. In order to avoid the wrongly POS tagging of long entities, we proposed a method called reduced POS tagging that reserves the tags of general words but not of the seemingly medical entities. The model proposed in this paper, named SM-LSTM-CRF, consists of three layers: self-matching attention layer – calculating the relevance of each character to the entire sentence; LSTM (Long Short-Term Memory) layer – capturing the context feature of each character; CRF (Conditional Random Field) layer – labeling characters based on their features and transfer rules. The experimental results at a Chinese EMRs dataset show that the F1 value of SM-LSTM-CRF is increased by 2.59% compared to that of the LSTM-CRF. After adding POS feature in the model, we get an improvement of about 7.74% at F1. The reduced POS tagging reduces the false tagging on long entities, thus increases the F1 value by 2.42% and achieves an F1 score of 80.07%. The POS feature marked by the reduced POS tagging together with self-matching attention mechanism puts a stranglehold on entity boundaries and has a good performance in the recognition of clinical entities.

Deep neural networks have recently been applied to the study of brain disorders such as autism spectrum disorder (ASD) with great success. However, the internal logics of these networks are difficult to interpret, especially with regard to how specific network architecture decisions are made. In this paper, we study an interpretable neural network model as a method to identify ASD participants from functional magnetic resonance imaging (fMRI) data and interpret results of the model in a precise and consistent manner. First, we propose an interpretable fully connected neural network (FCNN) to classify two groups, ASD versus healthy controls (HC), based on input data from resting-state functional connectivity (rsFC) between regions of interests (ROIs). The proposed FCNN model is a piecewise linear neural network (PLNN) which uses piecewise linear function LeakyReLU as its activation function. We experimentally compared the FCNN model against widely used classification models including support vector machine (SVM), random forest, and two new classes of deep neural network models in a large dataset containing 871 subjects from ABIDE I database. The results show the proposed FCNN model achieves the highest classification accuracy. Second, we further propose an interpreting method which could explain the trained model precisely with a precise linear formula for each input sample and decision features which contributed most to the classification of ASD versus HC participants in the model. We also discuss the implications of our proposed approach for fMRI data classification and interpretation.

Background: With the rapid development of sequencing technologies, collecting diverse types of cancer omics data become more cost-effective. Many computational methods attempted to represent and fuse multiple omics into a comprehensive view of cancer. However, different types of omics are related and heterogeneous. Most of the existing methods do not consider the difference between omics, so the biological knowledge of individual omics may not be fully excavated. And for a given task (e.g. predicting overall survival), these methods prefer to use sample similarity or domain knowledge to learn a more reasonable representation of omics, but it’s not enough. Methods: For the purpose of learning more useful representation for individual omics and fusing them to improve the prediction ability, we proposed an autoencoder-based method named MOSAE (Multi-omics Supervised Autoencoder). In our method, a specific autoencoder were designed for each omics according to their size of dimension to generate omics-specific representations. Then, a supervised autoencoder was constructed based on specific autoencoder by using labels to enforce each specific autoencoder to learn both omics-specific and task-specific representations. Finally, representations of different omics that generate from supervised autoencoders were fused in a traditional but powerful way, and the fused representation was used for subsequent predictive tasks. Results: We applied our method over TCGA Pan-Cancer dataset to predict four different clinical outcome endpoints (OS, PFI, DFI, and DSS). Compared with traditional and state-of-the-art methods, MOSAE achieved better predictive performance. We also tested the effects of each improvement, which all have a positive effect on predictive performance. Conclusions: Predicting clinical outcome endpoints are very important for precision medicine and personalized medicine. And multi-omics fusion is an effective way to solve this problem. MOSAE is a powerful multi-omics fusion method, which can generate both omics-specific and task-specific representation for given endpoint predictive tasks and improve the predictive performance.