Introduction With the advent of capsule endoscopy, automated detection of Barrett's columnar lined oesophagus (BE) becomes achievable. This might allow screening to be undertaken in a non-specialist setting. We have shown that elastic scattering spectroscopy, which could be incorporated into a capsule, accurately differentiates squamous from columnar mucosa but cannot reliably differentiate gastric from oesophageal columnar mucosa. We hypothesise that the shape of BE distribution may hold a key to this. The aim of this study, therefore, was to determine whether there are distinctive macroscopic shapes of Barrett's columnar lined oesophagus which could be used for automated detection.
Methods Sixty-five patients (53 men, median age 69, range 45–83) with dysplastic BE, who were being assessed for endoscopic therapy between 2006 and 2009, had visual mapping of each quadrant of BE every 1 cm at endoscopy. The maps were reconstructed in electronic format, as two-dimensional, colour-coded visual representations, using Microsoft Excel programming. This allows for easy observation and interpretation of the macroscopic BE pattern distribution. BE patterns were described as Circumferential, Islands and Tongues and basic parametric statistics were applied.
Results Fifty-nine patients (92%) had a hiatus hernia (mean length 2.9 cm). Mean length BE was 7.6 cm; only two patients had short segment (<3 cm). Overall, 55 patients (85%) had some element of circumferential BE (mean length 6.1 cm) and 50 patients (77%) had at least one tongue of BE (19 had 2 tongues). The mean tongue length was 2.7 cm and the mean circumferential extent of tongues was 49%. Most tongues were cephalad in shape, but 15% of tongues also had circumferential projections. In 11 patients (17%), there was also at least one island of squamous tissue and three patients had islands of columnar tissue cephalad to the confluent BE segment. If the presence of tongues or islands were used as the criteria for automatic detection of BE, 80% of our patients would have been correctly diagnosed.
Conclusion We have developed an electronic mapping tool which provides a quick and simple method of describing macroscopic visual patterns in BE. The morphology of BE usually involves tongues or islands. Automated detection based on these features would identify 80% of cases. We need to apply this mapping to patients with non-dysplastic and short segment BE, as well as to patients with oesophagitis, to see if its use is more widely applicable.