Up! - Intelligent Fall Prevention Companion

          One-third of the elderly population fall each year, resulting in over 600,000 deaths, the second leading cause of accidental deaths worldwide (WHO).  This places a significant financial burden on the healthcare system and the patients' family members.  However, existing solutions are either passive protectors, which are inconvenient and uncomfortable to use, or restraining devices that severely limit the mobility of patients. 

          Up!'s solution is a robot-pet with a depth camera that automatically follows the patient at home, monitoring their posture and gait.  It tracks instability events with the help of skeletal tracking software and alerts caretakers in the case of severe events.  After identifying the issue, the robot-pet also guides patients to do exercises targeting their weaknesses.  Up!'s friendly approach to fall prevention gives their family members peace of mind without taking away from mobility and independence from the patients. 

         As people lead longer lives, an unprecedented number of elderly people live in this world, with new estimates placing the number at over 650 million.  Over a third of the elderly population fall each year – a massive number for an aging world.  The elderly population is seriously affected in the event of falls, often suffering from hip fractures. Each year, 600,000 elderly deaths are caused by falls, making it the second leading cause of accidental deaths.

          On an individual level, fall victims suffer from reduced mobility and independence due to sustained injuries.  Traumatic memories resulting from falls further induce stress in the victims, which may discourage activity and outdoor social interactions.  

          On a systematic level, this places a huge burden on the already saturated healthcare systems, evident in the 1,000 USD that each fall costs the system (NSW Department of Health).  With over a third of the global elderly population falling each year, the financial burden it places on society is not to be overlooked.

Figure 1 (Data collected from 100 elderly aged over 65 in Hong Kong, reflecting the prevalence of falling amongst the elderly population)

         In a bustling metropolis such as Hong Kong, the younger generation can seldom afford the time to tend to their parents' needs, while the elderly themselves may not wish to be placed in elderly homes.  Therefore, over 150,000 elderly live alone in Hong Kong, a number that is on the rise (Hong Kong 2016 Population By-census). Many urban communities around the world experiencing aging population are faced with similar elderly communities that lack immediate support from caretakers.  This elderly population is particularly vulnerable to falls because their inability to access healthcare professionals in time leads to a delay in the treatment of such events.  

          Up!'s solution tackles the issue of the inadequate human support by automating the caretaking process.  The robot-pet enables healthcare practitioners and family members to remotely monitor the elderly who live alone.  Our product helps the elderly access immediate medical attention, motivates them to exercise, and prevents falls without limiting their mobility.  Best of all, the solution is presented as a friendly companion who interacts with the elderly at home, keeping the elderly company.  We believe that this is what the elderly need.

         Current solutions that aim to prevent elderly falls already exist in the market, but most adopt one of three approaches: fall detection, passive protection, or risk assessment.  Although fall detection devices reduce the delay in their treatment, the damage of the fall has already been inflicted on the elderly.  Passive protectors, such as hip padding, may reduce bodily injury from such events but are often uncomfortable and inconvenient to use.  On the other hand, conventional risk assessment approaches such as medical surveys require the involvement of physicians, which the elderly may not visit often, leading to outdated data.  

          Up!'s solution – an intelligent fall prevention companion – seeks to revolutionize fall prevention in elderly communities through early intervention techniques.  Our easy and fun-to-use product continuously monitors the patient's stability and offers prescriptive mobility training.  The product itself is a robot-pet equipped with a depth camera that automatically follows the patient at home with the help of LiDAR.  Using data collected by the depth camera, it applies AI-powered skeletal tracking algorithms to instantaneously assess the frequency and nature of unstable motions.  Only the skeletal body model is saved into our system, ensuring the privacy of the patient.  Family members are updated on the patient's status through an app, while proactive calls are triggered to emergency services for severe events.  It then guides patients to do exercises by demonstrating movements specifically catered to their weaknesses.  This preemptive strategy to fall prevention aids the patient in strengthening their weakest muscles, thus reducing the probability of falling due to instability.  

          The product is accompanied by an app that allows family members and physicians to remotely access the data collected by the robot-pet.  Families are alerted on the location and time of falls, who may address potential risks by making changes to the elderly’s home.  On the other hand, skeletal tracking data and posture analysis of the patient is available to the physician through the app, who may use the data to provide pinpoint diagnoses and customized treatment.

          At this stage, we have built a prototype of the robot-pet using the TurtleBot3 robot kit that can autonomously follow the patient and record skeletal tracking data using the RealSense depth camera equipped on it, powered by the Nuitrack software.  It can also detect falls instantaneously through an algorithm we devised.

         Up!'s solution is based on viable technology that currently exists in the market. What makes it innovative is the unique application of these technologies and how we combine them to create a working solution that solves a major health care problem.

          Robot pets such as Pyro and Sony's Aibo are in the market, though their main purpose is to provide company and entertainment. Moreover, they are very expensive. We are one of the first to use them directly in the healthcare industry.

