Image-guidance Software Development

To augment the proton hardware at Mayo Clinic Rochester with the capability of fluoroscopy-based motion verification and monitoring, I developed computer vision application (Code name: Wukong), based on high-speed acquisition and real-time processing of flat-panel detector (FPD) data streams. Built as a fully DICOM-aware utility, this .NET/C# Windows desktop application is integrated into treatment planning workflow with Varian Eclipse Scripting API. It also provides an optional ability to store live-captured fluoroscopic frames on a HIPAA-compliant scalable storage cloud.

FPD

The development necessarily requires projective geometry calibration (see image above) of the flat-panel imagers. While the room configuration is engineered to a given specification, the actual geometry may contain small errors. This project delivers an iterative algorithmic calibration routine to correct such errors, i.e. translation, rotation and non-affine distortions. The main motivation behind this routine is to enable high-fidelity perspective overlays of RT structure sets in our in-house clinical software. It may also be used as an independent QA of vendor’s own calibrations.

Publications and talks which resulted from this project are listed below.

Avatar
Zheng Zhang, Ph.D.
Clinical Assistant Professor