Skip to main content
Insights into Imaging

Table 4 Radiological features, algorithm predictions and inter-rater agreement in patients with ‘target reports’ (n = 146)

From: Emergency department CT examinations demonstrate no evidence of early viral circulation at the start of the COVID-19 pandemic—a multicentre epidemiological study

Characteristics

Patientsa

Inter-rater agreementb

SFR-like classificationc

 

0.656 (0.601–0.717)

 V-V-V

22/146 (15.1%)

 

 V-V-IV

3/146 (2.1%)

 

 V-IV-IV

9/146 (6.2%)

 

 IV-IV-IV

3/146 (2.1%)

 

 Others

109/146 (74.6%)

 

CART predicted probabilities

0.739 ± 0.141

0.288 (0.2–0.39)

Step-LR predicted probabilities

0.622 ± 0.243

0.532 (0.464–0.597)

GGO

144/146 (98.6%)

0.904 (0.885–0.923)

GGO predominant pattern

110/146 (75.3%)

0.367 (0.263–0.46)

Intralobular reticulations

60/146 (41.1%)

0.33 (0.243–0.417)

Subpleural predominant pattern

105/146 (71.9%)

0.399 (0.311–0.487)

bronchitis syndrome

41/146 (28.1%)

0.275 (0.188–0.382)

Diffuse lesions (≥ 2 lobes)

100/146 (68.5%)

0.498 (0.402–0.589)

Band-like condensations

35/146 (24%)

0.366 (0.255–0.472)

  1. Abbreviations: CART classification and regression trees, GGO ground glass opacities, SFR French ‘Société Française de Radiologie’, Step-LR stepwise logistic regression
  2. aData are number of patients (with percentage in parentheses), except for probabilities for positive RT-PCR according to the predictive models, which are mean probability ± standard deviation
  3. bInter-rater agreement is Krippendorff’s alpha with 95% confidence interval
  4. cThe findings for SFR-like classification correspond to the 3 readings by senior radiologists X1-X2-X3, where Xi ∈ {I, II, III, IV, V}, arranged in descending orders
Back to article page