SINGAPORE (July 1): Local entrepreneur Lim Meng Hui was working on an artificial intelligence (AI)-enabled security system to detect intruders when his grandmother fell and lay helpless. When her health deteriorated rapidly after that, he decided that detecting falls among the elderly was a far more critical need that his start-up could address.
“Although she was [only] slightly injured, she couldn’t get up by herself. She screamed for help but nobody was around. About 30 minutes later, someone found her and sent her to the hospital,” Lim recounts.
“After that incident, she recovered but she had a fear of walking. Although she could walk, she chose to rely on a wheelchair. We noticed that she became weaker and weaker. Her dementia got worse and she passed away 12 months later.
“That triggered [a thought]: that a lot of elderly people often die alone and are found several days later,” he says. “Why not develop an AI application to protect the elderly?”
Lim and co-founder Toh Ying Wei started SmartPeep, a tongue-in-cheek moniker for what is shaping up to be an essential tool for an ageing population. The system is currently on trial at a hospital and nursing home in Malaysia, as well as a public hospital in Singapore and a retirement village in Australia.
SmartPeep plugs into a facility’s closed circuit television system, and sets off an alarm when it detects that a person has fallen over, alerting caregivers to the situation. The system is also smart enough to differentiate between patients and nurses, and will turn off the alert when a nurse enters the room to attend to the patient. This allows caregivers to concentrate on their job rather than spend time dealing with the technology.
Healthcare facilities that have patients with chronic conditions, such as dementia, would typically need a nursing staff-to-patient ratio of one to one or one to two. SmartPeep aims to help ease the nurses’ workload of making multiple checks on wards.
“Falls and accidents happen in between rounds as well. How often can you do the rounds? Why not let technology take over that and give nurses and caregivers more time to provide a human touch to the patients,” Toh says. “[Hiring more nurses] is not a [viable] solution. Our aim is to provide a [viable] solution to assist the nurses to provide care in elderly homes and hospitals.”
Lim says: “The direction we’re going in is to support caregivers to monitor a larger group of elderly people more efficiently. We’re building tools to help them realise this.”
Lim and his team have been working on SmartPeep since 2017, and while it has already gained some traction among healthcare facilities, it has not been an easy journey. For starters, none of his team members had any healthcare experience. And, it has been tough getting the healthcare services providers on board.
As the company signs up partners to pilot the programme, however, it has been able to collect data that will help improve its system. The start-up has partnered with Bagan Specialist Centre in Penang and an undisclosed nursing home in Malaysia.
It is also venturing into Japan, with a pilot at an assisted living facility, and is targeting healthcare facilities in Hong Kong.
“Our accuracy at the end of the [Bagan] pilot was 95%,” Lim says. “We managed to fine-tune the model. We have achieved 200 cases of getting nurses to come and help.”
Toh adds: “The first step was to validate whether this was the right solution for nursing homes and hospitals.” During the trials, the start-up sought to understand the nurses’ workflow as well as hospital administrators’ concerns, and how to transform the elderly patients’ experience. It also began to understand the differences in caregiving provided in hospitals and nursing homes. For instance, the firm aims to help prioritise the nurses’ visits to patients who may have health issues other than a fall.
“We are also working with an Australian retirement village: Their compound is huge and they do not have enough caregivers to walk around. It is a dementia facility and falls can easily happen at night without people noticing,” says Lim.
SmartPeep has provided the retirement village with a fall detection system that leverages the facility’s existing camera system. By plugging SmartPeep servers into the cameras, it converts them into an intelligent fall detection device. There is no need to upgrade the cameras.
To allay concerns about privacy, SmartPeep anonymises the cameras’ recordings; no one at SmartPeep actually views the footage. Patients who are participating in the pilots also have to consent to being monitored by the cameras and sign consent forms. In any case, this is already standard procedure at care facilities with cameras.
The time taken for SmartPeep to fine-tune its data models will depend on the healthcare institution. For example, the pilot in Bagan took three to four months, but a pilot in Singapore would take longer, owing to the internet separation policy at public hospitals. This means the team members can only spot and rectify mistakes each time they visit rather than monitor the system remotely over the Internet.
And while SmartPeep does not have a consumer version yet, it is Lim’s vision to help the elderly age gracefully at home. “We found out when we talked to nurses who have tested a lot of sensors that nothing’s really worked perfectly yet,” Lim says. “A lot of them are [too] sensitive and have false alarms, creating alert fatigue. We want to make it accurate and acceptable to everyone. That’s one of the challenges right now.”
The firm’s plans involve using video analytics to detect anomalies in people’s health, including analysing the way they walk or picking up irregular behaviour. “We still need to work with medical professionals to verify some of the items, whether they are helpful. Involvement from physiotherapists and geriatricians will be very important,” says Lim.
Meanwhile, he is cognisant that more time is needed before the business can really take off. “Once the camera hardware is ready and our tech is stable, that is when we will partner with a camera manufacturer to plug our software into a camera to scale it out to consumers.”
