Video Analysis and Artificial Intelligence for the Analysis of Upper Limb Movement in Children

Video Analysis and Artificial Intelligence for the Analysis of Upper Limb Movement in Children. Validation of a Technique in a Pediatric Population Aged 6 to 17 Years

The principle of the study is to compare the data obtained using a shoulder movement analysis software with those obtained during a traditional clinical examination, that is, using a goniometer and the modified Mallet classification

The children are recorded performing 3 sets of shoulder movements (abduction, adduction, flexion, extension, external rotation 1, external rotation 2, internal rotation 2), first on the left and then on the right, at maximum active range of motion, chosen active range of motion, and maximum passive range of motion. The recordings are made by an RGB-D camera connected to a software (ShoulderLoc from B-com) equipped with artificial intelligence that, after image processing, determines the joint range angle of the shoulder for the given movement. This value is compared to the visual estimation of the examiner and its measurement using a goniometer. The hand-mouth, hand-neck, and internal rotation 1 movements are also performed and compared with the data from the modified Mallet classification.

Inclusion Criteria:

• Age between 6 and 17 years old at the time of inclusion
• No neurological pathology
• No history of upper limb surgery
• No upper limb trauma above the hand in the 6 months preceding the examination
• Ability to stand for a minimum of 2 minutes
• Consent from the child, both parents, and/or legal representatives for participation in a filmed clinical examination.

Exclusion Criteria:

• Inability to understand the different movements requested.