Determining risks: Insight to the Norwegian Cervical Cancer Data
In this talk, we will discuss a new collaborative research project between Lawrence Livermore National Laboratory (LLNL) and the Cancer Registry of Norway. Utilizing population-level cancer screening data from Norway, we are building predictive statistical models to assist researchers and doctors construct informative cancer screening schedules for individual patients. The current cancer screening dataset under consideration is composed of cervical cancer screening data for more than 1.7 million women over a twenty four year period. The dataset presents several difficulties, such as longitudinal observations with interval censoring, multiple diagnostic test types of varying accuracy, and low incidence rates of particular disease states. Our current methodology is to use a Mixture Hidden Markov Model with multiple observation types to capture population heterogeneity, disease progression/regression, diagnostic test type and misclassification. We will be presenting some initial results and discussing future directions for this research.
Thursday, September 14 at 2:00pm to 3:00pm
New Research Building Auditorium