Researchers use 'crowdsourcing' to improve colon cancer screening methods, training

Episode # 89
Uploaded: March 7, 2012
Running Time: 03:48

CROWN: From the National Institutes of Health in Bethesda, Maryland, this is CLINICAL CENTER RADIO.

Some people call it distributed human intelligence. Others call it clickworking or crowdsourcing. The concept is a growing phenomenon. And the basic idea is to connect businesses with people on the internet, sometimes called knowledge workers, who are willing to complete tasks for negotiated rewards. In many cases, the tasks are simple and don't require much expertise. For example, a crowdsourcing gig might be organizing an index or checking links in a document. However, researchers at the NIH Clinical Center recently explored harnessing the power of crowdsourcing when they wanted to learn more about polyp detection. Dr. Ronald Summers, chief of the Imaging Biomarkers and Computer-Aided Diagnosis Laboratory in Radiology and Imaging Sciences at the Clinical Center, explains the experiment:

SUMMERS: We showed them pictures of the inside of a colon and asked them to check a box of whether they thought the picture showed a polyp or not. Now, we had to give them some training because the average person wouldn't know what to do. So we gave them (for reference) five polyp pictures and five not polyp pictures. Then we sent them to work and we got something like 10,000 readings from these people in three days.

CROWN: To assess reproducibility, they did the entire experiment one month later. Their results were surprising, says Dr. Summers:

SUMMERS: On the whole, the people who participated in the experiment did quite well.

CROWN: But the idea of this experiment wasn't to see if the average person could be trained to detect polyps. Dr. Summers and his team were actually trying to better understand how people perceive images or scans of the colon. They hope to apply this information later on to improve training for doctors.

SUMMERS: These are the kinds of things that can lead to better performance. We believe that the same factors will lead to improved performance by the radiologists. So the idea is that you would test an idea using crowdsourcing. And once you got to a level of performance with the non-experts, you would then design an experiment to test the same scenario with the experts.

CROWN: Colorectal cancer is the second leading cause of cancer death in Americans. And colorectal cancer is a largely preventable disease if colorectal polyps are properly screened and removed. Current screening tests for polyps include colonoscopy and computed tomographic colonography, both of which are performed and interpreted by doctors. And, in the past few years, computer-aided polyp detection software has become available. But not all physicians use the software to the best advantage, and researchers are trying to understand why so they can improve colon cancer screenings.

SUMMERS: The answer is not fully understood. And, in part, that is the reason why I did this experiment, to try to understand it. This sort of problem crosses a lot of interdisciplinary boundaries. It's not just a question of radiology or medical knowledge, but there are also questions about psychology and perception.

CROWN: From America's Clinical Research Hospital, this has been CLINICAL CENTER RADIO. In Bethesda, Maryland, I'm Ellen Crown, at the National Institutes of Health, an agency of the United States Department of Health and Human Services.

Back to Clinical Center Radio