Publications

This paper discusses the problem of abstract­ ing conditional probabilistic actions. We identify two distinct types of abstraction: intra­action abstraction and inter­action ab­ straction. We define what it means for the abstraction of an action to be correct and then derive two methods of intra­action ab­ straction and two methods of inter­action ab­ straction which are correct according to this criterion. We illustrate the developed tech­ niques by applying them to actions described with the temporal action representation used in the drips decision­theoretic planner and we describe how the planner uses abstraction to reduce the complexity of planning.

Sketching is ubiquitous in medicine. Physicians commonly use sketches as part of their note taking in patient records and to help convey diagnoses and treatments to patients. Medical students frequently use sketches to help them think through clinical problems in individual and group problem solving. Applications ranging from automated patient records to medical education software could benefit greatly from the richer and more natural interfaces that would be enabled by the ability to understand sketches. In this paper we take the first steps toward developing a system that can understand anatomical sketches. Understanding an anatomical sketch requires the ability to recognize what anatomical structure has been sketched and from what view (e.g. parietal view of the brain), as well as to identify the anatomical parts and their locations in the sketch (e.g. parts of the brain), even if they have not been explicitly drawn. We present novel algorithms for sketch recognition and for part identification. We evaluate the accuracy of the recognition algorithm on sketches obtained from medical students. We evaluate the part identification algorithm by comparing its results to the judgment of an experienced physician.

We have developed a virtual reality (VR) and an augmented reality (AR) dental training simulator utilizing a haptic device. The simulators utilize volumetric force feedback computation and real time modification of the volumetric data. They include a virtual mirror to facilitate indirect vision during a simulated operation. The AR environment allows students to practice surgery in correct postures by combining the 3D tooth and tool models with the real-world view and displaying the result through a video see-through head-mounted display (HMD). Preliminary results from an initial evaluation show that the system is a promising tool to supplement dental training and that there are advantages of the AR over the VR approach.

 n this paper, we describe the operation of barter trade exchanges by identifying key techniques used by trade brokers to stimulate trade and satisfy member needs, and present algorithms to automate some of these techniques.  In particular, we develop algorithms that emulate the practice of trade brokers by  matching buyers and sellers  in such a way that trade volume is maximized while the balance of trade is maintained as much as possible.  We show that the buyer/seller matching and trade balance problems can be decoupled, permitting efficient solution as well as numerous options for matching strategies. 

We present a language for representing context­ sensitive temporal probabilistic knowledge. Con­ text constraints allow inference to be focused on only the relevant portions of the probabilistic knowledge. We provide a declarative semantics for our language and an implemented algorithm (BNG) that generates Bayesian networks to com­ pute the posterior probabilities of queries. We il­ lustrate the use of the BNG system by applying it to the problem of modeling the effects of medica­ tions and other interventions on the condition of a patient in cardiac arrest.

This paper describes COMET, a collaborative intelligent tutoring system for medical problem-based learning.  COMET uses Bayesian networks to model individual student knowledge and activity, as well as that of the group.  Generic domain-independent tutoring algorithms use the models to generate tutoring hints.  We present an overview of the system and then the results of two evaluation studies.  The validity of the modeling approach is evaluated in the areas of head injury, stroke and heart attack. Receiver operating characteristic (ROC) curve analysis indicates that, the models are accurate in predicting individual student actions. Comparison of learning outcomes shows that student clinical reasoning gains from our system are significantly higher than those obtained from human tutored sessions (Mann-Whitney, p = 0.011).

This paper describes COMET, a collaborative intelligent tutoring system for medical problem-based learning. The system uses Bayesian networks to model individual student knowledge and activity, as well as that of the group. It incorporates a multi-modal interface that integrates text and graphics so as to provide a rich communication channel between the students and the system, as well as among students in the group. Students can sketch directly on medical images, search for medical concepts, and sketch hypotheses on a shared workspace. The prototype system incorporates substantial domain knowledge in the area of head injury diagnosis. A major challenge in building COMET has been to develop algorithms for generating tutoring hints. Tutoring in PBL is particularly challenging since the tutor should provide as little guidance as possible while at the same time not allowing the students to get lost. From studies of PBL sessions at a local medical school, we have identified and implemented eight commonly used hinting strategies. We compared the tutoring hints generated by COMET with those of experienced human tutors. Our results show that COMET’s hints agree with the hints of the majority of the human tutors with a high degree of statistical agreement (McNemar test, p = 0.652, Kappa = 0.773).

