Awarded Best Application Paper Award
internationally by IEEE/RSJ IROS (2022)
Using AI & Robotics for
Medical Examinations
As the population grows, healthcare systems such as the NHS become increasingly saturated. An ageing population also means that more health checkups are needed, so that health issues can be identified at early stages when they are easily treatable.
​
To tackle this issue, more emphasis and resources should be put on the automation of common regular examinations.
​
This project presents the research and development of a proof-of-concept AI-enabled device to examine the state of underlying organs from the comfort of your home.
Pilar Zhang Qiu | Jacob Tan | Oliver Thompson | Ben Cobley
Impact
Relieve Pressure from Healthcare Systems
Healthcare systems such as the NHS suffer from long backlogs. Automating regular appointments can help ease the system.
Accelerate Research through Big Data
Increased number of data points can help generate more accurate AI models to accelerate medical research.
Accesibility
for the Elderly
More than 1.45M elderly in the UK find it difficult to travel to hospital appointments. [1]
Remote devices can facilitate access medical care.
Process
Stage 1
Problem Definition
Literature Review
Robot Ideation
Robot Prototyping
Stage 2
Experiment Design
Jig Prototyping
Data Collection
Stage 3
Data Processing
Data Analysis
ML Models
System Summary
Robotic Device:
Percusses Phantoms
The device impacts the silicone phantoms, which represent the human abdomen.
Data Collection:
Audio Response
The acoustic data resulting from the impact is collected for 15 different abdominal structures, with a total of 7,500+ recordings saved.
AI Models:
Estimating Organ State
Predicted organ consistency and "size" using Gaussian Mixture Modelling (GMM) and Neural Networks (NN).
Results
The AI models showcased
97.5%
accuracy
in identifying 3 different organ consistencies
and 5 different organ "sizes".
We focused on
Explainable AI.
The biggest issue with AI is that it can be a "blackbox".
Our innovative method presents a solution for understandable AI-enabled research.
at the highly-influential IEEE/RSJ IROS 2022 Conference (Kyoto, Japan) for our contribution to the field.
We were honored to receive the
International
Best Application Paper Award