AI Clinics on Mobile (AICOM)

Despite the progress achieved at the global level, many health systems are not sufficiently prepared to respond to populations' growing demands for healthcare services. Inequitable access to healthcare continues to impede progress toward achieving universal health coverage (UHC). According to UN, 30% of the world’s population lacks access to essential health services and over 1 billion people in the world today are forced to live in extreme poverty due to high health expenses. It is expected to see up to one-third of the world's population remain underserved by 2030, especially in disadvantaged population groups as well as least developed countries (LDCs). According to UN, only 17% of mothers and children in the poorest fifth of households in low- and lower-middle-income countries received at least six of seven basic maternal and child health interventions, compared to 74% for the wealthiest fifth of households.

As the first step to answering (UN) Sustainable Development Goal (SDG) 3, we have launched the AICOM project to develop health AI technologies on affordable mobile phones to improve healthcare access for the underserved and hard-to-reach population. 

We have completed phase one of AICOM and released AICOM-MP, which has achieved state-of-the-art results on monkeypox screening on affordable mobile phones. We chose to target the recent global monkeypox outbreak, a monkeypox is a highly contagious skin disease which causes rash-like lesions. Up to Feb 2nd 2023, there have been 85,449 global cases in 110 locations. Clinical diagnoses for Monkeypox can be conducted through a biopsy or a polymerase chain reaction (PCR) laboratory test. However, these tests are expensive, and LDCs may not be able to afford mass-scale PCR testing. 

Access to healthcare, especially primary care, is difficult in many countries. This problem further exacerbates in the current COVID-19 situation. Three challenges need to be overcome for us to address the healthcare access problem in the Western Pacific Region. The first challenge is access itself; people get denied access to healthcare services for various reasons, including lack of mobility, particularly for disabled people and the elderly, unpleasant user experiences, especially for countries with overloaded public healthcare systems, etc. The second challenge is equity. The reality is that healthcare resources are unevenly distributed, and this distribution is often strongly correlated with wealth distribution. The third challenge is efficiency; as the overall healthcare cost is climbing, people have high hopes for new technologies to improve healthcare access and quality while minimizing costs. 

Technological advancements have always been a driving force for progress in healthcare, we believe technology is the solution to improve healthcare access, equity, and efficiency.  We have already demonstrated health AI on affordable mobile phones is an effective way to improve healthcare access in the least developed countries. Particularly, we have observed a continuously rising rate of smartphone usage in the least developed countries in recent years, and the costs of smartphones have been dropping rapidly. Hence, providing health AI capabilities on affordable mobile devices can be a viable solution to reach the unreached.

In the first phase of the AICOM project, AICOM-MP combines mobile computing and health AI technologies to enable effective monkeypox screening on low-end mobile phones, and has achieved state-of-the-art screening results. In addition, we have generalized a methodology to expand AICOM-MP to cover other diseases, such as measles, chickenpox, eczema, etc.

More Information Regarding the Progress of Our Solution can be found below:

Technical Details: https://arxiv.org/abs/2211.14313

The United Nations ITU: https://aiforgood.itu.int/ai-clinics-on-mobile-aicom-universal-ai-doctors-for-the-underserved-and-hard-to-reach/

Bulletin of the World Health Organization: https://apps.who.int/iris/handle/10665/362183

World Economic Forum: https://www.weforum.org/agenda...

Note that We have open-sourced the dataset and source code of AICOM-MP to maximize coverage of the developed technologies.

AICOM aims to offer essential health technologies that’s safe, affordable, and effective, with a special emphasis on the poor, vulnerable and marginalized segments of the population, particularly people in the least developed countries (LDCs).

Among the list of LDCs provided by the United Nations, 33 countries are classified as least developed countries in Africa, they account for the largest proportion of the lDCs. Thus, to field-test our project’s feasibility in LDCs, we intend to target Africa as our primary client-based community for users to learn and experiment with AICOM. 

Following the development of AICOM-MP, we plan to extend our health AI software by developing additional AI-powered image processing functions to cover the most common infectious diseases with skin lesions in Africa.

AICOM prioritizes the development of medical-grade AI applications that are compatible with phones that are cheap, easy-to-use, and equipped with limited computing resources. According to online research, we summarized the top ten most used smartphones in Africa as the following: Tecno, Itel, Infinix, Samsung, Xiaomi, HUAWEI, OPPO, HMD Nokia, Apple, and REALME. Hence, we focus on running AICOM-MP APP on TECNO mobile devices which are usually installed with a MediaTek chip, to examine AICOM-MP's performance in such hardware environments, and validate its ability to operate under no or limited network connectivity. 

Dr. Shaoshan Liu's experience in fields of technology, entrepreneurship, and public policy enables him to take on great global challenges.  On technology, Dr. Liu has published 4 textbooks, 100+ research papers, and holds 150+ patents in artificial intelligence. On entrepreneurship, Dr. Liu has been the CEO of PerceptIn and has commercially deployed autonomous micro-mobility services worldwide. He is the Asia Chair of IEEE Entrepreneurship. On public policy, Dr. Liu has served on the World Economic Forum’s panel on Industry Response to Government Procurement Policy and is a member of the ACM U.S. Technology Policy Committee. He is an IEEE Senior Member, an IEEE Computer Society Distinguished Speaker, an ACM Distinguished Speaker, a member of MIT Technology Review’s Global Insights Panel, and a member of the Forbes Technology Council. Dr. Liu provides technical guidance and develops the roadmap for the AICOM project.