This paper compares the accuracy of Decision Tree and Bayesian Network algorithms for predicting the academic performance of undergraduate and postgraduate students at two very different academic institutes: Can Tho University (CTU), a large national university in Viet Nam; and the Asian Institute of Technology (AIT), a small international postgraduate institute in Thailand that draws students from 86 different countries. Although the diversity of these two student populations is very different, the data-mining tools were able to achieve similar levels of accuracy for predicting student performance: 73/71% for {fail, fair, good, very good} and 94/93% for {fail, pass} at the CTU/AIT respectively. These predictions are most useful for identifying and assisting failing students at CTU (64% accurate), and for selecting Very Good students for scholarships at the AIT (82% accurate). In this analysis, the Decision Tree was consistently 3-12% more accurate than the Bayesian Network. The results of these case studies give insight into techniques for accurately predicting student performance, compare the accuracy of data mining algorithms, and demonstrate the maturity of open source tools.

 In this paper we present TRADES, a data-driven agent-based simulator for barter trade exchanges.  We provide an overview of the barter trade exchange industry, focusing on the operational aspects of trade exchanges and motivating the design of our simulator.  Our simulator is built by learning probabilistic models of company purchase behavior using transaction history data from an operating trade exchange.  We quantitatively evaluate the accuracy of our simulator by comparing simulated trade to the transaction data, showing a high degree of agreement between the two.    We also demonstrate use of the simulator to evaluate the effectiveness of a particular trade brokering strategy.

We describe ongoing work toward development of a decision­theoretic agent to help users choose videos based on their preferences. The DIVA (Decision Theoretic Interactive Video Advisor) system elicits user preferences using a case­based technique. Hard constraints are used to permit the user to communicate temporary deviations from his basic preferences. If the user is not happy with the system's recommendations, he can provide feedback, which is used to modify the represented preferences and generate a new set of recommendations. We describe the fundamental algorithms, the implementation, and some results from some initial experimentation.

Clinical decision analysis seeks to identify the optimal management strategy by modelling the un­ certainty and risks entailed in the diagnosis, natu­ ral history, and treatment of a particular problem or disorder. Decision trees are the most frequently used model in clinical decision analysis, but can be tedious to construct, cumbersome to use, and computationally prohibitive, especially with large, complex decision problems. We present a new method for clinical decision analysis that combines the techniques of decision theory and artificial in­ telligence. Our model uses a modular represen­ tation of knowledge that simplifies model building and enables more fully automated decision mak­ ing. Moreover, the model exploits problem struc­ tures to yield better computational efficiency. As an example we apply our techniques to the problem of management of acute deep venous thrombosis.

We present an educational tool for bringing the in­ formation contained in a Bayesian network to the end user in an easily intelligible form. The banter shell is designed to tutor users in evaluation of hypotheses and selection of optimal diagnostic procedures. ban­ ter can be used with any Bayesian network contain­ ing nodes that can be classified into hypotheses, obser­ vations, and diagnostic procedures. We present algo­ rithms for determining optimal diagnostic procedures and for explanation generation.

This paper discusses techniques for perform­ ing efficient decision­theoretic planning. We give an overview of the drips decision­ theoretic refinement planning system, which uses abstraction to efficiently identify opti­ mal plans. We present techniques for au­ tomatically generating search control infor­ mation, which can significantly improve the planner's performance. We evaluate the effi­ ciency of drips both with and without the search control rules on a complex medical planning problem and compare its perfor­ mance to that of a branch­and­bound deci­ sion tree algorithm.

Intelligent tutoring systems assist medical faculty in training and equipping students with the required clinical reasoning skills. Plausible student solutions to a given problem are rejected by tutoring systems as being incorrect, if they do not match a specific solution accepted by the tutoring system. This leads to brittleness in evaluating student solutions. In this paper we describe a combination of knowledge base expansion and exploitation of existing knowledge structure to enhance robustness in an intelligent tutoring system for medical problem-based learning using UMLS. We present a tutoring system that enriches the solution space by collating different plausible solutions and exploiting the knowledge structure in UMLS to offer students a broader scope of reasoning.