Published In

Authors

Abstract

Purpose:
To assess the diagnostic performances of artificial intelligence (AI)-based software to perform automatic measurements on frontal and lateral foot radiographs.

Methods:
We retrospectively collected 112 consecutive frontal foot radiographs and 130 consecutive lateral foot radiographs from 2 private imaging groups.
Two senior skeletal radiologists independently annotated key points to calculate the hallux valgus angle, the M1-M2 angle, the M1-M5 angle on the frontal foot images and the talus and M1 angle, the medial arch angle, and the calcaneal inclination angle on the lateral foot images. The gold standard was defined as the mean of their two measurements.
Statistical analysis consisted of the mean absolute difference (MAE) between the gold standard and the AI prediction and intraclass coefficient (ICC) between the two manual ratings.

Results:
A total of 85 frontal foot images were included and 27 images were excluded. For the lateral foot images, 97 were included and 33 were excluded. The MAE for the hallux valgus angle, the M1-M2 angle, and the M1-M5 angle on the frontal foot images was respectively 1.2° (95% CI: [1 ; 1.4], ICC=0.98), 0.7° (95% CI: [0.6 ; 0.9], ICC=0.91) and 0.9° (95% CI: [0.7 ; 1.1], ICC=0.96). The MAE for the talus and M1 angle, the medial arch angle, the calcaneal inclination angle on the lateral foot images were respectively 3.9° (95% CI: [3.4 ; 4.5],
ICC=0.55), 1.5° (95% CI: [1.2 ; 1.8], ICC=0.95) and 1° (95% CI: [0.8 ; 1.2], ICC=0.99).

Conclusion:
The AI can accurately and automatically predict measurements on frontal and lateral foot radiographs.