Walter Sun

Segmenting and tracking the left ventricle by learning the dynamics in cardiac images

Having accurate left ventricle (LV) segmentations across a cardiac cycle provides useful quantitative (e.g. ejection fraction) and qualitative information for diagnosis of certain heart conditions. Existing LV segmentation techniques are founded mostly upon algorithms for segmenting static images. In order to exploit the dynamic structure of the heart in a principled manner, we approach the problem of LV segmentation as a recursive estimation problem. In our framework, LV boundaries constitute the dynamic system state to be estimated, and a sequence of observed cardiac images constitute the data. By formulating the problem as one of state estimation, the segmentation at each particular time is based not only on the data observed at that instant, but also on predictions based on past segmentations. This requires a dynamical system model of the LV, which we propose to learn from training data through an information-theoretic approach. To incorporate the learned dynamic model into our segmentation framework and obtain predictions, we use ideas from particle filtering. Our framework uses a curve evolution method to combine such predictions with the observed images to estimate the LV boundaries at each time. We demonstrate the effectiveness of the proposed approach on a large set of cardiac images. We observe that our approach provides more accurate segmentations than those from static image segmentation techniques, especially when the observed data are of limited quality.

Early identification of adverse drug reactions from search log data

The timely and accurate identification of adverse drug reactions (ADRs) following drug approval is a persistent and serious public health challenge. Aggregated data drawn from anonymized logs of Web searchers has been shown to be a useful source of evidence for detecting ADRs. However, prior studies have been based on the analysis of established ADRs, the existence of which may already be known publically. Awareness of these ADRs can inject existing knowledge about the known ADRs into online content and online behavior, and thus raise questions about the ability of the behavioral log-based methods to detect new ADRs. In contrast to previous studies, we investigate the use of search logs for the early detection of known ADRs. We use a large set of recently labeled ADRs and negative controls to evaluate the ability of search logs to accurately detect ADRs in advance of their publication. We leverage the Internet Archive to estimate when evidence of an ADR first appeared in the public domain and adjust the index date in a backdated analysis. Our results demonstrate how search logs can be used to detect new ADRs, the central challenge in pharmacovigilance.

Optimal Rebalancing for Institutional Portfolios

Institutional fund managers generally rebalance using ad hoc methods such as calendar periods or tolerance band triggers. Another approach is to quantify the cost of a rebalancing strategy in terms of risk-adjusted returns net of transaction costs. An optimal rebalancing strategy that actively seeks to minimize that cost uses certainty-equivalents and the transaction costs associated with a policy to define a cost-to-go function. Stochastic programming is then used to minimize expected cost-to-go. Monte Carlo simulations demonstrate that the method outperforms traditional rebalancing strategies such as periodic and 5% tolerance rebalancing.

Learning the Dynamics and Time-Recursive Boundary Detection of Deformable Objects

We propose a principled framework for recursively segmenting deformable objects across a sequence of frames. We demonstrate the usefulness of this method on left ventricular segmentation across a cardiac cycle. The approach involves a technique for learning the system dynamics together with methods of particle-based smoothing as well as nonparametric belief propagation on a loopy graphical model capturing the temporal periodicity of the heart. The dynamic system state is a low-dimensional representation of the boundary, and the boundary estimation involves incorporating curve evolution into recursive state estimation. By formulating the problem as one of state estimation, the segmentation at each particular time is based not only on the data observed at that instant, but also on predictions based on past and future boundary estimates. Although this paper focuses on left ventricle segmentation, the method generalizes to temporally segmenting any deformable object.

Using Dynamic Programming to Optimally Rebalance Portfolios

The authors propose a different framework that quantifies the cost of a rebalancing strategy in terms of risk-adjusted returns net of transaction costs. They then derive an optimal rebalancing strategy that actively seeks to minimize that cost. Certainty equivalents and the transaction costs associated with a policy to define a cost-to-go function are used, with the expected cost-to-go minimized using dynamic programming. They apply Monte Carlo simulations to demonstrate that their method outperforms traditional rebalancing strategies like monthly, quarterly, annual, and 5% tolerance rebalancing. They also show the robustness of our method to model error by performing sensitivity analyses.

Variational approaches on discontinuity localization and field estimation in sea surface temperature and soil moisture

Some applications in remote sensing require estimating a field containing a discontinuity whose exact location is a priori unknown. Such fields of interest include sea surface temperature in oceanography and soil moisture in hydrology. For the former, oceanic fronts form a temperature discontinuity, while in the latter sharp changes exist across the interface between soil types. To complicate the estimation process, remotely sensed measurements often exhibit regions of missing observations due to occlusions such as cloud cover. Similarly, water surface and ground-based sensors usually provide only an incomplete set of measurements. Traditional methods of interpolation and smoothing for estimating the fields from such potentially sparse measurements often blur across the discontinuities in the field.

Segmentation of the evolving left ventricle by learning the dynamics

We propose a method for recursive segmentation of the left ventricle (LV) across a temporal sequence of magnetic resonance (MR) images. The approach involves a technique for learning the LV boundary dynamics together with a particle-based inference algorithm on a loopy graphical model capturing the temporal periodicity of the heart. The dynamic system state is a low-dimensional representation of the boundary, and boundary estimation involves incorporating curve evolution into state estimation. By formulating the problem as one of state estimation, the segmentation at each particular time is based not only on the data observed at that instant, but also on predictions based on past and future boundary estimates. We assess and demonstrate the effectiveness of the proposed framework on a large data set of breath-hold cardiac MR image sequences.

A Statistical Analysis to Improve Basketball Strategy

Decision-making and competitive strategies in major American sports have often relied on heuristics and conventional wisdom. Recently, statistics and quantitative methods have been used in the sport of baseball to help teams improve their chances of winning. However, the use of mathematical methods to improve strategy has not been as prevalent in sports such as basketball. In this paper, we discuss a strategy based on statistical information that improves a teams probability of winning. In particular, we claim that the expected benefit of the strategy proposed could have affected the outcome of 17.5% of the 2004 NCAA tournament games.