Ms. Ao Kong has 17 years of international management experience in global strategy, technology for peace and development, and multi-sector partnerships working for the United Nations, and in ESG investing and leadership role for the Tech industry. She recently became Senior Programme Advisor and Chief of Resource Mobilization and Strategic Communications at UN Technology Bank for Least Developed Countries. She serves on MIT Solve judge panel guiding tech-based social entrepreneurs to solve world challenges.  She is awarded as the “Pacific Delegate” by Carnegie Council for Ethics in International Relations. Ms. Kong develops visions for the AICOM project and provides essential insights into how AICOM can solve the global healthcare access problem.

Dr. An Na is a Doctor of Public Health student and Prajna Fellow at the Harvard T.H. Chan School of Public Health, she has worked in the field of humanitarian aid for more than a decade, and as part of the hospital management team for the International Committee of the Red Cross in Yemen and Afghanistan since 2019. She covered multiple positions while facing all the challenges in resources limited settings during COVID-19. Dr. Na brings in-depth medical expertise to the AICOM project and provides public health analysis to address real-world healthcare access problems through AICOM.

Prof. Xue Liu is a Chair Professor at McGill University. He is a VP of R&D, Chief Scientist, and Co-Director of the Samsung AI Center Montreal. He is a Fellow of the Canadian Academy of Engineering, an IEEE Fellow, and an ACM Distinguished Member.  He has been conducting innovative research and product development in both academia and industry ranging from early-stage startups to blue-chip companies. Prof. Liu provides technical guidance to the AICOM project.

Tianyi and Tianze Yang are senior undergraduate students at McGill University majoring in Statistics and minoring in CS. They have had 4 years of research experience conducted in 7 labs supervised by well-known professors. These experiences prepared them to address convoluted problems in human society from a mathematical and computational perspective. They are the main developers of the AICOM project.

Fist, AICOM team has developed and open-sourced a SOTA Monkeypox screening model hosted as a web service for public access, a comprehensive monkeypox dataset. We have compiled and submitted an academic article within a short period of time.  

Second, AICOM team has developed a mobile application AICOM-MP that is deployable on resource-constrained mobile devices. AICOM-MP is embedded with a SOTA monkeypox screening machine learning model which underwent a float-16 quantization process to avoid an over-reliance on computing environments. AICOM-MP is equipped with the energy-constrained quantization optimization (EQO) technique which adjusts the number of CPU threads used by the app according to the mobile device's battery level, charging state, and available memory.

Through a series of development in mobile computing optimization techniques, deep learning techniques, as well as our open-sourcing capability, AICOM has evolved from the concept of enabling global healthcare access to the prototypical development of deploying medical-grade machine learning models on low-end mobile devices.

The AICOM-MP solution has been downloaded over 1000 times and our website traffic has reached over 30,000 two months after its launch. Also, through the United Nations ITU blog, the bulletin of Worl Health Organization, and the World Economic Forum agenda, the solution has reached over millions of people. 

We have incubated the AICOM project at McGill University, and we have demonstrated the potential of AICOM through achieving state-of-the-art results on monkeypox screening. At this stage, we need to maximize the outreach of AICOM so that more people, particularly those living in LDCs, can enjoy the benefits empowered by AICOM. MIT Solve is the perfect platform to maximize AICOM's outreach globally. 

Although many health AI solutions have been proposed in the past decade, enabling health AI on affordable mobile phones is the true solution to reach the underserved and unreached. 

AICOM addresses the healthcare access problem by bringing healthcare services to the underserved and hard-to-reach patient's mobile devices and addresses the healthcare efficiency problem by enabling AI doctors to conduct rapid disease screenings. Along with the embedded energy-constrained quantization optimization (EQO) technique which adjusts the number of CPU threads used by the app according to the mobile device's battery level, charging state, and available memory, AICOM-MP undergoes a float-16 quantization process which makes it capable of conducting monkeypox screening without an over-reliance on computing environments. In addition, users with limited or no connectivity can still access our medical AI technologies on mobile. In short, our AICOM-MP app has demonstrated the fundamental possibility of deploying a medical-grade machine learning model to low-end mobile devices with minimal loss in performance. 

Through combining mobile computing optimization techniques and deep learning techniques, we have achieved state-of-the-art monkeypox screening results. We have open-sourced the dataset and source code so that people around the globe can easily access the technology, either directly use AICOM-MP, or integrate AICOM-MP into their products and services.

Short-term impact goal - maximize download and monthly active users of AICOM to get people into the habit of utilizing their mobile phones for healthcare access. Particularly, we need MIT Solve to help us on public outreach and to achieve this short-term goal.

Long-term impact goal - to enable public health benefits particularly for people in LDCs, for instance, health performance and life expectancy for those who utilize the AICOM application vs. those who do not use the AICOM
application.

