Tufts ML Alumni
Current location: Medford, MA
Associated Publications: [+]
Authors: Yijun Zhao, Tanuja Chitnis, Brian C. Healy, Jennifer G. Dy, Carla E. Brodley
IEEE International Conference on Data Mining (ICDM)
Abstract: Predicting disease course is critical in chronic progressive diseases such as multiple sclerosis (MS) for determining treatment. Forming an accurate predictive model based on clinical data is particularly challenging when data is gathered from multiple clinics/physicians as the labels vary with physicians’ subjective judgment about clinical tests and further we have no a priori knowledge of the various types of physician subjectivity. At the same time, we often have some (limited) domain knowledge on how to group patients into disease progression subgroups. In this paper, we first present our rationale for choosing a Dirichlet mixture of Gaussian processes (DPMGP) model to address the subjectivity in our data. We then introduce a new approach to incorporating domain knowledge into the non-parametric mixture model. We demonstrate the efficacy of our model by applying it to two medical datasets to predict disease progression in MS patients and disability levels in early Parkinson’s patients.
Authors: Yijun Zhao, Carla E. Brodley, Tanuja Chitnis, Brian C. Healy
Proceedings of the 2014 SIAM International Conference on Data Mining
Authors: Bilal Ahmed, Thomas Thesen, Karen Blackmon, Yijun Zhao, Orrin Devinsky, Ruben Kuzniecky, Carla E. Brodley
Proceedings of the 31st International Conference on Machine Learning (ICML-14)
Abstract: We cast the problem of detecting and isolating regions of abnormal cortical tissue in the MRIs of epilepsy patients in an image segmentation framework. Employing a multiscale approach we divide the surface images into segments of different sizes and then classify each segment as being an outlier, by comparing it to the same region across controls. The final classification is obtained by fusing the outlier probabilities obtained at multiple scales using a tree-structured hierarchical conditional random field (HCRF). The proposed method correctly detects abnormal regions in 90\% of patients whose abnormality was detected via routine visual inspection of their clinical MRI. More importantly, it detects abnormalities in 80\% of patients whose abnormality escaped visual inspection by expert radiologists.
Current Research Topics:
Description: Our work is done in the context of expressive Bayesian probabilistic models (a.k.a graphical models), developing inference algorithms for them, developing a learning theory that explains why these algorithms work and applying them in interesting applications. Our theoretical results provide distribution-free guarantees on the risk of approximate Bayesian inference algorithms. Recent models include constrained clustering, multi-task learning, sparse Gaussian processes, mixture of expert models for label discretization, matrix facorization and topic models. Recent applications include land-cover clustering and classification, analysis of time series from Astronomy, and predicting contamination level in environmental engineering.
This work is partly supported by NSF grants IIS-1714440 and IIS-0803409
Past Research Topics: [+]
Description: We are working with researchers from Harvard Medical School to predict outcomes for multiple sclerosis patients. A focus of the research is how best to interact with physicians to use both human expertise and machine learning methods.