Here’s an interesting example of how researchers and product designers are combining AI with sensors to yield greater insights.
At the University of Missouri, researchers used in-home sensors combined with AI to monitor daily changes in ALS patients’ health in hopes of allowing earlier interventions to improve quality of life. They set out to track ALS (amyotrophic lateral schlerosis) progression real time and detect subtle shifts in health before doctors or patients notice the changes to predict how a patient might score on the ALS Functional Rating Scale Revised. The shifts include changes in gate while walking and sleeping patterns.
At the heart of the process are sensors in the home that transmit data wirelessly to university data centers for analysis. Two professors at University of Missouri originally developed the sensors to monitor the health of older adults living at home, and now researchers have adapted the sensors to use in ALS monitoring. ALS patients experience functional decline similar to older adults, but more rapidly and unpredictably.
The current focus is on verifying that the sensor data accurately reflects real-world changes in how patients function day to day, while the next phase will make senses of the collected data with predictive modeling using AI. The models are built to estimate each patient’s score on the ALSFRS-R tool, which includes walking, talking, swallowing and breathing.
“Our goal is to not just track changes after they happen; we’re also trying to see them in advance,” said Noah Marchal, a research analyst leading the project’s data science efforts.
The final stage of the project is to integrate the system into clinical workflows so ta clinician would receive an alert to check in with a patient and perhaps adjust medication, recommend assistive devices or suggest other treatment.
“Our vision is that one day clinicians will have a secure portal where they can view a patient’s daily health trends the way ICU teams monitor telemetry,” said Bill Janes, an assistant professor in Mizzou’s College of Health Sciences. “With these sensors, we can detect subtle shifts in health sooner, sometimes even before a patient feels them, and act before a crisis occurs. Right now, we’re essentially blind to what’s happening between clinic visits.”
Several of the researchers posted their findings in "Frontiers in Digital Health."
They noted that wearable sensor data, including from a wearable accelerometer, ECG and digital speech sensors for tracking ALS have been effective and suggested more frequent, remote (non-wearable) sensor-based tracking of changes in ALSFRS-R scales “would enable clinicians to better target interventions and detect acute events, such as falls or medication changes, between clinic visits.”
For the project, researchers used in-home sensor monitoring systems licensed by the University of Missouri to Foresite Healthcare LLC to provide three modes of continuous contactless data collection: bed mattress hydraulic transducers for recording respiration, pulse and sleep restlessness; thermal depth sensors to detect falls and collect walking speed, stride time and stride length measurements; and passive infrared motion sensors to provide room activity counts.
Researchers said their preliminary findings showed that integrating passive in-home sensor monitoring into routine ALS care “may help clinicians better detect and anticipate functional changes between quarterly assessments, differentiating stable periods from more rapid decline.”
The University of Missouri project currently focuses on ALS, but the same research could be adapted to help monitor other chronic conditions such as Parkinson’s disease or heart failure.