Artificial Intelligence and Machine Learning in Medical Sensor Technology
Artificial Intelligence (AI) is known for its ability to analyze big data sets, to continuously learn and to subsequently offer predictive analysis based on that learning. According to Business Insider Intelligence, AI in healthcare is projected to grow 48% annually between 2020 and 2023. The potential impact of this projected growth on how health services are delivered, particularly within the subset of medical sensor technology, is compelling.
Intelligent Dynamic Sensing
“Intelligent Dynamic Sensing” created by XSENSOR Technology, are sensors that provide real-time actionable information to help clinicians improve patient safety, with an integrated combination of smart data and artificial intelligence. Skin “pressure exposure” is continuously monitored and AI-powered algorithms inform advanced prevention strategies to maintain healthy skin for at risk patients. Sensors within surgical tables and hospital beds precisely identify where persistent pressure on the body is occurring, informing clinicians on where and how to reduce the pressure to prevent injuries from occurring in the first place.
The Cost of Hospital Acquired Pressure Ulcers/Injuries (HAPU/I’s)
In the US, more than 2.5 million people contract a HAPU/I every year and contribute to approximately 60 thousand patient related deaths as a result of those HAPU/I’s alone. In just this area, AI’s ability to quickly obtain and digest patient insights offers the opportunity for a radical improvement in patient safety and care.
Machine Learning (M/L) is a subset of AI, designed to identify patterns using algorithms and data to offer immediate and automated insights. In healthcare, M/L has the ability to provide data-driven clinical decision support to physicians and hospital staff. The integration of AI into the healthcare ecosystem generally allows for a multitude of benefits, including automation and the efficient analysis of massive patient data sets. This means better healthcare delivered faster, and at a lower cost.
Pioneering the capability of ML to alleviate HAPU/Is could improve patient outcomes and, in some circumstances, avoid pressure injuries altogether. The cost of treating later-stage pressure injuries can be as high as $70,000 each, and that the total treatment cost of HAPU/Is in the US is approximately $11 billion a year. Ultimately, AI and M/L also pave the way for dramatic cost savings as well as improved patient outcomes on the part of physicians and hospitals alike.
Wearable Sensing Technology
Data from medical sensing technology is central to an infrastructure where information is continuously collected and fed into algorithms to monitor vital signs, spot abnormalities and track treatments. Wearable or contactless sensors are an example of broadly adopted, unobtrusive, AI-enabled sensing technology. XSENSOR’s advanced pressure imaging for foot and gait assessment and measurement are one such example. With integrated intelligent dynamic sensing, they reveal accurate gait and motion data to help biomechanics professionals optimize human performance.
Risks and Benefits
With all the innovation and technical development in this area -- in the context of the overriding principal of healthcare providers to “Do No Harm”-- how do we ensure that integrated AI and M/L technologies provide certainty around labor savings, treatment optimization or the prevention of injuries like pressure injuries, while avoiding negative or unintended consequences?
As AI becomes increasingly easy to use and part of everyday healthcare decisions, perhaps the real issues have to do with the limits of its power and how we use it:
How do we ensure there’s a broad enough set of training and test data to begin with?
How do we make the algorithms’ decisions more transparent and their data more reliable?
Which decisions are safe to leave to technology, and which require human judgment?
Smart data from intelligent dynamic sensors can locate areas of higher probability of pressure injury.
Combine that with a hospital bed that has the mechanical capability to adjust the inclination or firmness and contours of the surface, and with M/L connected to the mechanics, the mattress can adjust to shift the pressure on the patient and avoid a potential HAPU/I.
Establishing the parameters for proactive M/L will take time and careful clinical testing. The potential to ensure patient safety, improve patient outcomes, and reduce the demands on critical, professional hospital staff, however, is a matter of time and study. Testing with human monitored and approved remedial measures will be an interim step.
Under ideal conditions, AI and the associated M/L within healthcare should be a complement to, rather than a replacement for human care. AI should help clinicians “predict” outcomes, but the decisions, actions and recommendations for future interventions and care will remain a human task until safe assignment to M/L can be assured. Until then, technology will act like a kind of ‘driver assistance’ system for clinicians.
The measurement of success in AI sensing integrations in healthcare should be their ability to consistently help detect and prevent injuries from occurring in the first place. Big data is vital to help arrive at and maintain this standard of care, and the more data that’s gathered, the greater AI’s rate of dependable success. Ultimately, however, it’s the conversion and interpretation of big data into smart data that’s key, and it’s the human analysis and action taken in response that will drive positive outcomes in patient injury prevention, care, and associated financial savings as a result.