Google pronounces funding for AI-enabled digital well being tasks

Google introduced that it’s funding 15 AI-powered tasks, together with digital well being initiatives to enhance supplier expertise and affected person entry to care, through its dedication to advancing the United Nations Sustainable Development Goals. 

Every challenge obtained $3 million in technical help, money assist and Google Cloud credit. A handful of tasks obtained Google.org Fellowships, the place a workforce of Google workers works with a corporation professional bono full time for as much as six months.

Of the 15 AI tasks funded, the next eight digital well being endeavors had been awarded funding:

RAD-AID gives low-source hospitals with an AI-enabled platform that helps triage sufferers, primarily relating to respiratory illness and breast most cancers. The platform additionally helps interpret X-rays and scans and supply take a look at outcomes. 

Wuqu’ Kawoq and secure+natal are collaborating to develop a machine learning-enabled instrument equipment to assist midwives in rural areas of Guatemala detect neonatal problems in real-time, akin to poor fetal progress and fetal stress throughout supply. The instrument equipment will encompass an ultrasound and blood stress monitor related to at least one’s smartphone. 

MATCH (Music Attuned Expertise – Care through eHealth) is a challenge constructed out of the College of Melbourne and CSIRO that mixes music and wearable sensor expertise to lower agitation in sufferers with dementia. Google’s grant will assist the workforce develop the sensor expertise and AI-enabled adaptive music system.

Makerere AI Lab will develop a 3D-printed adapter that processes photos utilizing AI and is suitable with a telephone or microscope. The aim is to assist suppliers in Uganda diagnose sufferers with diseases, akin to tuberculosis, malaria and most cancers in low- and middle-income nations the place lab technicians are scarce.   

IDinsight with Attain Digital Well being developed a pure language-enabled question-answering service for expectant moms in South Africa, which gives solutions to inquiries and important well being data.  

Causal Foundry seeks to develop a smartphone-based instrument that makes use of machine studying to assist neighborhood well being suppliers in Sub-Saharan Africa handle affected person data and habits modifications associated to being pregnant and childbirth.

Jacaranda Well being delivers an SMS-based digital well being platform that solutions questions for expectant moms in Sub-Saharan Africa. The platform gives behavioral nudges and features a pure language-powered assist desk that helps triage sufferers and join them to human brokers. The funding shall be used to refine the machine studying mannequin throughout the platform.  

The College of Surrey and Signapse will use generative AI to translate on-line and offline textual content in actual time for deaf folks within the U.S. and U.Ok. and supply photorealistic movies in signal language, allowing extra accessible entry to healthcare and different data.  

THE LARGER TREND

Google has its personal machine studying expertise, dubbed Med-PaLM 2, aimed toward bettering healthcare data entry. Med-PaLM 2 makes use of the tech firm’s massive language mannequin to reply medical questions. 

In March, Med-PaLM 2 was tested on U.S. Medical Licensing Examination-style questions and carried out at an “knowledgeable” test-taker stage with 85%+ accuracy. It additionally obtained a passing rating on the MedMCQA dataset, a multiple-choice dataset designed to deal with real-world medical entrance examination questions. 

One month later, Google introduced it will make Med-PaLM 2 accessible to pick out Google Cloud clients to discover use circumstances, share suggestions and for restricted testing. 

The corporate additionally announced a brand new AI-enabled Claims Acceleration Suite, created to assist with the method of prior authorization and claims processing in medical health insurance. The suite converts unstructured information (datasets not organized in a predefined method) into structured information (datasets extremely organized and simply decipherable).

In July, a examine carried out by Google researchers and printed in Nature revealed that Med-PaLM provided long-form answers aligned with scientific consensus on 92.6% of questions submitted, which aligns with clinician-generated solutions at 92.9%.