QA program cuts errors in 3D image processing lab

SAN FRANCISCO - Implementing a quality assurance program in a centralized 3D image processing lab can reduce the number of errors that occur when technologists reconstruct 3D images, according to a presentation on Wednesday at the International Society for Computed Tomography (ISCT) meeting.

The growing popularity of 3D reconstruction means that clinicians are becoming increasingly reliant on the studies, according to Laura Pierce, 3D lab manager at Stanford University in Stanford, CA. In fact, some surgeons won't take their patients to surgery until they've reviewed 3D images, she said.

At the same time, 3D image processing suffers from a number of challenges, including the fact that it can be highly operator dependent due to variations in operator experience and training. This can make it relatively easy to misrepresent anatomy.

Due to the volume of 3D imaging performing at Stanford, the radiology department operates a centralized lab with seven full-time employees. At the 2007 MDCT meeting, Pierce discussed how a QA program could be implemented to address quality assurance issues. At this year's show, she discussed how Stanford implemented the program, how it affected the lab's error rate, and whether it impacted workflow.

At the outset, Stanford wanted to make sure that the QA program didn't affect the workflow at the lab. Researchers also wanted to ensure that the program was something that the facility could continue to follow perpetually, as opposed to a one-time initiative.

The researchers first observed 3D lab staff over a three-month benchmark period without the staff's knowledge to determine the lab's error rate prior to retraining. They then classified the errors and brainstormed to find ways to reduce them. Many of the most common errors involved 3D technologists rushing through studies, being interrupted by colleagues, skipping image review, and failing to confirm protocols.

Techniques for reducing errors included advising staff to seek advice when processing studies, improving technologist training, more thorough exam review, and linking error rates to competency ratings. Lab managers even had staff listen to music on iPods to reduce external distractions and help technologists focus on their work.

They applied the techniques during a three-month retraining period, with lab managers improving their mentoring of staff through peer-review sessions, literature review, and radiologist case reviews. Mandatory rest breaks were enforced to prevent fatigue, and staff members were discouraged from interrupting each other -- they even turned off their cell phones. Reconstruction protocols were revised and made available electronically. Error rates were also tracked during this period.

Finally, Pierce and colleagues tracked error rates during a nine-month postintervention period to see if the training techniques stuck. The entire study took place between September 2008 and November 2009.

To classify errors, the research team developed a list of 13 error tags representing the most common mistakes in 3D image processing, ranging from postprocessing errors such as incorrect annotation to PACS workflow errors such as 3D images not found on the institution's PACS network. Errors were also classified as major, minor, or trivial.

During the study period, the seven full-time 3D technologists processed a total of 13,438 CT and MR images. During the benchmark period, the lab staff produced 1,236 errors, for an aggregate error rate of 16.1%. Error rates during follow-up periods were as follows:

3D processing error rate by survey period
Survey period Period duration Error rate
Observation 3 months 16.1%
Retraining 3 months 8.2%
Follow-up 9 months 7.4%

The QA program reduced error rates nearly by half, and the lab staff appeared to retain the knowledge by driving the error rate even lower during the follow-up period. All this occurred during a time in which 3D lab volume grew 10.2% in the post-training period, and the lab's turnaround time for stat studies fell 21% (p < 0.01).

"Even though we were processing more studies, our error rate dropped, so we were very encouraged," Pierce said.

The number of major errors at the lab fell from 0.6% to 0.1%, the number of minor errors dropped from 5.2% to 1.6%, and the number of trivial errors declined from 10.3% to 5.8% (p < 0.01 for all categories).

The staff also tracked the types of errors and the level of error rate reduction by error type. Their findings were as follows:

Most common 3D processing errors
Error type Error rate reduction after retraining
Incorrect measurement 73%
Annotation 72%
Missing images 52%
Positioning/technical 51%
Database entry 11%

Pierce said that the researchers were a bit disappointed that the rate of database entry errors did not fall more significantly following the training period. Pierce believes that the phenomenon occurred because data-entry errors typically do not reflect technical ability.

The QA program was also able to bring the error rates for new technologists down to the level seen with more experienced staff. At the beginning of the study, 3D technologists with less than four years of experience had error rates from 20% to 25%, while the lab's most experienced technologists (with 10 to 12 years of experience) had error rates of less than 10%. By the end of the program, all but one of the technologists had error rates of 10% or lower.

Pierce concluded by saying that even though the current study applied specifically to 3D imaging, its lessons could be applied to other areas of the healthcare enterprise.

"We feel that measuring quality with these metrics could be transferred to any type of clinical unit," she said.

By Brian Casey
AuntMinnie.com staff writer
May 21, 2010

Related Reading

Electronic technologist peer review program enhances QC, October 16, 2009

Quality control in the digital environment: Not a luxury, a necessity, September 7, 2009

Turning your 3D lab into a profit center, May 21, 2009

3D technologists: You've trained them -- now retain them, August 15, 2007

Copyright © 2010 AuntMinnie.com

Page 1 of 653
Next Page