wang_george.jpgPicture archiving and communication systems (PACS) have been hailed as the state of the art in storage, retrieval and access to digital medical images and are a critical part of the effort to convert the nation to electronic medical records. Because they archive the images from CT and PET scans, X-rays, MRI exams, etc. on computers and allow practitioners at multiple sites to access them remotely, PACS are a boon to physicians and are being phased in at medical centers and practices throughout the country. But like any technology, they are subject to human and computer error. Just as we sometimes accidentally save a file to the wrong folder on our computers, hospitals are finding that the PACS images sometimes go missing, with potentially catastrophic results.

To address problems with misfiled and misplaced images, Olive View–UCLA Medical Center has turned to a team from CSUN’s Department of Computer Science for help in designing a digital watermarking system that will tag the images so that “orphaned” images can be identified and filed where they belong. Over a year ago, graduate students Christopher Gutierrez and Gautam Kakani began working with George Wang, associate professor of computer science, and Ramesh Verma, M.D., the medical center’s nuclear medicine director, to develop a system that will “stamp” each image with a unique digital imprint containing key information about the patient.

“A watermark fluctuates values slightly so they are invisible to the human eye, but a computer program can be designed to extract and reconstruct hidden information,” explains Gutierrez. “It’s currently being used in the film industry. If a company wants to distribute a few movies before their release to the public, it will add a unique, hidden watermark—like a visual serial number or barcode that is invisible to a person viewing the film.”

CT_brain_MR8075_spectra0002.jpgCT_brain_MR8075_spectra0003.jpgCT_abdomen_MR8074_spectra0003.jpg
Olive View–UCLA staff are not sure why images sometimes become orphaned, but a network that becomes saturated with traffic may cause transmission to get bogged down, prompting a technician to abort the process and try again and possibly causing the information to be only partially recorded in the database. Sometimes the problem seems to take place over a period of time in storage, possibly due to noise spikes in the database or power problems during transmission. If the images are not reunited with the patient records, the missing information can, in a worst case, lead to misdiagnosis. Fixing such errors, which happen a few times a month, currently requires a radiologist to identify the error and then inform the PACS administrator, who will sometimes need hours to locate the point of error and even more time to fix it.

Kakani, who is working on the project for his master’s thesis, is trying to find a reversible algorithm that will insert the watermark but remove it once the image is retrieved so the image is not degraded by the watermark. Ideally, the project will flag the error when it occurs and immediately notify the PACS administrator, who will be responsible for fixing it. “We want to show that this can be used not only from a security point of view but also from an error management point of view,” he says.