PROVIDENCE, RI - Creating detailed pictures of the face from 3D surface-reconstructed images could be construed as a patient privacy issue under the provisions of the Health Insurance Portability and Accountability Act (HIPAA). But the ability to identify patients from their 3D facial images is still relatively low, according to researchers from the Imaging Informatics Lab at the VA Maryland Health Care System in Baltimore.
"Preliminary data suggests that 3D surface-reconstructed images are poor substitutes to digital photographs," said Dr. Joseph Chen. He presented the results Friday at the 2007 annual meeting of the Society for Imaging Informatics in Medicine (SIIM).
To determine whether these images could be considered as comparable images and need to be protected under HIPAA provisions, the researchers performed a prospective study to evaluate whether 3D reconstructed images of a patient's facial features could be readily associated with that specific patient in a manner analogous to a photograph.
Fifty-two patients participated in the study. After informed consents were obtained, 10 patients received digital photos using a Panasonic Lumix ZDMC-FX3K digital camera, as well as 3D facial images created using AquariusNet (TeraRecon, San Mateo, CA). The remaining patients received only digital photos of the face to serve as controls.
Using a Web-based program, 35 image reviewers then examined the 3D images and sought to match the 3D image with a digital photo of a face, or to determine if no match was among the randomized photos.
The reviewers produced an overall accuracy in matching digital photos and 3D facial images of 56.4%. No statistical significance was seen by gender (p = 0.31), age (p = 0.60), or ethnicity (p = 0.30).
This was a surprisingly poor result, as no restrictions were placed on the reviewers, Chen said.
"Everybody can spend as much time as they want on the computer screen. They could compare and look," he said. "And we didn't isolate particular sex, age, or ethnicity. If you group people together, that would make it harder for people to recognize and differentiate between another."
The lack of accuracy is good news currently, but the situation might change in the future, Chen said.
Technology advances, such as 256-slice CT, will lead to increased resolution and could potentially obtain superb images, he said. Facial features such as hair or scars could be visualized.
To ensure privacy and security, solutions such as a MRI defacer (utilized by the Morphometry Biomedical Informatics Research Network [BIRN]) might need to be employed. Some type of "spray painting" technology could also be an option, Chen said.
By Erik L. Ridley
AuntMinnie.com staff writer
June 8, 2007
Related Reading
Image postprocessing boosts performance of multidetector CT coronary angiography, May 31, 2007
Axial-loaded imaging of lumbar spine heightens diagnostic accuracy, May 29, 2007
PACS results in more incidental cervical spine MR findings, November 29, 2006
Copyright © 2007 AuntMinnie.com
