Google has announced a new tool to help users detect skin diseases through the company's AI algorithms, which is set to be launched later this year. According to Peggy Boy, the director of the new tool at Google Health, many users of the global search engine are searching for information about skin, hair, and nail problems, totaling up to 10 billion search queries each year, but they often cannot access accurate information.
Boy adds that nearly two billion people worldwide suffer from skin diseases, and the first step for patients is often turning to Google. However, words alone cannot accurately describe their condition, leading to misleading information.
**How to Use It**
During a Google conference this week, the company showcased the new skin disease detection tool. Users are required to upload three well-lit and clear photos of their skin, hair, or nails from different angles. The tool then asks a series of questions about the user's skin type, the duration of the problem, and other symptoms that help narrow down the possibilities.
The next step involves analyzing the AI model based on the information provided, relying on a variety of dermatological symptoms reviewed by specialists. According to Peggy Boy, the tool is based on 288 criteria to provide a list of potential matches for the user's condition along with predictions.
Following this, the tool presents information reviewed by dermatologists, answers to frequently asked questions, and matching photos. Users can choose to save their results, delete them, or contribute to Google's research and development processes.
The new product manager at Google confirms that the information provided by users is securely stored and encrypted, and the company does not use that data for targeted advertising. Boy points out that the tool is intended to help users initially identify skin diseases, but it cannot replace professional medical advice.
**Work Journey**
It took three years of machine learning research and development to produce this new tool, according to project lead Yuan Liu. So far, Google has published numerous peer-reviewed research papers, contributing to the validation of the AI model, with more studies ongoing.
A study published in the journal Nature last year found that the AI system is as good as a dermatologist in identifying 26 skin conditions and is more accurate than primary care physicians and nurses. Similarly, a recent research paper published in JAMA Network Open showed that the new tool could assist doctors in diagnosing skin diseases more accurately during preliminary evaluations.
The AI model powering the tool has passed "clinical validation" and has been classified as a Class I medical device in the European Union. The Google Health team developed the AI model to consider factors such as age, gender, race, and skin type, ranging from light non-tanned skin to brown skin.
Liu and Boy concluded, "We have developed our model and enriched it with unspecified data, including around 65,000 images and case data for diagnosed skin diseases during trials, millions of curated images related to skin concerns, and thousands of examples of healthy skin across different population demographics."
