Can we objectively assess upper limb activity?

December 6, 2023

Can we objectively assess upper limb activity?

Upper limb function is mostly assessed using self-reporting outcomes such as the Disabilities of the Arm, Shoulder, and Hand Questionnaire (DASH) and its Short Form (Quick-DASH), and the Shoulder Pain and Disability Index (SPADI). As these questionnaires are mostly valid and reliable, there are some downsides to them. They rely for example, on response bias of the patient (i.e., are they scoring the items to satisfy the treating clinician?) or recall bias (i.e., Do they not remember what the actual movement behavior was like?). Therefore, adding objective upper limb functional monitoring in daily life to self-reported outcomes could lead to a more comprehensive, all-encompassing upper limb assessment. Wearable motion sensors could be used to objectively assess functional daily life upper limb activities. Afterwards machine learning techniques could be used to analyze and predict the amount of functional and non-functional upper limb activity. Machine learning is a subgroup of artificial intelligence, were mathematical algorithms are used for predictions based on identifying patterns in a data set. To study the accuracy of this machine learning technique, breast cancer survivors were video recorded at home and these results were compared with the machine learning technique. The participants were asked to perform a kitchen activity, a laundry activity, a shopping activity and a bed making activity as they would usually do. 

UL functional activities  

The accuracy of the machine learning model is 0.83 for the left side and 0.85 for the right side. These results indicate a good accuracy to detect upper limb functional activity from wearable motion sensors in daily life. However we noticed that there is still an overestimation in the prediction of functional activity using the machine learning model. Therefore we are conducting further research developing a new, more accurate, machine learning model. If you want to learn more about the used machine learning model and how we performed this research you can check out the full text over here.

By Nieke Vets