Patients with diabetes must continually monitor their blood glucose levels and adjust insulin doses, striving to keep blood glucose levels as close to normal as possible. Blood glucose levels that deviate from the normal range can lead to serious short-term and long-term complications. An automatic prediction model that warned people of imminent changes in their blood glucose levels would enable them to take preventive action. In this paper, we describe a solution that uses a generic physiological model of blood glucose dynamics to generate informative features for a Support Vector Regression model that is trained on patient specific data. The new model outperforms diabetes experts at predicting blood glucose levels and could be used to anticipate almost a quarter of hypoglycemic events 30 minutes in advance. Although the corresponding precision is currently just 42%, most false alarms are in near-hypoglycemic regions and therefore patients responding to these hypoglycemia alerts would not be harmed by intervention.

As the amount of data collected from individuals increases, there are more opportunities to use it to offer personalized experiences (e.g., using electronic health records to offer personalized treatments). We advocate applying techniques from batch reinforcement learning to predict the range of effectiveness that policies might have for individuals. We identify three sources of uncertainty and present a method that addresses all of them. It handles the uncertainty caused by population mismatch by modeling the data as a latent mixture of different subpopulations of individuals, it explicitly quantifies data sparsity by accounting for the limited data available about the underlying models, and incorporates intrinsic stochasticity to yield estimated percentile ranges of the effectiveness of a policy for a particular new individual. Using this approach, we highlight some interesting variability in policy effectiveness amongst HIV patients given a prior patient treatment dataset. Our approach highlights the potential benefit of taking into account individual variability and data limitations when performing batch policy evaluation for new individuals.

Convolutional neural networks (CNNs) are flexible, biologically-inspired variants of multi-layer perceptrons that have proven themselves to be exceptionally suited to discriminative vision tasks. However, relatively little is known on whether they can be made both more efficient and more accurate, by introducing suitable transformations that exploit general knowledge of the target classes. We demonstrate this functionality through pre-segmentation of input images with a fast and robust but loose segmentation step, to obtain a set of candidate objects. These objects then undergo a spatial transformation into a reduced space, retaining but a compact high-level representation of their appearance. Additional attributes may be abstracted as raw features that are incorporated after the convolutional phase of the network. Finally, we compare its performance against existing approaches on the challenging problem of detecting lesions in retinal images.

Resting state functional magnetic resonance imaging (rsfMRI) is a powerful tool for investigating intrinsic and spontaneous brain activity. The application of univariate and multivariate methods such as multi voxel pattern analysis has been instrumental in localizing neural correlates to various cognitive states and psychiatric disease. However, many existing methods of rsfMRI analysis are insufficient for investigating the true mechanism of brain activity since they make implicit assumptions that are agnostic of the temporal and spatial dynamics of brain activity. The proposed method aims to create a superior feature space for representing brain activity using k-means and to create interpretable generalizations on these features for studying group differences using support vector machine classifiers.​

Social robots can be used to tutor children in one-on-one interactions. Because students have different learning needs, they consequently require complex, non-scripted teaching behaviors that adapt to the learning needs of each child. As a result of this, robot tutors are more effective given a means of adaptively customizing the pace and content of a student's curriculum. In this paper we propose a reinforcement learning-based approach that affords such capabilities to a tutoring robot, with the goals of fostering measurable learning gains and sustained engagement. We outline an architecture in which the robot uses reinforcement learning to adapt the difficulty of its exercises. Further, we describe a proposed study capable of evaluating the effectiveness of our Intelligent Tutoring System.

This paper examines classification of shots in video streams for indexing and semantic analysis. We describe an approach to obtain shot motion by making use of motion estimation algorithms to estimate camera movement. We improve prior work by using the four edge regions of a frame to classify No Motion shots. We analyze a neighborhood of shots and provide a new concept, middle-scale classification. This approach relies on automated labeling of frame transitions in terms of motion across adjacent frames. These annotations form a sequential scene-group that correlates with narrative events in the videos. We introduce six middle-scale classes and the corresponding likely sequence content from three clips of the movie The Lord of the Rings : The Return of the King , demonstrate that the middle-scale classification approach successfully extracts a summary of the salient aspects of the movie. We also show direct comparison with prior work on the full movie Matrix .

