Kawasak.AI

COVID-19 has unveiled the problem of disease early detection and diagnosis worldwide. Efficient and cost-effective early diagnosis systems, particularly among the rising incidence of Kawasaki disease (KD) and outbreaks of Multisystem Inflammatory Syndrome in Children (MIS-C) in COVID-19 hotspots, may help decrease adverse health outcomes for children.

Although an uncommon pediatric disease, KD is the leading cause of acquired heart disease in children globally. Recent links during the pandemic may reveal SARS-CoV-2 as a potential trigger of KD. Despite the importance of timely treatment, it is commonly misdiagnosed, leading to significantly higher risk of coronary artery aneurysms.

Kawasak.AI is a robust, cost effective, and versatile artificial intelligence-based image diagnosis framework to aid doctors in the differential diagnosis of the pediatric vasculitis with links to COVID-19. Scaled globally, it has the potential to assist doctors and ERs in underserved regions for accurate and robust early diagnosis.

Kawasaki disease (KD) is the leading cause of acquired heart disease and coronary artery aneurysms in children globally, with estimates of over 6,000 cases annually. Within the last 15 years, its incidence rate has increased by approximately 5 times, and especially during COVID-19, an outbreak of the Multisystem Inflammatory Syndrome in Children (MIS-C) resembling clinical features of KD has targeted children in COVID-19 hotspots. Since mid-May of 2020, the CDC has reported over 4,000 cases of MIS-C in the US. SARS-CoV-2 may even be one of the triggers of KD, although its etiology remains unknown, revealing the importance of KD and MIS-C diagnosis in combating the global pandemic.

KD early diagnosis is crucial for effective intravenous immunoglobulin treatment within a rapid time frame of ten days, significantly reducing risk of coronary artery aneurysms from 25% to <5%. However, KD is often misdiagnosed as it has no specific diagnostic test and shares clinical findings with look-alike diseases, leading to risk of myocardial infarction or death of a child. Thus, the differential diagnosis of KD, as well as MIS-C, from look-alike diseases becomes an increasingly crucial problem for timely diagnosis and decreasing its morbidity and mortality.

Kawasak.AI is a robust, cost effective, and versatile artificial intelligence-based image diagnosis framework to aid doctors in the differential diagnosis of a rare heart disease with links to COVID-19. A deep learning ensemble with joint detection of multiple clinical criteria performs image analysis on photos that can be taken on a smartphone or a cheap computing device. Initial prototypes on a Raspberry Pi with live image capture demonstrates capability of Kawasak.AI to run on a $50 device while producing high accuracy results in <2 seconds.

When applied to early diagnose Kawasaki disease, the Kawasak.AI model resulted in an average accuracy of over 82% and Area under the ROC Curve of 0.90 (where 1 is a perfect classifier), despite challenges of distinguishing Kawasaki disease from extreme look-alike diseases and limited data for the commonly misdiagnosed and rare heart disease. Results demonstrate clear discriminatory capabilities above all metric thresholds recommended by a Kawasaki disease expert, and this framework can be potentially extended to MIS-C diagnosis for COVID-19 and other diseases relying on visual clinical diagnosis.

The target population is two-fold: Kawasaki disease and MIS-C patients and parents of children who face risk of misdiagnosis and acquired heart disease, and doctors and hospitals to aid in diagnosis especially in underserved or disadvantaged areas.

Over 400 million individuals globally lack access to essential health services; given the importance of timely treatment for Kawasaki disease and its related MIS-C during COVID-19, Kawasak.AI can provide a cheap and versatile method to help prevent missed diagnosis in areas where a pediatrician may not be readily available. In underserved communities who face the greatest difficulties from COVID-19 and MIS-C outbreaks, the adaptable framework can be loaded onto existing infrastructure in communities or through a $50 Raspberry Pi implementation for live image analysis of clinical criteria.

I am extremely grateful to be working with the UCSD Kawasaki Disease Research Center and Rady Children’s Hospital, one of the leading centers for Kawasaki disease research, as well as the international KD Foundation, in discussing potential implementations of Kawasak.AI in hospitals or emergency rooms and assisting underserved populations.

The problem of misdiagnosis of KD and its related MIS-C emergence during COVID-19 identifies a weakness in current health systems in efficiently responding to onsets of diseases for timely treatment, especially in remote and underserved regions which have been disproportionately affected. An efficient and robust system for differential diagnosis can help decrease deaths during pandemics by catching patients early in disease progression, assisting communities in responding to health threats. Kawasak.AI addresses health inequities across communities through democratizing an accessible and affordable tool for early detection and diagnosis, assisting healthcare providers with diagnosing emerging diseases during a global pandemic.

