Virtual colonoscopy (VC) allows a physician to virtually navigate within areconstructed 3D colon model searching for colorectal polyps. Though VC iswidely recognized as a highly sensitive and specific test for identifyingpolyps, one limitation is the reading time, which can take over 30 minutes perpatient. Large amounts of the colon are often devoid of polyps, and a way ofidentifying these polyp-free segments could be of valuable use in reducing therequired reading time for the interrogating radiologist. To this end, we havetested the ability of the collective crowd intelligence of non-expert workersto identify polyp candidates and polyp-free regions. We presented twenty shortvideos flying through a segment of a virtual colon to each worker, and thecrowd was asked to determine whether or not a possible polyp was observedwithin that video segment. We evaluated our framework on Amazon Mechanical Turkand found that the crowd was able to achieve a sensitivity of 80.0% andspecificity of 86.5% in identifying video segments which contained a clinicallyproven polyp. Since each polyp appeared in multiple consecutive segments, allpolyps were in fact identified. Using the crowd results as a first pass, 80% ofthe video segments could in theory be skipped by the radiologist, equating to asignificant time savings and enabling more VC examinations to be performed.
展开▼
