PTU-227 Improving efficiency in capsule endoscopy: can reading times be reduced without sacrificing diagnostic accuracy? A self-assessment. (28th May 2012)

• Record Type: Journal Article
• Authors: Nakamura, M; Murino, A; Fitzpatrick, A; Fraser, C
• Is Part Of: Gut. Volume 61(2012)Supplement 2
• Page Start: A278
• Page End: A278
• Publication Date: 2012-05-28
• Subjects: Gastroenterology -- Periodicals; 616.33
• DOI: 10.1136/gutjnl-2012-302514c.227
• Languages: English
• ISSNs: 0017-5749

Abstract

Abstract : Introduction: Capsule endoscopy (CE) is a time consuming procedure. The RAPID 7 Access reading software (Given Imaging Ltd) has three patterns of view modes (VM) (one view, VM1; double views, VM2; quadruple views, VM4) and an adjustable frame rate (AFR) from 5 to 40 fps. The appropriate settings for VM and AFR depend on capsule endoscopist's experience, and a consensus has not been achieved yet. The aim of this study was to investigate how different VM's and AFR's could influence diagnostic accuracy. Methods: An entire capsule endoscopy procedure consisting of 27 small bowel angioectasias was selected from our database. This was read by a single expert capsule endoscopist repeatedly using 11 different randomised combinations of VM and AFR (1, 2 and 4 VM × 10, 15, 25 and 40 fps). Reading times and number of angioectasias detected for each combination were recorded and then compared. Results: The small bowel transit time was 321 min. Mean reading times (all VM's) at 10, 15, 25 and 40 fps respectively were 34, 22, 14 and 10 min. Considering 10 fps as the gold standard for reading, the reduction in reading time at 15, 25 and 40 fps was 33%, 60% and 70% respectively. No significant differences were noticed in reading times between VM's at the same AFR. A mean of 23, 16, 7 and 6 angioectasias were detected at 10, 15, 25 and 40 fps respectively (all VM's combined). Diagnostic accuracy at 25 and 40 fps was significantly lower than 10 fps (p=0.04, 0.01). The mean numbers of detected angioectasias according to VM were 14, 17 and 16 for VM1, VM2 and VM4 respectively. The lowest number of angioectasias (5) was detected using VM2 × 40 fps. The highest number of angioectasias (25) was detected using VM2 × 10 fps and VM4 × 10 fps. Using VM2 × 15 fps, 18 angioectasias were detected, meaning that diagnostic accuracy was reduced to 72% (compared with VM2 × 10 fps), although the reading time decreased by 33%. Conclusion: Our findings suggest that the highest diagnostic accuracy was achieved with VM2 × 10 fps or VM4 × 10 fps. The AFR influences both diagnostic accuracy and reading time. As the AFR increases, reading times are reduced but this is associated with a reduction in diagnostic accuracy and a concomitant increase in miss rates. Capsule endoscopists need to be aware of this phenomenon. Competing interests: None declared.