We present several epistemic views of ideation in scientific discovery that we have investigated: conceptual classification, abductive explanation, conceptual modeling, analogical reasoning, and visual reasoning. We then describe an experiment in computational ideation through model construction, evaluation and revision. We describe an interactive tool called MILA–S that enables construction of conceptual models of ecological phenomena, agent-based simulations of the conceptual model, and revision of the conceptual model based on the results of the simulation.   The key feature of MILA–S is that it automatically generates the simulations from the conceptual model. We report on a pilot study with 50 middle school science students who used MILA–S to discover causal explanations for an ecological phenomenon. Initial results from the study indicate that use of MILA–S had a significant impact both on the process of model construction and the nature of the constructed models.  We posit that MILA–S may enable scientists to construct, evaluate and revise conceptual models of ecological phenomena.

It is generally challenging to obtain the exact disease prevalence, as the true cases of a disease in the population level are not easy to identify. Available and relevant data sources such as administrative or clinical health data are used in public health surveillance as a proxy to estimate the disease prevalence. Traditionally, these data sources span through healthcare utilization information such as emergency department visits, pharmacy drug sales, or laboratory test orders. In addition to incompleteness, these data sources are not usually available in a timely manner. Timeliness is an important factor for prevalence estimation for some conditions such as infectious diseases, especially at the time of an epidemic. For instance, in an influenza pandemic such estimates must be obtained within a day or two. In recent years several non-clinical and non-traditional data sources have been introduced to public health with the potentials to provide signals on a disease rate or to provide a feedback on the trends of a disease. Ideally, combining these new sources with the ones routinely used should help to identify disease cases more efficiently. However, building a construct capable of incorporating data from these various sources in a coherent manner is not trivial. In this research, we consider the case of H1N1 pandemic as the infectious disease of interest and we use media reports of deaths from H1N1 on the web as a non traditional data source. We propose to use dynamic Bayesian networks from the class of probabilistic graphical models in order to combine this new data source with traditional ones through exploration of the possible probabilistic relationships between these data streams. This is an initial step towards building a framework that can potentially support aggregation of heterogeneous data for a real-time estimation of a disease prevalence. Our preliminary results show that the proposed model generalizes well.

Nowadays, many health groups and forums are established on the Internet, where health consumers discuss health issues and interact with each other. Although there is a large amount of user generated content about healthcare on different social media sites, few studies have applied data mining or artificial intelligence techniques for knowledge discovery on a large scale of data in this particular emerging area. In online health forums, it is difficult for users to find relevant topics or peers due to the large amount of information. Traditional recommendation systems may not work well for health online forums, because health consumers have different intentions of participation or may be interest in different types of supports even if the content matches their interest. To help solving this problem, we apply Naïve Bayes methods in this study to classify posts and comments on QuitStop forum, which is an online community for smoking cessation intervention. Classifiers are built on different text features and health features of user quit status. Two different classification tasks are investigated: (1) classification of user intentions, and (2) classification of types of social support exchanged in interactions. We developed classifiers for posts and comments separately, and conducted experiments to compare classifiers with different text and health feature sets. It is found that using thread title or post content can achieve the highest classification accuracy on both posts and comments for user intention classification with text features. On the other hand, using the content of post or comment itself performs the best for the classification of social support types. In particular for the post, integrating health features of the post author can boost the text classifications of user intention and support type. However, user health features cannot help in improving text classifiers for the comments.

Improving public health is a major responsibility of any government, and is of major interest to citizens and scientific communities around the world. Here, one sees two extremes. On one hand, tremendous progress has been made in recent years in the understanding of causes, spread and remedies of common and regularly occurring diseases like Dengue, Malaria and Japanese Encephalistis (JE). On the other hand, public agencies treat these diseases in an ad hoc manner without learning from the experiences of previous years. Specifically, they would get alerted once reported cases have already arisen substantially in the known disease season, reactively initiate a few actions and then document the disease impact (cases, deaths) for that period, only to forget this learning in the next season. However, they miss the opportunity to reduce preventable deaths and sickness, and their corresponding economic impact, which scientific progress could have enabled. The gap is universal but very prominent in developing countries like India.   In this paper, we show that if public agencies provide historical disease impact information openly, it can be analyzed with statistical and machine learning techniques, correlated with best emerging practices in disease control, and simulated in a setting to optimize social benefits to provide timely guidance for new disease seasons and regions. We illustrate using open data for mosquito-borne communicable diseases; published results in public health on efficacy of Dengue control methods and apply it on a simulated typical city for maximal benefits with available resources. The exercise helps us further suggest strategies for new regions that may be anywhere in the world, how data could be better recorded by city agencies and what prevention methods should medical community focus on for wider impact.