CT/CBCT Alignment and Long-Term Engineering Support

Mobius Medical Systems (recently acquired by Varian Medical Systems) produces radiation oncology treatment planning quality assurance (QA) software. The software is capable of double checking the planned dose calculated by a treatment planning system, ensuring the planned dose is within safe limits for critical radiosensitive structures (e.g. spinal cord, optic nerve, etc), and ensuring the dose actually delivered during treatment matches what was intended.

We were brought on by Mobius to help implement an automatic, quantitative comparison between planning CT images and pre-treatment CBCT images to detect patient positioning errors and anatomical changes. Patients usually have planning CT images taken days or even weeks before starting radiation therapy. The treatment plan is computed and optimized based on the planning CT images. Fast CBCT images are taken minutes before treatment to ensure the alignment of the patient relative to the planning CT images is acceptable. We helped turn the once manual and qualitative process of alignment assessment into an automatic and quantitative one.

After the CT/CBCT alignment project, our role grew to encompass the following:

• Worked as an extension to the in-house engineering on an FDA-regulated product following all existing regulatory procedures.
• Furnished data and results for the FDA and marketing materials.
• Wrote high quality design documents suitable to include in regulatory documentation.
• Interfaced with hospital PACS to automatically retrieve relevant DICOM files for further processing.
• Automated parts of the quality management system to reduce labor intensive and error prone manual preparation.
• Wrote code that was flexible enough to respond to fast changing market pressures and customer demands.
• Discovered and worked around vendor-specific quirks in input data.
• Helped implement and test a 3D/3D image registration algorithm for CBCT/CT images to suggest a table movement to bring the patient into alignment.
• Supported features we wrote; this included writing defect reports, attending support calls with users, gathering additional data from complaints, and pushing updates on live production servers.
• Read and wrote DICOM files; generated radiotherapy DICOM objects from scratch.
• Helped with software maintenance and customer support duties, even for features not originally written by us.
• Came up with improvements to testing infrastructure and developer tools that have increased developer productivity and saved countless developer-hours.
• Created automation tests and harnesses for complex environments such as simulated PACS and network shares.
• Wrote performant log anonymization scripts to ensure HIPPA compliance.
• Filed issue reports and tracked down solutions to bugs outside of the codebase in third party libraries such as CouchDB and PyDICOM.
• Helped create a web frontend to an existing product written in MATLAB.