|
Abstract:
|
Atomic force microscopy provides a primary means of determining the mechanical properties of soft biomaterials. A key technical problem is to determine precisely the moment when the probe comes in to contact with the sample---a problem of detecting a change between two parametric regression models. By employing a hierarchical Bayesian approach, we formulate this changepoint detection problem in a way that is amenable to posterior sampling and Rao-Blackwellization, and subsequently demonstrate improvements relative to existing techniques for controlled experiments that we designed and performed. Our approach accounts for experimentally observed changepoint smoothness constraints and measurement error variability and allows for direct inference of material stiffness parameters. We show data analysis results obtained for silicone synthetic breast material, neurons, and red blood cells.
|
