Integrating digital photos within PACS may cut ID errors

It's an old adage that a picture is worth a thousand words. This seems to hold true for radiology, where patient photos can help decrease patient misidentification errors, researchers reported at the Society for Imaging Informatics in Medicine (SIIM) meeting.

Researchers from Emory University have created a prototype of an Android-based camera device that can take digital photos of patients at the time they're being imaged and then integrate them within the PACS environment. A clinical observer study with the prototype showed a significant increase in radiologist detection of misidentification errors.

"Integrating photos with medical imaging studies is technically feasible and inexpensive," said Dr. Srini Tridandapani, PhD. "We believe this can lead to reduction of patient misidentification and lead to increased throughput and improve diagnostic capabilities."

Tridandapani shared the results during a scientific session at the Orlando, FL, meeting.

Misidentification of patients is a major problem in medicine, and radiology is no exception, Tridandapani said. Published research has reported detection rates for errors (misidentification or mislabeling) in imaging exams ranging from 0.26% to 0.73%. The actual error rate is likely even higher, he said.

However, current regulatory or voluntary solutions to avoid mislabeling of imaging studies are insufficient, as they are based on extrinsic identifiers and fraught with errors, he said.

The Joint Commission currently mandates a dual-identifier requirement for patients from possible identifiers such as patient name, date of birth, social security number, hospital-assigned number, and telephone number. It can be very difficult, though, to get this sort of information from trauma patients or those who have been sedated, for example, Tridandapani said.

To address this problem, the Emory group sought to develop a way to obtain facial photos with all medical imaging studies at the time they are obtained. These photos would be integrated with the acquisition, storage, and display of medical images, he said. In the future, photographs of other body regions could also be added.

Thanks to continued advances in CCD/CMOS camera technologies, cheaper cameras are becoming available. Memory is also becoming continuously cheaper, and the DICOM Standard already allows the ability to store and display visible-light images, Tridandapani said.

The Emory team developed a system that features an Android-based camera device and an integration server that communicates wirelessly with the institution's PACS.

The camera device was custom-built using off-the-self components, built around a BeagleBoard-xM platform with an ARM microprocessor. It includes a three-megapixel camera, a Bluetooth-connected radiofrequency identifier (RFID) reader, and an 802.11 b/g wireless module for connection to the integration server.

Mounted on a portable x-ray imaging system, the camera device operates in a transparent fashion to the technologist. When the x-ray system trigger is pressed by the technologist, it also automatically activates the camera device. The photo is taken and the RFID reader scans the cassette ID.

The integration server retrieves current studies from the PACS, reads the cassette ID from the DICOM header, and matches the patient photo based on the cassette ID and time stamp. A new series is created with the matched, DICOM-converted photo and sent to the PACS, according to Tridandapani.

The researchers also devised a hanging protocol to display the patient photo along with the images. To test the system, they performed the clinical observer study.

In the first phase, 10 recently board-certified radiologists were each provided with a worklist for 20 pairs of radiographs without patient photographs. Unbeknownst to the readers, the pairs included two to four misidentification errors. The same 10 readers were asked to read 20 other pairs of radiographs, but this time with patient photos included. No additional instructions were provided.

Of the 24 total errors in the first phase of the study, only three errors (12.5%) were detected. When asked to read the pairs that included patient photographs, the readers detected 16 of 25 errors (64%).

"We found that seven of the readers missed even with the photographs, because they were actively avoiding looking at the photographs because they thought we were trying to distract them from the x-rays," he said. "But that was because we didn't give them any instructions."

The readers spent about 35 minutes reading the studies without the photographs, and only 26 minutes with the photographs.

"I think this might be related to the fact that if you're looking for lines and tubes, this may help you speed up the process," he said. "But that's something that needs to be tested."

Engineering challenges remain for integrating cameras with all imaging modalities, he said. A clinical trial should also be performed to evaluate the technique's effectiveness and to explore workflow issues and impact on radiologist efficiency, he said.

Page 1 of 775
Next Page