The world’s largest search engine, Google, receives approximately 1 billion health-related queries. Not monthly, not weekly, but daily. On Google, patients enter their symptoms before making an appointment. Not everyone is aware that the results have nothing to do with medical knowledge but depend on algorithms—and they favour the most popular content. The result: many people consider Google their primary care physician because the internet knows everything. However, for some people, especially those struggling with financial difficulties or without health insurance, this is the only way to diagnose their disease. According to WHO, half of the world’s population does not have access to essential health services.
When access to a doctor is limited, and each unnecessary visit is associated with high costs, it is easier to ask Google. Others do it for the sake of convenience—because Google is at hand. However, the truth is that Google can mislead patients, lead to delayed treatment and diagnosis, and sometimes endanger their health and life. Doctors have been dealing with the problem for years—patients bring printouts of results displayed by search engines, expecting, for example, to prescribe a specific drug or argue about the disease.
Encyclopedia with the problem of information evaluation
Google search engine is today the largest repository of knowledge in the world—every day, it receives 5.4 billion queries. People ask for a cake recipe, weather forecast, restaurant website, TV program. But also about what fever and rashes mean and how to treat a long-lasting cough. Google will find this information in a second. The value of access to information offered by Google is beyond dispute.
The challenge is that AI algorithms cannot separate scientifically verified information from individual opinions and judgments unrelated to medicine. At the same time—and it needs to be made clear—Google plays a vital role in preventing, diagnosing, and treating diseases even if it is inconvenient for the health sector.
But all signs point to significant change ahead. In May 2021, the California-based technology giant announced the progress on the first app that will allow you to assess the health of skin, nails, and hair based on photos taken with a smartphone and answering several questions. Does this mean that Dr. Google, who is synonymous with misdiagnosis, aspires to become an MD present in every corner of the world, in every home?
The application will classify dermatological problems
Two billion people worldwide suffer from skin, nail, and hair diseases. Every year, 10 billion dermatological queries go to the Google search engine—phrased differently, using different words, and therefore very often imprecise. The new web application, which will be launched as a pilot later this year, works like a simplified symptom checker. First, the user takes three photos of their skin, hair, or nails from different angles. In the next step, the system will ask additional questions, e.g., about the lesions and other symptoms. An AI model will analyse this information and display the results: information verified by dermatologists, a list of answers to frequently asked questions, and similar photos posted on the internet.
“We hope it gives you access to authoritative information so you can make a more informed decision about your next step,” declares Google, emphasizing that it is not a diagnostic tool. On the other hand, we learn that the system can achieve the same precision as dermatologists.
An algorithm based on many years of scientific research
The AI-assisted dermatology tool is the result of over three years of machine learning research. The results have been published in Nature Medicine. In JAMA Network Open, researchers working with Google prove that by using tools based on artificial intelligence, non-dermatologist physicians can improve their skills in interpreting the symptoms of skin conditions.
Google has trained its model on 65,000 photos and cases of diagnosed skin diseases, millions of photos of various dermatological problems, and thousands of photos of healthy skin—in groups of different age, sex, race, skin types. The AI model has been CE marked as a Class I medical device in the EU.
Can we expect more health assessment systems?
Creating an algorithm that recognizes lesions cannot be called a revolution—such apps are already available. However, their developers do not have the same access to potential consumers as Google does. Assessing skin problems is one of the easiest tasks for AI because it relies on objective data—photos.
The problem gets complicated in other diseases, where the symptoms are non-specific or may be described differently by the patient and therefore require careful evaluation by a physician. Nevertheless, the AI system can assess simple health ailments by asking detailed questions—several startups worldwide are currently developing such tools, f.e. Mayo Clinic, WebMD, Symptomate, Your.MD, Ada.
Increasing pressure on Google will force the big tech company to embed such an app in the search engine to end the rising problems of misdiagnosis on the internet.