          Skeletal tracking/ depth cameras are mainly used for gaming (Xbox Kinetic 360) and occasionally in sport teams to analyse athletes’ posture. However, they are seldom used in the healthcare community.

          One known problem to applying skeletal analysis to track the elderly at home is that the depth camera is often fixed in one place, limiting the field and angle of vision.  To make this work, the camera must be able to follow the patient around the home.  We believe a robot-pet is the perfect conduit to carry the camera and take continuous pictures at the correct angle necessary for effective skeletal tracking analysis.  Presenting the solution as a friendly robot-pet further improves the chances that the elderly will be receptive of this technology.

Our solution utilizes three core technologies:

Skeletal tracking with Intel RealSense depth camera and Nuitrack software

• Monitor gait and posture of the elderly
• Inputs skeletal data into algorithm 

Robot-pet (prototype built with Turtlebot 3 “Burger” model)

• Provides company and entertainment for the elderly
• Autonomously follows user to monitor user at a constant, optimal angle
• Shows user-specific exercises via movements (future feature)

Algorithm

• Compares skeletal data with database to identify weaknesses and abnormal motions
• Determines corrective exercises based on identified weaknesses

         Many elderly people are resistant to technology (Figure 1).  This explains why elderly security cameras are unpopular among the elderly community. We overcome this problem by presenting a monitoring platform in the form of a robot-pet that offers a personal touch to help raise the elderly’s receptiveness towards technology.  It allows us to place the camera near them all the time to analyse their posture from a constant, optimal angle. Having the robot-pet perform exercises with the elderly also makes it seem like they are playing with the pet, reducing their reluctance to exercise. This interaction allows us to accurately identify weaknesses in the gait of the elderly, increasing the effectiveness of the prescriptive exercise and allowing physicians to provide pinpoint treatment.

          Furthermore, the elderly needs constant monitoring.  In cases where the elderly does not have relatives or friends to take care of them at home, their only alternatives are either hire professional senior caregivers or simply have no one take care of them.  However, hiring professional senior caregivers tend to be costly, and is therefore not a viable option for most seniors.  Instead, most opt for the latter option, which leaves seniors vulnerable to falls.  Our robot-pet is a more economical and enticing alternative that gets the job done.

Figure 1 (Data collected from 100 people aged over 65 in Hong Kong)

Our product is still in the prototyping stage; therefore, we are not yet serving any customers.

We intend to conduct pilot tests with 100 elderly participants from Hong Kong by June 2020, working with local universities and physical therapy clinics.

We intend to serve over 10,000 elderly in Hong Kong and the Greater Bay Area by June 2024.

Jul - Dec 2019:

Intense R&D primarily focused on hardware prototyping (robot-pet), AI algorithm (skeletal analysis and exercise demonstrations) and mobile app for family members and physicians.  Ongoing product testing through collaboration with local universities, clinics and elderly centers.

Jan - Jun 2020:

Continue to refine product design based on feedback from collaborators.

Jul - Dec 2020:

Reach out to potential distributors and manufacturers to establish supply chain and distribution channel.  Launch product in Hong Kong in December 2020.

Jan - Dec 2021

Continue to refine product based on feedback from customers.  Launch second version of product in Greater Bay Area in December 2021.

2022 and beyond:

Continue to refine product to eventually launch globally.

Financial Barrier

Finding adequate funding will be a challenge that Up! will face in the years to come.  Prototyping both the hardware and software will require financial resources from investors and prize money.

Technical Barrier

Creating our product will require a wide range of expertise, ranging from hardware designing, programming and medical knowledge.  Our team will need to collaborate with robotics engineers, AI experts and medical practitioners to further advance our product.  Furthermore, having the robot-pet mimic actions in a lifelike way will require expertise in mechatronics and physiotherapy. 

Market Barrier

Finding the right go-to-market collaborators to test and distribute our solution will be a challenge, as we need to overcome stringent certification requirements and procurement processes for medical devices.

Financial Barrier

We plan to actively pursue potential investors by showcasing our product through events and trade fairs and apply for financial support through incubation programs.

Technical Barrier

We will begin working with universities and medical clinics to tap into like-minded experts interested in solving this problem.  Once we receive adequate funding, we will begin recruiting the expertise required.

Market Barrier

We will develop connections through active networking at events such as SOLVE.  From there, we will reach out to potential collaborators and form partnerships with stakeholders in this field.

N/A

Our solution team currently consists of 3 core-members and 2 part-time staff.

We will expand our team to include hardware engineers, software developers, data scientists and medical practitioners in the coming year.

Bryan Ho is a sophomore at Northfield Mount Hermon School.  He also studies data analytics and Python programming at Coursera.  His passion in 3D CAD and software engineering originated from his invaluable experiences at CIS MIT STEAM Camp and MIT 2019 Hong Kong HealthHack.  Bryan takes a leadership role at Up! and oversees hardware and software development.

