Computer-aided detection for computed tomographic colonography screening: a prospective comparison of a double-reading paradigm with first-reader computer-aided detection against second-reader computer-aided detection

Invest Radiol. 2014 Mar;49(3):173-82. doi: 10.1097/RLI.0000000000000009.

Abstract

Objectives: The objective of this study was to prospectively compare diagnostic performance and time efficiency of a double-reading paradigm in which a first-reader computer-aided detection (CAD) is followed by a fast 2-dimensional review (DR FR-CAD) with those of a double reading with second-reader CAD (SR CAD).

Materials and methods: The local ethical committee approved this study. Consecutive immunological patients who have positive results for fecal immunological test who were scheduled for colonoscopy were enrolled for a 10-month period. Computed tomographic colonography studies were read with CAD (CAD COLON-1.20; im3D, Turin, Italy) by using both SR CAD (applied after unassisted interpretation primary 2-dimensional) and DR FR-CAD (CAD-prompts evaluation followed by a fast 2-dimensional review) in randomized order with the radiologist for each reading paradigm masked to the other reader's results.Per-patient sensitivity and specificity of unassisted and CAD-assisted readings for detecting 6-mm adenomas or larger were calculated by using unblinding colonoscopy as reference standard. Reporting times were also calculated. Pairwise comparisons were performed.

Results: A total of 182 participants (median age, 65 years; range, 58-76) were included in the final analysis. Of these, 93 (51%) had at least 1 cancer or a 6-mm adenoma or larger. At the 6-mm threshold, sensitivity of unassisted reading (79.6%; 95% confidence interval [CI], 69.9-87.2) increased significantly with the use of both SR CAD (86.0%; 95% CI, 77.3%-92.3%) and DR FR-CAD (89.2%; 95% CI, 81.1%-94.7%), without differences between CAD readings (P = 0.500). No significant differences in specificity among the 3 paradigms were observed. Double reading with first-reader CAD required less reading time than that for SR CAD (378 vs 496; Δ118 seconds; P < 0.001) and was 59 seconds longer than the unassisted reading (P = 0.058).

Conclusions: When compared with unassisted reading, a double-reading paradigm in which first-reader CAD is followed by a fast 2-dimensional review improves the adenoma detection rate to the same level achieved by a second-reader CAD while decreasing reporting times.

Publication types

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

MeSH terms

  • Aged
  • Algorithms*
  • Colonography, Computed Tomographic / methods*
  • Colonography, Computed Tomographic / statistics & numerical data
  • Colorectal Neoplasms / diagnostic imaging*
  • Early Detection of Cancer / methods*
  • Early Detection of Cancer / statistics & numerical data
  • Female
  • Humans
  • Male
  • Middle Aged
  • Pattern Recognition, Automated / methods*
  • Prospective Studies
  • Radiographic Image Enhancement / methods*
  • Radiographic Image Interpretation, Computer-Assisted / methods*
  • Reproducibility of Results
  • Sensitivity and Specificity