Table 2 Metrics of inter-reader difference and agreement for femoral and tibial torsion as determined by the Lee and ellipses methods.

From: Artificial intelligence-based automatic assessment of lower limb torsion on MRI

 

Method

Comparison

Absolute difference [°]

Pearson's r

ICC

R2

Alg

R2

Alg

R2

Alg

Femoral torsion

Lee

R1

2.6 (1.9; 3.3)

3.3 (2.5; 4.0)

0.984 (< 0.001)

0.968 (< 0.001)

0.974 (0.96, 0.99)

0.963 (0.93; 0.98)

R2

na

3.7 (3.0; 4.5)

na

0.971 (< 0.001)

na

0.966 (0.94; 0.98)

Tibial torsion

Ellipses

R1

5.0 (4.0; 6.0)

4.2 (3.2; 5.1)

0.871 (< 0.001)

0.904 (< 0.001)

0.865 (0.77; 0.92)

0.867 (0.78; 0.92)

R2

na

5.1 (4.1; 6.2)

na

0.867 (< 0.001)

na

0.894 (0.82, 0.94)

  1. Readers were the two radiologists (radiologist 1 [R1], radiologist 2 [R2]) and the algorithm (Alg). In a pair-wise manner, absolute differences [°] and agreement were quantified in terms of Pearson’s correlation coefficient r and the intraclass-correlation-coefficient (ICC, single scorings [not adjusted]). Means (95% confidence intervals). na – not applicable. Results of the other methods of determining femoral (according to the Reikeras, Tomczak, and Murphy methods) and tibial torsion (according to the bimalleolar and talus methods) are detailed in Supplementary Table 3.