Skip to main content

Table 1 Statistical comparison of the values of reference manual parameter measurements (GT) and those obtained automatically by the prediction model (PR)

From: Landet: an efficient physics-informed deep learning approach for automatic detection of anatomical landmarks and measurement of spinopelvic alignment

 

MAE

Mean PR (±SD)

Mean GT (±SD)

Acc (%)

DS1

DS2

DS1

DS2

DS1

DS2

DS1

DS2

SS\(^\circ \)

4.8

6.2

42.5 (8.7)

34.9 (11.8)

38.9 (10.3)

30.5 (10.4)

90.2

84.7

PT\(^\circ \)

1.8

2.4

18.1 (10.6)

21.7 (21.4)

19.3 (9.4)

22.1 (19.5)

91.1

89.2

PI\(^\circ \)

5.7

7.9

53.6 (10.2)

62.4 (10.7)

50.8 (12.6)

56.7 (13.7)

90.1

86.6

LL\(^\circ \)

6.7

7.5

48.5 (21.4)

33.9 (17.1)

46.2 (16.2)

37.1 (17.8)

88.6

81.2

SVA\(_{mm}\)

2.5

3.6

38.2 (3.6)

33.1 (2.5)

34.9 (3.2)

29.8 (2.1)

92.8

83.1

  1. The mean average error (MAE) are presented along with the standard deviation (SD) of each parameter evaluated in the two datasets (DS1 & DS2). The relative accuracy of prediction (Acc) is also calculated and presented in last two columns