Kawasak.AI has been tested off of crowdsourced data in collaboration with the international KD Foundation for 400+ patient images with consent, representing over 14 countries across the globe. Prototypical testing among a greater population is ongoing and forthcoming for validation larger patient datasets and across hospitals in the United States and Mexico. I am grateful to be working with the UCSD Kawasaki Disease Research Center and Rady Children’s Hospital in discussing potential implementations of Kawasak.AI in hospitals and ERs.

Kawasak.AI introduces new innovations in two dimensions: 1) medical application, in unlocking doors for efficient, reliable, and robust predictions for KD, a disease with relation to COVID-19 outbreaks for which a visual diagnostic tool has not been previously developed, and 2) technology, in developing a method for deep learning to solve challenging problems lacking large amounts of data.

Kawasak.AI is the first visual diagnosis tool for early diagnosis of KD and utilizes the first image database for KD created from scratch through crowdsourcing data from hundreds of KD parents with consent. The innovation arises in combining novel computer vision and AI techniques to overcome the bottleneck of data for deep learning in medical diagnosis. Contrary to using complex and often expensive medical scans, Kawasak.AI uses only smartphone images for efficient yet accessible diagnosis powered by deep learning and computer vision.

The solution presents a big improvement for the current diagnosis of KD. As a relatively untouched and overlooked area in medical innovations during the pandemic, KD and MIS-C has affected thousands of children worldwide, particularly in underserved areas which have been hit hardest by COVID-19. Currently, KD has no specific diagnostic test and relies on doctors carefully ruling out look-alike diseases, and as a result is commonly misdiagnosed. Kawasak.AI presents a solution to help assist doctors in timely diagnosis by distinguishing KD from extreme look-alike diseases. To the author’s knowledge, it is the only visual diagnosis tool and image-based diagnosis system for KD.

Currently, Kawasak.AI is in the prototype stage of development with 108 parents and patients as well as collaboration with 1 KD research center and hospital. In the next year, Kawasak.AI will continue to rapidly iterate during additional validation with external hospital and patient data to ensure safety and robustness of predictions, and conduct diagnostic trials to move into the pilot phase. The scalable smartphone app or on a Raspberry Pi computing device has the potential to serve a wide market of patients and doctors during pilot and growth.

Within 5 years, Kawasak.AI will continue to grow while continuously performing R&D to adapt to new findings and make improvements. A large customer base of patients, doctors, and hospitals will utilize Kawasak.AI to better diagnose KD and MIS-C patients, along with other difficult diagnose diseases incorporated into the holistic approach. The solution is envisioned to easily integrate into current workflows for hospitals as well as allow patient-facing software to identify potential risk in its earliest stages. Through collaboration with the KD Foundation and other international organizations, Kawasak.AI’s goal is to expand internationally and benefit communities across diverse demographics and socioeconomic backgrounds.

Progress and impact will be measured through:

• Total software downloads and demographics

• Number of patients diagnosed and timing/outcome (i.e. early diagnosis, late diagnosis; early treatment; recovered without complications, coronary artery aneurysms, etc.)

• Number of hospitals using Kawasak.AI technology and number of individual doctors employing the technology per hospital

• Change in number of diagnosed treated patients in communities reported before and after Kawasak.AI implementation and use

• Impact on the number of MIS-C and KD cases in COVID-19 hotpots and underserved communities

• Average time of diagnosis and hospitalization/treatment compared to traditional diagnosis

• Feedback from patients and doctors (through conducting interviews, feedback through app, etc.)

The solution aligns with the UN Sustainable Development Goals of 3) Health & Wellbeing through accelerating innovations for child health and heart disease, 10) Reduced Inequalities through helping one of the most vulnerable groups of children facing MIS-C and KD, and 11) Sustainable Cities and Communities through democratizing healthcare diagnosis access and assisting communities in COVID-19 recovery.

I deeply value the incredible network of leaders across MIT and numerous fields of expertise dedicated towards pioneering innovation, as well as mentorship opportunities to learn from experts. Specifically, the world class mentors and network that MIT Solve provides would foster connections through opportunities for partners and advisors for the growth of Kawasak.AI, particularly in expanding internationally and in underserved regions. Gaining exposure to many incredible peers and Solver teams unlocks many doors for accelerated growth and different perspectives on world challenges, which I hope to become a part of.

Kawasak.AI wishes to engage with partners and mentors across numerous disciplines of business, law and regulation, public relations, and product and service distribution to gain exposure to different perspective and receive advice for accelerated growth and improvement.

