Pittsburgh is at the nexus of tech and healthcare

Dr. Joseph Aracri, Chair of the Allegheny Health Network Pediatric Institute; Dr. Kelly J. Shields, Senior Research Data Scientist for Highmark Health; and Dr. Samarjit Das, leader of the Intelligent Internet of Things group at Bosch Research in Pittsburgh; hold a microphone that will be coupled with artificial intelligence and machine learning to conduct sound analysis and potentially detect pediatric pulmonary conditions such as asthma.


This is the kind of innovation that happens when a city is a nexus of high tech and health care. Pittsburgh’s Highmark Health and Allegheny Health Network will be pioneering the use of an audio AI, created by Bosch Research, to help detect childhood asthma.

The device was originally used on the International Space Station (ISS) to monitor critical infrastructure through audio signal processing and machine learning.

We spoke to the team behind this initiative to find out more: Dr. Samarjit Das, leader of the Intelligent Internet of Things group at Bosch Research; Christopher Martin, Director of Research and Development for Bosch’s Research and Technology Center; and Kelly J. Shields, PhD, Senior Research Data Scientist at Highmark Health.

Describe the basis of the sound tech you’re developing.

Dr. Das: We’re using state-of-the-art AI to extract information or cues out of audio patterns. This is already being used, of course — smart speakers can recognize and understand speech. But that’s just one kind of sound. There are so many others that are giving you a lot of information about the state of things. For example, if you hear an interruption in the humming sound of your HVAC system, you might be suspicious something might be wrong. We’re developing tech to decode the language of things by decoding meaning out of noisy patterns. We came into contact with Kelly and her team to apply our concept to machine health monitoring.

How did the folks at Highmark Health get connected?

Dr. Shields: We hadn’t necessarily been looking for this type of technology. But we had a meeting with Bosch where they allowed our team to come into their facility and lab space to see quite a few projects they’re working on. I honed in on this one in particular and was very excited about the potential — listening for cues and applying it to the human body as a machine.

After a few internal conversations, it didn’t take long to make people understand the potential here. We are now on the path of figuring out what the next steps look like. This was a first for us at Highmark, to collaborate with a group like Bosch. We’re excited for this opportunity.

How did you isolate asthma as the condition on which to use the tech?

Dr. Shields: We held a discovery workshop. From there, we whittled down to some specific areas and then we approached a few physicians to see who might be interested in partnering with us. We used their expertise to identify use cases. We landed on asthma. It’s something where we can detect issues without having to put a sensor directly on someone. We’re just listening to the child cough and looking for sound clues.

Dr. Das: We are looking at really low frequency sounds that you and I cannot hear. There are some smartphone apps that claim to do some asthma monitoring or any breathing disorder monitoring, but we’re looking at frequencies at less than 10 hertz.

Where are you with the research?

Dr. Shields: We’ve done quite a bit of work up to this point. We’ve done a simple proof of concept with the configured microphone and in the software and were able to see differences between speech, coughing, and breathing. We’re going to start enrolling our patients this month.

Sometimes a parent can ‘hear’ something in a child’s breathing that a doctor may not.

Martin: We should never underestimate exactly what you’re talking about. We did work a while back with Carnegie Mellon University with the facility management professionals. They could walk into these very complex machine rooms and hear that something was wrong. It could be a vibration or sound, but they had a natural human intuition that was built up over months and years. They could feel something was wrong but couldn’t fully explain the relationship. These different AI techniques help us make that link that we as humans do all the time. We’re learning how to train machines or AI to do that.

It seems like Pittsburgh is the perfect spot to develop this tech.

Martin: This is a great example of the type of collaboration that’s possible in Pittsburgh. Bosch has been operating in this ecosystem since the late 90s. One of the reasons we’re here is because of the collaboration potential that we see and have been able to take advantage of. Our largest partner over the years has been Carnegie Mellon University. You mentioned the space station—our role there came about through a partnership with Astrobotic, a spinoff out of CMU.

Highmark is able to bring in the expertise of domain experts, the clinicians and data scientists. Whether it’s academia’s top tier institutions like Carnegie Mellon or the incredible startup scene here in Pittsburgh, there is great opportunity in this area.

For more information about Astrobotic, click here to read our interview with CEO John Thornton. We also interviewed Dr. Jeff Cohen and Rich Lunak about Highmark Health and the AlphaLab Health accelerator in Pittsburgh.