Google launches AI health tools for skin conditions

[ad_1]

Google is taking the most important step a large technology company has taken so far, launching an artificial intelligence-driven tool that will help consumers self-diagnose hundreds of skin conditions.

Derm Assist is the first product of its kind and will be launched in Europe this year, targeting nearly 2 billion people worldwide with skin conditions ranging from acne to melanoma.

Users upload images of their medical conditions through the Derm Assist website and answer questions about their symptoms. The AI ​​model then analyzes this information and generates a list of possible matching conditions. The service will be provided free of charge to all Internet users, regardless of whether they are Google users or not.

“This tool is not designed to provide a diagnosis… On the contrary, we hope that it will give you authoritative information so that you can make a more informed decision about the next step.”

Google launched the product after three years of development. Google has long regarded healthcare as a mature market disrupted by advanced artificial intelligence technology. Competitors Apple, Amazon and Microsoft have also entered potentially lucrative fields to provide medical services to consumers, doctors and pharmaceutical companies.

Google chose skin disease as the first target of AI-driven healthcare because a large number of people are affected by skin conditions. The search giant said that about 10 billion Google searches per year are related to skin, nails and hair, and research shows that people can diagnose themselves correctly only 13% of the time.

Google’s new AI 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.” Google Health Product Manager , Said Dr. Peggy Bui, an internal medicine expert at the University of California, San Francisco.

The Derm Assist system is based on machine learning algorithms trained on more than 16,000 real-world dermatology cases.according to A study Since last year, the tool has been able to accurately identify skin conditions like a dermatologist certified by the US Board of Directors.

Certain information provided to users has been reviewed by human dermatologists. If the user mentions any shocking symptoms, such as being unable to breathe, other alerts will advise them to see a doctor immediately.

A study Published In JAMA Network Open, AI tools have also significantly improved the diagnostic accuracy of non-professionals (such as GPs and nurses) who use the tool to help them diagnose skin conditions.

“Our observations show that artificial intelligence has the potential to enhance the capabilities of artificial intelligence.[generalist doctors and nurses]. . . So as to diagnose and classify skin conditions more effectively. “The study author Yuan Liu and her team wrote in a peer-reviewed paper. “Improved the diagnostic accuracy of non-referral cases. . . May have a huge impact on the healthcare system. ”

Eric Topol, a professor of molecular medicine at the Scripps Research Institute and an expert in artificial intelligence and healthcare, said: “This is destined to happen at some point, because it is the first in medicine. A major deep learning artificial intelligence use case, and it has been verified to a certain extent. In 2017.”

In order to avoid missed diagnosis of skin cancer due to false alarms, the algorithm should be cautious in its decision-making. Dr. Bui said: “When designing, we stated that we should optimize for high sensitivity, especially for alarms or terrible situations.”

In order to solve the privacy issues related to user health data, Google said that it will not use uploaded pictures as advertising targets, and will only save pictures to further train the Derm Assist algorithm, provided that users have given them explicit permission.

Dr. Bui said: “Users can choose to save, delete or donate data to control their own data,” “We hope to encourage donations, because the algorithm is only as good as the trained data… In addition to the donated data, we will also pass Procurement of other data sets from other sources to continue to improve the model.”

[ad_2]

Source link