Table 3 Metal artifacts in conventional and deep learning images by crown type.

From: Evaluation of deep learning MRI reconstruction for dental implant crowns in a phantom study

Type of crown

T1- weighted

T2-weighted

Visual score (conventional/DL)

Artifact ratio (%) (conventional/DL)

Visual score (conventional/DL)

Artifact ratio(%) (conventional/DL)

Zirconia

4.5/4.5

15.21/27.64

4.5/4.5

10.38/9.31

PMMA

4.5/4.5

22.60/39.93

4.5/4.5

9.86/9.94

Gold

4.0/4.0

28.96/48.23

4.0/4.0

20.28/21.36

Ni–Cr metal

1.0/1.0

47.36/59.41

1.0/1.0

69.16/78.89

  1. PMMA polymethyl methacrylate, Ni–Cr nickel-chrome, DL deep learning.
  2. Visual scores were assessed by two radiologists using a 5-point ordinal scale (1 = severe artifacts, 5 = minimal artifacts), and the average of the two scores was used for analysis. Inter-observer agreement was moderate (Cohen’s κ = 0.429).
  3. Artifact ratios were measured by one radiologist who repeated the analysis twice; the average of the two measurements was used. Intra-observer reproducibility was excellent (ICC = 1.0, 95% CI 0.999–1.000).
Back to article page