As we are developing a health AI application on mobile, the key indicators are number of downloads and monthly active users, which are well-established metrics to measure the success of a mobile application. We have already
developed the technologies as well as the application, and we enter MIT Solve to maximize the outreach of the AICOM application.

Many people, even those in LDCs, have mobile phones, but many in LDCs do not have access to healthcare.  A change we aim for is to deliver healthcare access through mobile phones. The short-term change is to promote AICOM so that people in LDCs can get used to use healthcare services through their phones.  This can be measured by the number of installs in LDCs.  A long-term change can be measured through public health research techniques, which gauge whether AICOM has improved health conditions in LDCs.  We have public health expertise on our team to continuously monitor the progress and public health benefits brought by AICOM.

AICOM integrates deep learning methods, mobile computing optimizations, medical know-hows, and public health tracking techniques into one product, we are not aware of any other products achieving the same degree of integration. This integration is practical in solving real-world healthcare challenges. 

Our advantage has been demonstrated through AICOM-MP, the first prototype of AICOM. First, we constructed a robust dataset with wide coverage, diversification, and generalization. Our dataset underwent a rigorous data selection process. We removed several artificially generated monkeypox images. We added 18 healthy face and 18 healthy hand high-quality images from the 11k hands dataset and Flickr-Faces-HQ dataset respectively, and collected 22 newly-found monkeypox images of high quality from seven published articles. Thus, the medical knowledge embedded in AICOM-MP AI is ensured to be precise, objective, and professional. Furthermore, compared to existing AI-based monkeypox detectors, AICOM-MP has achieved state-of-the-art performance.  Furthermore, to allow health AI professionals around the globe to integrate AICOM-MP into their services, AICOM-MP’s source code and dataset have been open-sourced. Moreover, aiming to enable the deployment of medical-grade machine-learning models to low-end mobile devices, we resort to a Float-16 quantization process which reduces the model size to 24 Mega Byets, which is less than 4\% of the original model size, while ensuring a minimal loss in disease screening performance.

To maintain the stability of our app when operating under limited computing resources, our app is implemented with an energy-constrained quantization optimization (EQO) technique which adjusts the number of CPU threads used by the app according to the mobile device's battery level, charging state, and available memory (heapsize) at the time of diagnosis. EQO adjusts higher and lower compute units based on mobile devices' hardware environment and ultimately helps our app to deploy the finite computing resources more effectively.

More technical details regarding AICOM-MP can be accessed here: https://arxiv.org/abs/2211.143...

A demo of AICOM-MP can be accessed here: https://aicom-hc.web.app/

Diversity: AICOM welcomes partners, scientists and development agencies from every social and cultural background. 

Equity: Everyone in the team will be provided with the same opportunities, resources, and expectations to thrive. Policies and practices will be designed in a unbiased manner such that everyone is treated objectively without discrimination.

Inclusion: AICOM is dedicated to creating an inclusive and aspiring working environment in which everyone is respected and valued. To ensure long-lasting progress, AICOM makes sure that everyone's proposal, ideas, and opinions are heard and treated in a transparent and just manner.

AICOM has been incubated as an open-source project to achieve ubiquitous adoption of health AI technologies on mobile devices. In the initial phase, AICOM will rely on outside funding to sustain its growth. Meanwhile, we are developing proprietary, fee-based version of the application for users in more advanced economies to become self-sufficient and to enable organic growth.  

Technological advancements have always been a driving force for progress in healthcare, we believe technology is the solution to improve healthcare access, equity, and efficiency globally.  Particularly, we have observed a continuously rising rate of smartphone usage in LDCs in recent years, and the costs of smartphones have been dropping rapidly. Hence, providing health AI capabilities on affordable mobile devices have tremendous market potential, starting with empowering healthcare access for people in LDCs.

In the initial phase, AICOM will rely on outside funding to sustain the growth of AICOM. Meanwhile, we will develop proprietary, fee-based version of the application for users in more advanced economies to become self-sufficient and to enable organic growth.

Specifically, at the current stage, AICOM operates based on available funding and mainly focus on non-profit activities, the support mainly comes from McGill University. AICOM will work on gathering reputations among the healthcare marketplace and creating a credible image among the user communities through public speeches on globally recognized platforms, such as UN, and research publications. 

In the second stage, AICOM will develop a proprietary version of its AI doctor to serve customers with more specialized medical needs. It will incorporate a new subscription business model that offers high-end healthcare services targeted toward more specific diseases for people in developed countries. 

This project has been incubated in Prof. Xue Liu's research group at McGill University and is supported by UN Technology Bank. With limited resource, we have already proven the feasibility of our project and its potential. Specifically, we have developed and open-sourced a SOTA Monkeypox screening model hosted as a web service for public access, a comprehensive monkeypox dataset. We have compiled and submitted an academic article within a short period of time. As shown in the pitching video, we have developed a workable demo app that is equipped with the EQO technology and the quantized Monkeypox screening model. It is deployable on resource-constrained mobile devices and is capable of conducting medical inference while avoiding an over-reliance on computing resources.