Thermal time constant: optimising the skin temperature predictive modelling in lower limb prostheses using Gaussian...

Area of Science:

• Biomedical Engineering
• Materials Science
• Machine Learning

Background:

• Elevated skin temperature at the body/device interface of lower-limb prostheses negatively impacts tissue health.
• Heat dissipation in prosthetic sockets depends on the thermal properties of socket and liner materials.
• Monitoring skin-level interface temperature in prosthetic sockets is challenging due to flexible liners and sensor positioning issues.

Purpose of the Study:

• To explore predicting residual limb temperature by monitoring the interface between the socket and liner.
• To investigate the use of machine learning, specifically Gaussian processes, for temperature prediction.
• To highlight the importance of the thermal time constant of prosthetic materials in improving prediction accuracy.

Main Methods:

• Utilized Gaussian processes, a machine learning algorithm, for temperature prediction.
• Incorporated thermal time constant values of common prosthetic socket and liner materials into the model.
• Focused on monitoring temperature between the socket and liner as a proxy for skin temperature.

Main Results:

• Demonstrated the relevance of the thermal time constant in Gaussian processes for predicting prosthetic socket temperature.
• Showcased a method for non-invasively monitoring residual limb skin temperature.
• Indicated that incorporating the thermal time constant optimizes and generalizes the prediction model.

Conclusions:

• Predicting residual limb temperature via socket/liner interface monitoring is a viable strategy to reduce heat and perspiration complaints.
• The thermal time constant is a crucial parameter for enhancing the reliability of machine learning models in prosthetic thermal management.
• This approach offers a more reliable and generalized method for monitoring prosthetic socket temperatures.