Researchers have demonstrated AI tools could play an important part in the development of simpler, lower-cost technologies for monitoring foot health, particularly in settings where access to specialised equipment may be limited.
A collaboration between The University of Queensland, iOrthotics and Healthia Limited has produced an AI model capable of reconstructing detailed foot pressure maps using information about foot shape and a small number of anatomical pressure points.
Applied mechanics engineer UQ Emeritus Professor Martin Veidt said plantar pressure analysis was widely used to assess a person's foot function and balance, gait mechanics and foot health, and to inform the design of orthotics that aim to minimise the risk and progression of foot-related pathologies.
"But existing measurement methods have limitations and are often costly and inaccessible for people living in rural and remote regions," Professor Veidt said.
UQ materials engineer Dr Stuart McDonald said in-shoe systems offer greater mobility and extended pressure monitoring but typically rely on a large number of sensors which can increase cost, complexity and power requirements.
"This study looked at the potential for AI to overcome some of the challenges associated with traditional plantar pressure monitoring systems," Dr McDonald said.
The researchers from UQ's School of Mechanical and Mining Engineering collaborated with iOrthotics and Healthia Limited to explore how a multimodal deep learning system could help unlock practical and more accessible methods to reconstruct dense plantar pressure information from sparse sensing.
Using anatomical foot information and plantar pressure measurements from 35 study participants, UQ PhD student Chongguang Wang was able to build an artificial neural network framework capable of generating accurate, high-resolution pressure maps using significantly fewer physical sensors.
The proposed deep learning model achieved its best performance using just 16 anatomical 'landmarks' from the bottom of the foot and achieved a comparable result using only 2 landmarks, indicating promising reconstruction performance even under very limited sensing conditions.
"This research demonstrates that by combining information about foot shape with only a small number of anatomical inputs, it is possible to reconstruct detailed plantar pressure distributions with a high degree of accuracy," Professor Veidt said.
"The beauty is that data collection could feasibly take place anywhere, including in isolated communities where health outcomes are poor and services may be limited."
The UQ-led study was part of a suite of research initiated by iOrthotics and parent company Healthia, together with researchers from QUT, through a $2.2 million Federal Government Cooperative Research Centres Projects (CRC-P) grant to develop smarter orthotic technology for people living in rural and remote regions.
Healthia's group chief education and research officer Kerrie Evans said the research formed part of a broader program at Healthia and iOrthotics exploring how emerging technologies could improve access to foot-health assessment and monitoring.
"Foot complications, including diabetic foot ulcers and amputations, continue to have a significant impact on individuals and health systems," Associate Professor Evans said.
"Our goal is to support the development of practical, affordable technologies that can help clinicians better understand foot function and identify potential problems earlier."
"While more research is needed, particularly in clinical populations and real-world settings, these findings demonstrate the potential for AI to play an important role in future foot-health monitoring technologies".
The research is published in Sensors.
