We are currently focusing on a copyrighted method of processing image data for integration into a mobile app. This project is supported by the UT Health Catalyst.
Radiologic imaging is a key variable in many clinical trials, but collecting these data can be cumbersome, time consuming and expensive. Clinical trial sites bill sponsors thousands of hours to copy, de-identify and either write CDs to mail or electronically transfer image files. Many are improperly de-identified, corrupted or lost leading to more staff hours and higher costs. These large files also contain metadata that is potentially identifying and difficult to remove. We have developed a method for medical image capture and transfer that is 10 times faster, cheaper and better than traditional methods. Our preliminary data show this method does not sacrifice image quality, that is, the scans are readable, scalable and high quality. The project will involve work on the image processing softare in Python.
We will create an imaging repository for clinical trials with a special application that streams the images into a HIPAA-compliant cloud where we can download them for processing. We can create repositories for a tenth the cost sponsors are currently paying, creating value and improving data quality and thus patient care. But greater value will be achieved when we have provided the service to 100 clinical trials, combined their imaging data and built the largest imaging repository in existence. These data may be robust enough to train artificial intelligence programs to read the scans with skill approaching that of a human radiologist, and they will likely have many other innovative applications that add value to and improve patient care.
Python experience
This project can begin immediately and can extend to meet needs of student.
Image data organization and QA
Develop image processing algorithms
