Dual-energy iterative methods for CT reconstruction

Working with the ALERT COE, we are extending recently-developed shape based methods and applying them to scanner data. We recently demonstrated noticeable gains by incorporating patch-based regularization terms in the reconstruction method.

A paper on our work is found on arXiv by clicking here, and a recent conference paper is here:  imageFormationConf2014_TraceyMiller

Patch-based methods for image and time-series denoising

Several of our projects relate to the improvements that can be achieved using patch-based methods such as non-local means. In collaboration with BBN and other partners, we are extending these techniques to reduce parameter sensitivity and to tailor the methods to better preserve edge information. We have also seen promising results in applying these techniques to time series data. such as ECG.

You can find code for patch-based ECG denoising here.

A recent paper related to this work is here.

Cough sounds analysis for tuberculosis patient monitoring

With collaborators at Johns Hopkins and in Lima, Peru, I am helping to develop software for counting cough rates in TB patients. This work may eventually allow clinicians without access to laboratory facilities to identify patients who are not responding to treatment, whether due to drug-resistance or other problems. We have seen promising results, and are moving on to second-generation systems.

Here’s a preprint on this work: coughAlgoEmbc2011_revApril10

Image processing for airport security screening

We are developing image processing algorithms and false alarm reduction methods for airport screening, allowing more rapid and secure airline travel. This work is funded through the  ALERT center as well as by industrial collaborators.