Pleural is an autonomous mucus management device for the 3.9 million Britons suffering from debilitating airway mucus that can lead to infections and exacerbations that increase the likelihood of hospitalisation.
Pleural uses a silicone cup to apply a calculated force to the chest at particular frequencies to mobilise mucus from the lungs.
The team at Pleural observed that a doctor’s finger-tapping percussion on a patient’s chest during auscultation generates impulse signals, which are interpreted alongside breathing sounds using a stethoscope to analyse the chest.
This takes years of training but advances in machine learning have allowed the team at Pleural to train AI to perform the task. They found they could use the same percussive mechanism to provide the input for AI-powered auscultation and for the percussive chest clearance therapy itself. Users can perform airway clearance therapy at home, with the device providing guidance and progress tracking through the companion app to ensure thorough treatment.
Daniel Hale, engineering lead at Pleural, explained that the device requires calibrating to each user.
“It is preloaded with a set of training data to provide a basis for informed treatment, but this is generic and lacks insight, so it really comes into its own after collecting data on each specific user,” said Hale.
“An initial in-depth calibration cycle will take lots of samples from different positions around the chest, record sounds and ask the patient a series of health and wellbeing questions. This builds a profile of the congestion characteristic of each user.”
Hale continued: “This initial phase can be done under supervision by a medical practitioner and/or a relative. Similarly, the user's first experience will involve setting up the app and tutorial, which they may feel more comfortable doing with supervision.”
The device and dock could be set on a bedside table, next to where the user keeps their smartphone. In the morning they can open their pleural app and start a treatment cycle through a simple touch interface.
“The user would rest their phone on the dock and face towards the device, such that the front-facing camera can see the user's torso, and overlay guidance graphics on to it in real-time,” said Hale. “They would then be guided through a set of positions in which to place the handheld device on the chest, where it would cycle through analysis and therapeutic percussion, giving time for the user to breathe and cough effectively in between, to release and bring up their mucus.”
Hale added that data ‘lives with a docking station’, which serves several functions including charging, providing a smartphone stand and housing much of the processing power. Initially data could be stored offline but for advanced features, it would be necessary for the dock to have WiFi connectivity to send data to the machine learning model housed remotely on cloud servers.
“The user would provide consent for their data to be used in such a way,” said Hale. “Outputs from the machine learning would include condition insights, trends, and alterations to the treatment plan. Data and conclusions can also be provided to a healthcare provider through a partner app. This may inform the medical professional of the need for an appointment or further investigation.”
Hale added: “There would also be benefits in user data being used to train a much larger model which combines data from all users. We believe this would ultimately provide the highest quality of insight but we only want users to contribute to such a model if they are comfortable with the data privacy implications.”
Hale worked with Will Eliot, Fergus Laidlaw and Yihan Dong on Pleural, which has won the team £5000 to produce a fully functional, technologically integrated prototype. Pleural will now progress to the international stage of the James Dyson Award.