West meets East: preliminary results of narrow band imaging with optical magnification in the diagnosis of colorectal lesions: a multicenter Australian study using the modified Sano's classification

Dig Endosc. 2011 May;23 Suppl 1:126-30. doi: 10.1111/j.1443-1661.2011.01107.x.


Introduction: Narrow band imaging with optical magnification (NBI-Z) enables mucosal morphology to be assessed in real time by using light with narrowed band width and magnification of up to 115×.

Methods: Colorectal lesions detected were assessed with NBI-Z. Histology was predicted using the modified Sano's classification based on capillary network patterns (cn); type I: absent cn (hyperplastic polyp), type II: cn present, surrounding mucosal glands (adenoma), type IIIa: high density cn with tortuosity and lack of uniformity (intramucosal cancer) and type IIIb: nearly avascular cn (invasive cancer). Each lesion was also graded with a confidence level (low/high). High-definition videos (mean 28.2 s; range 12-55) of each lesion assessed with NBI-Z were then taken. This was followed by polypectomy, endoscopic or surgical resection. NBI-Z diagnosis was compared with the final histopathology. To test for interobserver agreement, an endoscopist blinded to the video acquisition process/histology was invited to grade the videos.

Results: A total of 50 lesions (2 assessors: 100 studies), with an average size of 8.4 mm (range 3-30), in 32 patients were assessed. Twenty were hyperplastic, 25 adenomas, 2 intramucosal and 3 invasive cancers of which 19 were located in the right and 31 in the left colon. The overall accuracy of NBI-Z in predicting histology was 90%, which increased to 95% (88/93) when lesions were predicted with high confidence. The sensitivity (Sn), specificity (Sp), positive (PPV) and negative predictive values (NPV) in differentiating neoplastic from non-neoplastic lesions with high confidence were 98%, 89%, 93% and 97%, respectively, while the Sn, Sp, PPV and NPV in predicting endoscopic resectability (type II, IIIa vs type I, IIIb) was 100%, 90%, 93% and 100%, respectively. The interobserver agreement between both assessors (κ value) was substantial at 0.89.

Conclusions: Using confidence levels, NBI-Z permits prediction of colorectal neoplasia with high accuracies and might allow prompt decisions to be made if a lesion should be left in situ, resected and discarded or biopsied. This approach might lead to substantial time and cost savings and could potentially reduce complications associated with polypectomy and endoscopic resections.

Publication types

  • Comparative Study
  • Multicenter Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Australia
  • Colonoscopy / methods*
  • Colorectal Neoplasms / classification
  • Colorectal Neoplasms / diagnosis*
  • Diagnosis, Differential
  • Humans
  • Image Enhancement / methods*
  • ROC Curve