Google launches AI health tool for skin conditions


Google is taking one of the biggest steps yet by a leading healthcare tech company, launching an AI-powered tool that will help consumers self-diagnose hundreds of skin conditions.

Derm Assist is the first of its kind and will launch in Europe this year before reaching the nearly 2 billion people worldwide with skin conditions ranging from acne to melanoma.

Users will need to log into their Google accounts, upload images of their condition through the Derm Assist website, and answer questions about their symptoms.

An AI model then analyzes the information and generates a list of possible match conditions. The service will be free for all Internet users, whether or not they are Google users.

“The tool is not intended to provide a diagnosis. . . rather, we hope this will give you access to authoritative information so you can make a more informed decision on your next step, ”Google said.

The launch follows three years of development at Google, which has long viewed healthcare as a market ready to be disrupted by advanced artificial intelligence. It comes as rivals Apple, Amazon and Microsoft also enter the potentially lucrative space, creating healthcare services for consumers, doctors and pharmaceutical companies.

Google chose dermatology as its first target for AI-driven healthcare due to the large number of people affected by skin conditions. About 10 billion Google searches are performed on skin, nail and hair problems each year, and studies have shown that people only diagnose themselves correctly 13% of the time, the search giant said.

Google’s new artificial intelligence tool Derm Assist © Google

“Skin diseases as a category are a huge global burden – people are turning to Google to study their skin problems. Most cases are curable, but half of the world’s population faces a severe shortage of dermatologists, ”said Dr. Peggy Bui, product manager at Google Health and internal medicine specialist at the University of California at San Francisco.

The Derm Assist system is based on a machine learning algorithm trained on over 16,000 real dermatology cases. According to a study as of last year, the tool is able to identify skin conditions as accurately as certified dermatologists in the United States.

Some of the information provided to users is reviewed by human dermatologists. If a user reports alarming symptoms, such as an inability to breathe, additional alerts advise them to seek immediate medical attention.

A study published in JAMA Network Open found that the artificial intelligence tool also significantly improved the diagnostic accuracy of non-specialists such as general practitioners and nurse practitioners, who used it to help them diagnose skin conditions.

“Our observations suggest that AI has the potential to increase the ability to[generalist doctors and nurses]. . . to diagnose and sort out skin conditions more effectively, ”study author Yuan Liu and her team wrote in a peer-reviewed article. “Improve the accuracy of the diagnosis of unreferenced cases. . . could have huge implications for health systems.

Eric Topol, professor of molecular medicine at the Scripps Research Institute and expert in AI and healthcare, said: “It had to happen at some point, as this was the first major use case for AI in medicine with some validation. in 2017. ”

To avoid missing false negative skin cancer cases, the algorithm was designed to be careful in its decision making. “When we designed this, we said we wanted to optimize for high sensitivity, especially for alarming or frightening conditions,” Dr Bui said.

To address privacy concerns over users’ health data, Google said it will not use the uploaded images to target advertising and will only save the images to perfect the Derm Assist algorithm, if users request them. gave explicit permission.

“Users have control over their data with the ability to save, delete or donate data for research,” said Dr Bui. “We hope to encourage donation, because the algorithms are only as good as the data they were trained with. . . We will continue to improve the model by sourcing other datasets from other sources, in addition to the data provided. “



Source link

Leave a Reply

Your email address will not be published. Required fields are marked *