Sub-millisievert CT colonography: effect of knowledge-based iterative reconstruction on the detection of colonic polyps

Eur Radiol. 2018 Dec;28(12):5258-5266. doi: 10.1007/s00330-018-5545-5. Epub 2018 Jun 8.

Abstract

Objectives: To assess the feasibility of ultra-low dose computed tomography colonography (CTC) using knowledge-based iterative reconstruction (IR) and to determine its effect on polyp detection.

Methods: Forty-nine prospectively-enrolled patients underwent ultra-low dose CTC in the supine (100 kVp/20 mAs) and prone positions (80 kVp/20 mAs), followed by same-day colonoscopy. Thereafter, images were reconstructed using filtered back projection (FBP) and knowledge-based IR (IMR; Philips Healthcare, Best, Netherlands) algorithms. Effective radiation dose of CTC was recorded. Pooled per-polyp sensitivity and positive predictive value of three radiologists was analysed and compared between FBP and IMR. Image quality was assessed on a five-point scale and image noise was recorded using standard deviations.

Results: Mean effective radiation dose of ultra-low dose CTC was 0.90 ± 0.06 mSv. Eighty-nine polyps were detected on colonoscopy (mean, 8.5 ± 4.7 mm). The pooled per-polyp sensitivity for polyps 6.0-9.9 mm (n = 22) on CTC reconstructed with IMR (36/66, 54.5%) was not significantly different with that using FBP algorithm (34/66, 51.5%) (p = 0.414). For polyps ≥10 mm (n = 35), however, the pooled per-polyp sensitivity on CTC with IMR (73/105, 69.5%) was significantly higher than that with FBP (55/105, 52.4%) (p < 0.001). In particular, the difference of per-polyp sensitivity was statistically significant in intermediate (p = 0.014) and novice (p = 0.003) reviewers. Furthermore, mean image noise of IMR (8.4 ± 6.2 HU) was significantly lower than that of FBP (37.5 ± 13.9 HU) (p < 0.001) and image quality with IMR was significantly better than with FBP in all evaluated segments in all reviewers (all ps < 0.001).

Conclusions: Sub-mSv CTC reconstructed with IMR was feasible for the detection of clinically significant polyps, demonstrating 70% per-polyp sensitivity of polyps ≥10 mm, while allowing significant noise reduction and improvement in image quality compared with FBP reconstruction.

Key points: • Sub-mSv CTC using IMR demonstrated 70% per-polyp sensitivity for polyps ≥10 mm. • CTC using IMR significantly outperformed CTC reconstructed with FBP. • IMR allows significantly more noise reduction and improvement in image quality than FBP.

Keywords: Colonic polyp; Colonography, computed tomographic; Feasibility studies; Image reconstruction; Sensitivity.

MeSH terms

  • Adult
  • Aged
  • Algorithms*
  • Clinical Competence*
  • Colonic Polyps / diagnosis*
  • Colonography, Computed Tomographic / methods*
  • Colonoscopy
  • Female
  • Humans
  • Male
  • Middle Aged
  • Radiation Dosage
  • Radiographic Image Interpretation, Computer-Assisted / methods*