Frank Ho is founder of AutoCognita, an EdTech business that develops mobile literacy education apps (Global Learning XPRIZE semi-finalist / Adult Literacy XPRIZE finalist).  He also founded Contrendian, a FinTech business that revolutionizes how people invest.  Frank advises the team on product direction, design, and business model.  He is a graduate of MIT and INSEAD.

Timothy Ho is a 10th Grader at Po Leung Kuk Choi Kai Yau School. He gained experience in CAD, CAM and coding from the Engineering Course of the Dual Program (organized by HKUST) and the MATE Underwater ROV Competition. Timothy leads the mechanical design and development at Up! and assists the team in hardware and software development.

Falcone Tsang is a physiotherapist focusing on sports, geriatric and orthopaedic rehabilitation.  His expertise is integrating exercise with manual/electrotherapy.

Sam Law is a prosthetist orthotist at Queen Elizabeth Hospital in Hong Kong.  His expertise is applying orthoses for postural stability and fall prevention.

Our team and solution has earned initial validation by winning the Grand Prize of the inaugural MIT HealthHACK Hong Kong competition.  

We have been invited by, and are currently in discussion with, Hong Kong Science & Technology Park (HKSTP) and Cyberport to join their initiatives supporting start-up organizations.  We are also in discussion with the Hong Kong University of Science & Technology (HKUST) for technical collaboration on gait analysis.

          As we are developing our product, we will work with medical practitioners to further refine our product. We will also allow them to use our product for free in the pilot stage. Afterwards, they can help us promote our products to the elderly and their families.

          Our target is to price our robot-pet at US$5000. Each minor fall costs over US$1000 in medical fees, while major falls cost over US$10000. As a third of the elderly population suffer from at least a fall each year, our product can save the elderly over US$3000 in 9 years. As the human lifespan continues to increase, we project that our product will be able to reduce medical spending of our customers by much more. 

          As a comparison, wheelchairs, a popular solution to prevent falls, cost only US$2000, yet confine the elderly to the chair and vastly restrict their movement.

          We plan to sell our product to consumers in Hong Kong at the early stages to gain recognition in the healthcare community. We will then expand our market to China and the US, before launching our product worldwide. We estimate that we can penetrate 0.01% of the elderly market within 5 years.

          First, we plan to obtain funding through grants to develop our product into an attractive option for consumers. This is achieved by applying to incubation programs, such as the Innovation and Technology Fund (ITF) of Hong Kong, the Incu-Bio and Incu-Tech programs of the Hong Kong Science and Technology Park, and the Cyberport Incubation program. Additionally, we will participate in competitions such as SOLVE and invest the prize money into UP!.

          Additionally, we plan to raise seed capital from angel investors and venture capitalists. Thus, we will promote our product in conventions to better reach out to potential investors and establish networks with future collaborators.  

          In the long term, we will achieve profit through our various revenue streams: sales of robot-pets, leasing robot-pets to elderly care centers, subscription to our application, and subscription to our skeletal analysis database for physicians and medical practitioners.  

We believe that applying to SOLVE will help us address each of the challenges our team currently faces.  We may obtain funding to further develop our solution and build connections with like-minded experts with technical skills and market access.  Our goal is to expand our idea into a viable business and serve the elderly community.

N/A

We would like to partner with the following organizations to accelerate development, data collection/analysis and clinical trials.

United States:

- National Council on Aging - Fall Prevention Resource Center

- National Institute on Aging

- AgeLab (MIT)

- Fall Prevention Center of Excellence (USC)

Hong Kong:

- Department of Health - Elderly Health Service

- Hospital Authority of Hong Kong

- Hong Kong Society for Rehabilitation

We are currently developing AI algorithms to map skeletal tracking data to known types of instabilities.  Our plan is to start with easily observable patterns that have a direct mapping based on spatial and temporal measures, and gradually advance to more subtle movement patterns where a chain of events could be used to assess fall risk in real-time.  This would require a rather large amount of labelled data with known cases of instabilities, and training the deep neural network to identify patterns.  Our clinical development effort will focus on acquiring and analyzing such data.

We will utilize the prize to build out the data collection, labeling, and training of the deep neural network to ultimately create a system that matches or exceeds a human medical professional's ability to evaluate fall risk.

N/A

N/A

N/A

We are currently developing AI algorithms to map skeletal tracking data to known types of instabilities.  Our plan is to start with easily observable patterns that have a direct mapping based on spatial and temporal measures, and gradually advance to more subtle movement patterns where a chain of events could be used to assess fall risk in real-time.  This would require a rather large amount of labelled data with known cases of instabilities, and training the deep neural network to identify patterns.  Our clinical development effort will focus on acquiring and analyzing such data.

We will utilize the prize money to build out the data collection, labeling, and training of the deep neural network to ultimately create a system that matches or exceeds a human medical professional's ability to evaluate fall risk.

N/A