RobO2: A High Performance Intelligent Ventilator with Distributed Data

Scalable, intelligent ventilator that utilizes distributed learning to deal with fast escalating situations where resource and skills are not available.

Mark Gilligan is the CEO of Blacksheep Group, and a business and engineering leader with 30 years of experience in development and manufacture of scientific and medical products. 

Scalability: During COVID-19, the demand for ventilators increased significantly world-wide. Major ventilator manufacturers increased production by approximately 50%, however, supply chain issues strongly affected the production as existing high-cost ventilators have specially-manufactured components and they could not be made quickly.

Affordability: Current medical ventilators are high-cost devices. This leads to the major problem of unavailability of ventilators in low- and middle- income countries (LMICs), where they face limitations in local infrastructure, personnel expertise, and budget constraints.

Reliability: During the COVID-19 pandemic, many low-cost ventilators were developed. However, these ventilators were designed quickly and did not meet the medical requirements to enable their use in ICUs.

Medical data: it is still not known how best to treat patients with COVID-19 and respiratory failure despite millions of people requiring respiratory support over the last year. Sharing and learning from ventilator data would allow respiratory support of patients to be optimized and patient outcomes to be improved.

Privacy: RobO2 requires the Ventilator’s data, as well as other medical data of patients to be collected to a centralized server, to build machine learning models. This raises privacy concerns about the data during the sharing process which this project addresses from the start. 

Patients: The ultimate goal of RoBO2 is to protect patients’ lungs from any damage when being ventilated. Controlling breath by breath will ensure the ventilation is adapted every single breath to improve the outcomes. Additionally, our distributed learning mechanism will allow the data to be used effectively and privately, hence providing meaningful suggestions about the patient conditions based on enormous library of healthcare data.

Doctors and nurses: The ventilator can be used easily with intelligent mode. This will save time for doctors and nurses overloaded hospitals during a pandemic as well as in LMICs, where there are a lack of well-trained doctors and even facilities to support ICU ventilators. The machine’s mechanism also allows nurses to service and maintain it without the help of technicians or special tools.

Scientists and policy makers: Although we do not collect any data from the ventilator, scientists and policy makers will still benefit from our distributed learning approach, as we can utilize a large amount of medical data in a decentralized manner. Our learning approach will allow the science community to make better decisions without worry about the privacy concerns when sharing the medical data. 

The next pandemic is likely to be a respiratory viral infection (including influenza) and therefor this will allow rapid learning of this and ways to manage patients with resource constraints. However in the mean time the advantage of our project is that it can be used in LMICs and benefit patients with ‘normal’ causes of respiratory failure. There is still huge uncertainty how to best treat these patients and carrying out clinical trials is expensive and usually unfeasible in LMICs.

Knowledge sharing is also an important goal of the solution. The ventilator provides a platform to extract clinical data for further medical research, which helps advance the knowledge in the field. A Joint PhD programme is being created between academia and industry. This is a new education model, and through the programme, knowledge from the project will be shared via publishing peer reviewed papers on data collection/ machine learning for ventilators. To scale up the impact, the learning framework will be made open-source and available for doctor, nurses, researchers and policy makers. Additionally, training workshops and materials for doctors will be conducted (online and offline) to change their practice of using the ventilators, in cooperation with hospitals and medical schools.

There are different stakeholders in the successful outcomes of the RobO2 project:

Patients: the prognosis for patients in LMICs with respiratory problems requiring ventilation is poor and during a Pandemic well resourced countries trend this way. The fact that RobO2 is cost effective and able to operate autonomously means that patients will not only have a higher chance of overcoming acute respiratory problems, but also lead a healthier life afterwards.

Doctors/Nurses: Since RobO2 is an easy-to-use ventilator, medical staff can be trained quickly. During a pandemic when hospitals are overloaded, is thus reduced for doctors and nurses and allows them to utilize man effort for other critical tasks. The data history will also allow doctors to have a real-time update of the patient condition and changes overtime.

Researchers: Data collected from RobO2 ventilator will be a publicly available resource of ventilation data allowing researchers to build models of how to treat patients.

Health Agencies: For a given limited budget in the order of 5-10x more patients will be able to be treated with RobO2 than existing ventilation solutions. Also the future improved health of patient’s lungs after overcoming disease reduces ongoing burden on healthcare services and save future costs.

In the first year RobO2 will create the base platform to enable initial clinical trials to take place on a small number of patients.

In the second year, RobO2 will be deployed across at least 3 countries (Vietnam, Laos and the UK) to start to gain the underlying data to allow the autonomous ventilation of patients.

By the end of the 3 year project, RobO2 ventilators will have been rolled out in volumes of thousands and have collected a large amount of data to begin to make significant medical insights.

Outcomes for the usability component of the project are qualitative and quantitative. Focus groups and interview data will be transcribed and translated and analysed qualitatively. Detailed field notes will be kept for observation periods.

Usability feedback summaries will be passed to the Blacksheep team through multidisciplinary meetings. Additionally, we aim to use all data obtained to explore formal and informal clinical decision-making roles, responsibilities and processes encompassed in ICU ventilation care. We will use frameworks such as the Non-adoption, Abandonment, Scale-Up, Spread, and Sustainability (NASSS) Framework to understand behaviour and requirements specifically around ventilator use. Details of the outcome measurements for the non-human participant parts of this project are contained in the main proposal.

As at the proof of concept stage, the focus now is on the hardware and firmware development of the machine. Many tests have been run and the collected data has been processed through R data analysis tools to understand the machine’s performance, the shape of the data, as well as areas for improvement.

The primary significant challenge is funding for the project. We can raise capital from financially motivated investors. However, this will come with financial pressures that could derail the primary humanitarian objectives of the solution.

The RobO2 project requires financing to ensure that all the required team members are able to focus on the project and deliver the required result. The cost of developing the solution in Vietnam is significantly lower than doing so in the UK (perhaps ¼ of the cost) or USA (perhaps 1/10 of the cost). These costs mean that the product can be developed and launched and create a financially sustainable platform even during a non-pandemic period while in the USA or UK, it is difficult to make a platform that could be financially viable during a non-pandemic period and yet still be able to be cost effective during a pandemic.

The legal challenges to be faced will be those for any medical device certification, but beyond this, there are issues around Patient Data confidentiality and the conflicts around this and a global data centric platform. This is usually able to be overcome with varying levels in different countries based on anonymization of data.

Oxford University Clinical Research Unit

Tropical Disease Hospital in Vietnam 

Blacktrace Holdings Ltd. in the UK

Imperial College London

Blacksheep Company Vietnam 

We seek support from Trinity Challenge because we share the mission of the fund to protect people worldwide, which is to fight against the devastating pandemic using the best of our ability. The support of Trinity Challenge could help overcome the financial and technical barrier, and ensure the project to be delivered on schedule.

First, the fund provided in the Trinity Challenge provides stable financial resources for us to deliver the product. This allows us to recruit and collaborate with more talents for the project, which can speed up the development process. Furthermore, there is no intention to raise external investor funds because this leads to an extremely different focus. Funding through the Trinity Challenge allows the key components of this project to be realized: enabling equitable access to safe mechanical ventilation in LMICs, building local capacity and ultimately improving outcome from critical illness. Our financial viability analysis demonstrates that external investors motivations to increase profits are likely to result in high product prices.

Secondly, the wider network and collaboration from different sectors and various stakeholders will enable teams to access a bigger pool of knowledge. As a result, we can create the bigger global impact using advanced technology. 

Team have been completed and will seek further collaboration