Kawasak.AI utilizes easily accessible and equitable image-based diagnostics powered by deep learning to bridge gaps in early diagnosis. Through developing a solution to aid doctors in the differential diagnosis of the pediatric vasculitis with links to COVID-19, Kawasak.AI directly addresses a prominent and rising problem adversely affecting children in COVID-19 hotspots. Kawasak.AI is the first visual diagnosis tool for Kawasaki Disease and Multisystem Inflammatory Syndrome in Children, the #1 cause of acquired heart disease in children especially during the pandemic, and has the potential to assist doctors and ERs in underserved regions for accurate and robust early diagnosis.

Kawasak.AI will use the Robert Wood Johnson Foundation Prize to increase outreach and efforts to assist underserved areas which have been disproportionately affected by COVID-19, helping prevent misdiagnosis and thus decrease morbidity and mortality of COVID-19 associated pediatric diseases.

Kawasak.AI utilizes easily accessible and equitable image-based diagnostics powered by deep learning to bridge gaps in early diagnosis. Through developing a solution to aid doctors in the differential diagnosis of the pediatric vasculitis with links to COVID-19, Kawasak.AI directly addresses a prominent and rising problem adversely affecting children in COVID-19 hotspots. Kawasak.AI is the first visual diagnosis tool for Kawasaki Disease and Multisystem Inflammatory Syndrome in Children, the #1 cause of acquired heart disease in children especially during the pandemic, and has the potential to assist doctors and ERs in underserved regions for accurate and robust early diagnosis.

Kawasak.AI will use the Andan Prize for Innovation to increase outreach and efforts to assist underserved areas which have been disproportionately affected by COVID-19, helping prevent misdiagnosis and thus decrease morbidity and mortality of COVID-19 associated pediatric diseases.

Kawasak.AI utilizes easily accessible and equitable image-based diagnostics powered by deep learning to help prevent acquired heart disease in COVID-19's most vulnerable populations — children. Through developing a solution to aid doctors in the differential diagnosis of the pediatric vasculitis with links to COVID-19, Kawasak.AI directly addresses a prominent and rising problem adversely affecting children in COVID-19 hotspots. Kawasak.AI is the first visual diagnosis tool for Kawasaki Disease and Multisystem Inflammatory Syndrome in Children, the #1 cause of acquired heart disease in children especially during the pandemic, and has the potential to assist doctors and ERs in underserved regions for accurate and robust early diagnosis.

My younger sister was first misdiagnosed when she had Kawasaki disease at 3 years old, motivating me to address the recent spike in Kawasaki disease-like cases of MIS-C during COVID-19. Kawasak.AI can help save lives of children and girls like my younger sister who risk permanent heart damage when misdiagnosed.

Kawasak.AI will use the Innovation for Women Prize to increase outreach and efforts to assist underserved areas which have been disproportionately affected by COVID-19, helping prevent misdiagnosis and thus decrease morbidity and mortality of COVID-19 associated pediatric diseases.

Kawasak.AI leverages deep learning and computer vision for efficient image-based early diagnosis of a potentially fatal pediatric vasculitis to aid patients and doctors during COVID-19. Through developing Artificial Intelligence-based solutions to tackle problems in healthcare and diagnosis, Kawasak.AI combines technological innovations in overcoming the bottleneck of data in deep learning and real-world impact with benefiting humanity.

Kawasak.AI will use the AI for Humanity Prize to invest in processes for rigorous R&D and development of equitable and robust deep learning technologies. Through high quality clinical testing and outreach towards underserved regions, Kawasak.AI has the potential to greatly improve conditions for thousands of children facing acquired heart disease and misdiagnosis especially in areas disproportionately affected by COVID-19.

Kawasak.AI utilizes easily accessible and equitable image-based diagnostics powered by deep learning to bridge gaps in early diagnosis. Through developing a solution to aid doctors in the differential diagnosis of the pediatric vasculitis with links to COVID-19, Kawasak.AI directly addresses a prominent and rising problem adversely affecting children in COVID-19 hotspots. Kawasak.AI is the first visual diagnosis tool for Kawasaki Disease and Multisystem Inflammatory Syndrome in Children, the #1 cause of acquired heart disease in children especially during the pandemic, and has the potential to assist doctors and ERs in underserved regions for accurate and robust early diagnosis.

Kawasak.AI will use the Global Fund Prize to increase outreach and efforts to assist underserved areas which have been disproportionately affected by COVID-19, helping assist doctors and healthcare professionals in preventing misdiagnosis and thus decrease morbidity and mortality of COVID-19 associated pediatric diseases.