Ambient Intelligence and Emotion AI are two, somewhat related concepts, and both have the potential to significantly improve future healthcare in many areas, such as addressing capacity issues and preventing staff burnout, improving patient outcomes and supporting the autonomous living of the elderly.
Let’s see what these terms mean and what their practical applications will be.
Ambient intelligence means – in laymen’s terms – a technology-supported environment that is sensitive and responsive to human presence with the technology part remaining as close to invisible as possible.
To cite a simple example, a sensor- and AI-based fall detection and prevention solution in an elderly care facility is an example of ambient intelligence: the sensors and physical devices, as cameras are embedded in the environment, the algorithm runs in the background and it provides added value to the patient (more safety, help to arrive faster) and to healthcare providers (more accurate information, immediate alerts in case of a patient being in risk of falling or of a fall).
Emotion AI is often also called affective computing and is about detecting emotions using artificial intelligence and interacting with humans in a way that suits their current emotional state. It can identify emotions by measuring micro-expressions that are too fast and subtle for a human eye to capture and is used in myriad areas of life from designing more efficient sales techniques to providing better healthcare.
Both of these concepts can potentially introduce a number of problematic ethical and/or privacy issues, but for now, we focus on the upside and on what they can deliver.
This is what ambient intelligence can do for us in the healthcare universe
While ambient intelligence is not a new concept, (it was first set as a goal in the 1990s by Eli Zelkha), the required technologies only matured enough to provide practical value in recent years. Here come a few examples of what it is/can be capable of:
Intensive care unit (ICU) use cases
Ambient intelligence can be used to closely monitor patient mobilisation, and thus help prevent ICU-acquired weaknesses. Sensors and machine learning algorithms can effortlessly keep track of the activities of patients, freeing up time and mental stress for healthcare personnel.
Another potential use case is keeping track of proper compliance with hand hygiene protocols, promising better control over hospital infections. AI turning off non-essential alerts in an emergency setting also belongs here: getting rid of all the beeps that don’t require their attention, alarm fatigue of ICU personnel can be successfully reduced – which in turn generates better patient outcomes for meaningful alerts.
Use cases in the operating room
Ambient intelligence solutions can be used to keep track of surgical count, preventing objects – sponges, needles, instruments – from being accidentally retained inside the patients and thus significantly lowering post-operative complications.
These solutions can also be capable of assessing surgical skills, potentially leading to better training and better feedback for practising professionals.
Use cases in general medical practice
A system built using deep learning models and various sensors and devices, such as microphones can assist medical professionals with administration, performing tasks similar to a medical scribe. We all know how huge an improvement it would mean if healthcare workers could be freed of the burden of administration after each patient’s visit.
Another – although ethically shaky and as of now technically challenging – potential of ambient intelligence is to use it to assess the general condition of patients while they are waiting for their appointment. In theory, various sensors and algorithms can be used to keep track of a number of vital signs. They could create a preliminary report of the basics before the patient actually enters the doctor’s surgery.
An exciting application of ambient intelligence is to use cameras, depth sensors (etc) to measure the gait patterns of patients in a “natural” environment (eg: a room or corridor), without the need of attaching sensors to the patients. Accurate gait analysis could improve the health outcome of patients in many cases, from earlier Parkinson’s detection to the correct assessment of postoperative health in several conditions.
Use cases in elderly care facilities and at home
As mentioned above, fall prevention is a prime example of using ambient intelligence in elderly care. With cameras, sensors and algorithms it is possible to accurately and automatically track the movement and activity patterns of patients, easing the mental burden of patients, healthcare professionals and relatives. These systems alert caregivers in case they detect a risk of fall or a fall, prolonged inactivity, or lack of movement in a position that is not “natural”, like a patient sitting on the floor instead of the chair or the bed.
Contactless ambient sensors can also be used to assess activities of daily living of an individual – activities like eating, dressing up or taking a shower. These are critical factors to the independence and well-being of a person, but typically they are not tracked very reliably – as monitoring mostly relies on self-reports and/or infrequent personal checks. If one’s surroundings were to track these automatically, timely intervention could slow the functional decline of individuals.
What emotion AI can do for us
Detecting subtle emotional states and changes can offer valuable meta-information in the healthcare setting.
- Emotion AI can monitor non-verbal cues in a remote healthcare appointment, such as video calls, giving more information to doctors.
- It can help autistic people to interpret others’ feelings, and thus improve their social skills. With emotion AI, they are able to better understand the emotional condition of those with whom they are communicating.
- It can monitor waiting patients for signs of discomfort to determine the ones in need of urgent care.
- Emotion AI can look for facial expressions, speech- and behavioural patterns in dementia patients, accurately predicting their emotional state. It can not only notify caregivers in advance of the situation but can also adjust the patients’ environment accordingly, like selecting the right kind of music.
Many aspects of both emotion AI and ambient intelligence will be a regulatory nightmare. Comprehensive regulations will be crucially important – after all, you don’t want your convenience store filled with sensors which can better determine what you will buy after the right stimuli than you yourself.
On the other hand, as the capabilities of various sensors and devices develop, both concepts will provide a growing number of healthcare applications, easing the burden and burnout of doctors and nurses, increasing patient safety, and offering better screening among others. It is a fascinating field, we will certainly keep our eyes (and sensors) on it.