Feasibility study. A phantom was designed to test the accuracy of three-dimensional reconstruction. Thereafter, 192 small-bowel capsule endoscopy images (of vascular: 50; inflammatory: 73; protruding structures: 69) were reviewed with the aid of a purpose-built three-dimensional reconstruction software. Seven endoscopists rated visualisation improved or non-improved. Subgroup analyses performed for diagnostic category, diagnosis, image surface morphology and colour and SBCE equipment used (PillCam? vs. MiroCam?).
Overall, phantom experiments showed that the three-dimensional reconstruction software was accurate at 90 % of red, 70 % of yellow and 45 % of white phantom models. Enhanced visualisation for 56 % of vascular, 23 % of inflammatory and <10 % of protruding structures was noted (P = 0.007, 0.172 and 0.008, respectively). Furthermore, three-dimensional software application enhanced 53.7 % of red, 21.8 % of white, 17.3 % of red and white, and 9.2 % of images of lesions with colour similar to that of the surrounding mucosa, P < 0.0001.
Application of a three-dimensional reconstruction software in capsule endoscopy leads to image enhancement for a significant proportion of vascular, but less so for inflammatory and protruding lesions. Until optics technology allows hardware-enabled three-dimensional reconstruction, it seems a plausible alternative.