Qiuling Suo

Risk Factor Analysis Based on Deep Learning Models

Accurate rendering of diagnosis and prognosis for a disease with respect to a patient requires analysis of complicated, diverse, yet correlated risk factors (RFs). Most of the existing methods for this purpose are based on handcraft RFs by calculating their statistical significance to the disease. However, such methods not only incur intensive labor but also lack capability to discover or infer previously unknown complex relationships and combined effects among correlated RFs. Nowadays, deep learning models have emerged as a hot topic, due to its ability to automatically extract useful and complex features from raw data. In this paper, we explore the effectiveness of deep learning on medical data by building a deep learning based framework to analyze risk factors and study its prediction performance in disease diagnosis. Specifically, we investigate the application of deep learning with a special focus on interpreting the latent features extracted or created from raw data by the model. Experimental results demonstrate that deep learning based methods are able to aggregate features sharing same characteristics, and reduce effects from unimportant and uncorrelated RFs. The abstract features obtained by deep learning methods can represent the essentials of raw inputs, and give a good prediction performance in disease diagnosis.

Risk Prediction on Electronic Health Records with Prior Medical Knowledge

Predicting the risk of potential diseases from Electronic Health Records (EHR) has attracted considerable attention in recent years, especially with the development of deep learning techniques. Compared with traditional machine learning models, deep learning based approaches achieve superior performance on risk prediction task. However, none of existing work explicitly takes prior medical knowledge (such as the relationships between diseases and corresponding risk factors) into account. In medical domain, knowledge is usually represented by discrete and arbitrary rules. Thus, how to integrate such medical rules into existing risk prediction models to improve the performance is a challenge. To tackle this challenge, we propose a novel and general framework called PRIME for risk prediction task, which can successfully incorporate discrete prior medical knowledge into all of the state-of-the-art predictive models using posterior regularization technique. Different from traditional posterior regularization, we do not need to manually set a bound for each piece of prior medical knowledge when modeling desired distribution of the target disease on patients. Moreover, the proposed PRIME can automatically learn the importance of different prior knowledge with a log-linear model.Experimental results on three real medical datasets demonstrate the effectiveness of the proposed framework for the task of risk prediction

Metric Learning from Probabilistic Labels

Metric learning aims to learn a good distance metric that can capture the relationships among instances, and its importance has long been recognized in many fields. In the traditional settings of metric learning, an implicit assumption is that the associated labels of the instances are deterministic. However, in many real-world applications, the associated labels come naturally with probabilities instead of deterministic values. Thus, the existing metric learning methods cannot work well in these applications. To tackle this challenge, in this paper, we study how to effectively learn the distance metric from datasets that contain probabilistic information, and then propose two novel metric learning mechanisms for two types of probabilistic labels, i.e., the instance-wise probabilistic label and the group-wise probabilistic label. Compared with the existing metric learning methods, our proposed mechanisms are capable of learning distance metrics directly from the probabilistic labels with high accuracy. We also theoretically analyze the two proposed mechanisms and provide theoretical bounds on the sample complexity for both of them. Additionally, extensive experiments based on real-world datasets are conducted to verify the desirable properties of the proposed mechanisms.

Wave2Vec: Deep representation learning for clinical temporal data

Abstract Representation learning for time series has gained increasing attention in healthcare domain. The recent advancement in semantic learning allows researcher to learn meaningful deep representations of clinical medical concepts from Electronic Health Records (EHRs). However, existing models cannot deal with continuous physiological records, which are often included in EHRs. The major challenges for this task are to model non-obvious representations from observed high-resolution biosignals, and to interpret the learned features. To address these issues, we propose Wave2Vec , an end-to-end deep representation learning model, to bridge the gap between biosignal processing and semantic learning. Wave2Vec not only jointly learns both inherent and temporal representations of biosignals, but also allows us to interpret the learned representations reasonably over time. We propose two embedding mechanisms to capture the temporal knowledge within signals, and discover latent knowledge from signals in time-frequency domain, namely component-based motifs. To validate the effectiveness of our model in clinical task, we carry out experiments on two real-world benchmark biosignal datasets. Experimental results demonstrate that the proposed Wave2Vec model outperforms six feature learning baselines in biosignal processing. Analytical results show that the proposed model can incorporate both motif co-occurrence information and time series information of biosignals, and hence provides clinically meaningful interpretation.