The End Of Knowledge - Vault 4/46 - AI For Good - 2020 - New Ways of Thinking of the Mobile Phone for Healthcare and the current Pandemic

graph LR classDef main fill:#f9f9f9, font-weight:bold, font-size:14px classDef phones fill:#ffcc99, font-weight:bold, font-size:14px classDef ai fill:#ccff99, font-weight:bold, font-size:14px classDef healthcare fill:#99ccff, font-weight:bold, font-size:14px classDef diagnostics fill:#ff99cc, font-weight:bold, font-size:14px classDef challenges fill:#ccccff, font-weight:bold, font-size:14px A[New Ways of
Thinking of the
Mobile Phone for
Healthcare and the
current Pandemic] A --> B[Mobile phones
and AI
for healthcare. 1] B --> B1[Professor at
University of Washington,
director at Google. 2] B --> B2[AI for Good
Summit: practical
AI for SDGs. 3] B --> B3[ACM is
a key partner. 4] B --> B4[Pandemics, technology
enable health
paradigm shifts. 5] A --> C[Mobile devices,
wearables
for health monitoring. 6] C --> C1[Phones enable telemedicine,
diagnostics, continuous
measurement. 7] C --> C2[Current apps: manual
data entry
by users. 8] C --> C3[Modern smartphones have
many sensors
for health. 9] C --> C4[Provide health insights,
not clinical
accuracy. 10] A --> D[Phone sensors
monitor lung function,
hemoglobin, bone density. 11] D --> D1[App for lung function
with phone
microphone. 12] D --> D2[Microphone detects lung
obstructions,
disease. 13] D --> D3[Spirometry app tested
on 10,000
subjects. 14] D --> D4[Cough assessment
app for TB. 15] D --> D5[Phone cameras screen
for jaundice. 16] A --> E[Noninvasive hemoglobin
screening with phone
camera. 17] E --> E1[Bone density via
vibrational
resonances. 18] E --> E2[Temperature sensors
for fever
detection. 19] E --> E3[AI interprets rapid
diagnostic kit
results. 20] A --> F[Challenges: AI
as a medical device,
trust. 21] F --> F1[On-device AI
for privacy,
federated learning. 22] F --> F2[Test on low-end
phones for
equity. 23] F --> F3[Rigorous studies,
field deployments
needed. 24] F --> F4[Early detection, prevention
with mobile
health. 25] F --> F5[Mobile tech: triage,
adapt interfaces,
monitor treatment. 26] A --> G[Handset makers
include health
sensors. 27] G --> G1[Consider unintended
consequences of
insights. 28] G --> G2[Demonstrate accuracy,
trustworthiness
of apps. 29] G --> G3[Mobile health, AI
can revolutionize
healthcare. 30] class A main class B,B1,B2,B3,B4 phones class C,C1,C2,C3,C4 ai class D,D1,D2,D3,D4,D5 healthcare class E,E1,E2,E3 diagnostics class F,F1,F2,F3,F4,F5 challenges class G,G1,G2,G3 future

Resume:

1.- Dr. Shwetak Patel discusses using mobile phones and AI for healthcare, especially in light of the current coronavirus pandemic.

2.- Patel is a computer science professor at University of Washington and director of health at Google.

3.- The AI for Good Global Summit, organized by ITU and partners, aims to identify practical AI applications for sustainable development.

4.- ACM, the Association for Computing Machinery, is a key partner and gold sponsor of the AI for Good Summit.

5.- Major health paradigm shifts have been enabled by pandemics and technological advances in areas like vaccines, treatment, robotics, imaging.

6.- A current paradigm shift involves using mobile devices, wearable sensors, and AI/machine learning to enable health monitoring outside hospitals/clinics.

7.- Mobile phones are ubiquitous and personal, enabling telemedicine, disease screening, new diagnostics, treatment monitoring, and continuous physiological measurement.

8.- Most current mobile health apps involve manual data entry by users, e.g. for nutrition tracking.

9.- Modern smartphones have many sensors that can be leveraged for health sensing using AI, without additional hardware.

10.- Goal is to provide helpful health insights, not necessarily clinical-grade accuracy, to guide next steps for users.

11.- Patel's lab has worked on using phone sensors like microphones, cameras, accelerometers for monitoring lung function, hemoglobin, bone density etc.

12.- For lung function, they developed an app using the phone microphone to conduct spirometry tests of breathing capacity.

13.- Vocal tract resonances captured by microphone while breathing out can indicate lung obstructions and disease.

14.- Spirometry app was tested on 10,000 subjects, found comparable to clinical spirometers. Enables screening by community health workers.

15.- For cough assessment to screen TB, modeled cough sounds to identify abnormal lung conditions. On-device analysis preserves privacy.

16.- Used phone cameras to screen for jaundice in newborns by analyzing skin tones. Close to accuracy of $10K clinical devices.

17.- Similar noninvasive hemoglobin screening by analyzing colors of finger pressed against phone camera and flash. Tested in Peru.

18.- Analyzing bone density through vibrational resonances from tapping elbow, captured by accelerometer. Screening for osteoporosis.

19.- Exploring use of phones' temperature sensors to screen for fevers, e.g. for coronavirus, by detecting elevated skin temperature.

20.- AI enables interpreting color changes in rapid diagnostic kits for malaria, COVID-19 etc by taking phone photos of test strips.

21.- Key challenges: Regulating AI software as a medical device. Building trust in phone health apps as accurate and private.

22.- On-device AI avoids sending sensitive health data to the cloud. Federated learning enables insights while preserving user privacy.

23.- Not everyone has smartphones, so testing on low-end phones is important to develop equitable solutions for global health.

24.- Engaging healthcare providers requires rigorous clinical studies, field deployments showing health outcomes. Entrenched systems change slowly.

25.- Mobile health enables early detection and prevention, not just disease management. AI can spot physiological anomalies preceding symptoms.

26.- Three key healthcare roles for mobile tech: triaging care, adapting interfaces for impairments, monitoring treatment efficacy remotely.

27.- Engaging handset makers to include sensors and capabilities that enable global health equity through these mobile solutions.

28.- Considering unintended consequences, e.g. insights into sensitive conditions users haven't disclosed, as these powerful tools are developed.

29.- Demonstrating to providers and patients that these "cheap" mobile apps can be accurate and trustworthy is an ongoing challenge.

30.- Huge opportunity for mobile health and AI to revolutionize healthcare access, quality and equity worldwide, if challenges can be navigated.

Knowledge Vault built byDavid Vivancos 2024