Computer-aided diagnosis of neoplastic colorectal lesions using 'real-time' numerical color analysis during autofluorescence endoscopy

Eur J Gastroenterol Hepatol. 2013 Apr;25(4):488-94. doi: 10.1097/MEG.0b013e32835c6d9a.

Abstract

Objective: Differentiating non-neoplastic colorectal lesions from neoplastic lesions during screening colonoscopies is essential to reduce the unnecessary treatment of non-neoplastic lesions. The present study was conducted to verify the diagnostic yields of the computer-aided diagnostic system that enables 'real-time' color analysis of colorectal lesions when applied to autofluorescence endoscopy (AFE).

Patients and methods: Consecutive patients who were scheduled to undergo a therapeutic colonoscopy in our department were enrolled in this study. The encountered lesions were evaluated in AFE and color-tone sampling was performed. Lesions with green/red (G/R) ratios less than 1.01 were judged to be neoplastic and those with G/R ratios of at least 1.01 were considered to be non-neoplastic. All lesions greater than 5 mm were endoscopically removed and lesions less than 5 mm were biopsied.

Results: During the study period, a total of 32 patients with 102 colorectal lesions were evaluated with AFE. The mean G/R ratio for all neoplastic lesions was 0.86 [95% confidence interval (CI), 0.63-1.01], which was significantly lower than the mean G/R ratio for non-neoplastic lesions (1.12; 95% CI, 0.98-1.26; P<0.001). The mean G/R ratios were 1.36 (95% CI, 1.21-1.57) in normal mucosa, 1.12 (95% CI, 0.98-1.26) in hyperplastic lesions, 0.88 (95% CI, 0.69-1.02) in adenomas, and 0.61 (95% CI, 0.54-0.73) in intramucosal cancers. A G/R ratio cutoff value of 1.01 was applied for discriminating between neoplastic lesions and non-neoplastic lesions, and yielded sensitivity, specificity, positive and negative predictive values of 94.2, 88.9, 95.6, and 85.2%, respectively.

Conclusion: This diagnostic tool may lead to the reduction of unnecessary treatments for non-neoplastic lesions.

MeSH terms

  • Adenoma / diagnosis
  • Adenoma / pathology
  • Adenoma / surgery
  • Colon / pathology
  • Colonoscopy / methods
  • Color
  • Colorectal Neoplasms / diagnosis*
  • Colorectal Neoplasms / pathology
  • Colorectal Neoplasms / surgery
  • Computer Systems
  • Diagnosis, Computer-Assisted / methods*
  • Diagnosis, Differential
  • Humans
  • Hyperplasia / diagnosis
  • Image Interpretation, Computer-Assisted / methods*
  • Optical Imaging / methods
  • Predictive Value of Tests
  • Prospective Studies
  • Software Design