RATIONALE AND OBJECTIVES: The authors developed and tested automated and semiautomated bowel-lumen tracking and colon-unraveling techniques for determining the central axis of the bowel. METHODS: A computer-simulated gastrointestinal tract phantom was used to test the accuracy of an automated algorithm for central axis determination and bowel unraveling. Variations in cross-sectional features between straight and unraveled formats were compared in a canine bowel segment in vitro and a human colon in vivo by using spiral computed tomography. Three readers each performed three semiautomated evaluations. RESULTS: Accuracy of the automated algorithm was confirmed by the high degree of correlation in the cross-sectional feature measurements (length error, < 1%). For the canine colon segment, accuracy of the semiautomated algorithm was confirmed by comparison with the automated tracing. For the human colon, readings were reproducible with 3.3% (+/- 1.9 standard deviation) mean variation in length. CONCLUSION: An automated algorithm for central axis deterioration and unraveling the colon has been validated in a gastrointestinal tract phantom. A semiautomated algorithm has been shown to be reproducible and time-efficient.
RATIONALE AND OBJECTIVES: The authors developed and tested automated and semiautomated bowel-lumen tracking and colon-unraveling techniques for determining the central axis of the bowel. METHODS: A computer-simulated gastrointestinal tract phantom was used to test the accuracy of an automated algorithm for central axis determination and bowel unraveling. Variations in cross-sectional features between straight and unraveled formats were compared in a caninebowel segment in vitro and a human colon in vivo by using spiral computed tomography. Three readers each performed three semiautomated evaluations. RESULTS: Accuracy of the automated algorithm was confirmed by the high degree of correlation in the cross-sectional feature measurements (length error, < 1%). For the canine colon segment, accuracy of the semiautomated algorithm was confirmed by comparison with the automated tracing. For the human colon, readings were reproducible with 3.3% (+/- 1.9 standard deviation) mean variation in length. CONCLUSION: An automated algorithm for central axis deterioration and unraveling the colon has been validated in a gastrointestinal tract phantom. A semiautomated algorithm has been shown to be reproducible and time-efficient.
Authors: Jeff Goldsmith; Brian Caffo; Ciprian Crainiceanu; Daniel Reich; Yong Du; Craig Hendrix Journal: Ann Appl Stat Date: 2011-01-01 Impact factor: 2.083