A team of doctors and computer scientists based in Boston and New York have developed a free-to-use intelligent AI platform, called Buoy, which can help determine what illness a person has based on their symptoms.
The founding team of Harvard and Rice alums started developing the eHealth tool in 2014 at the Innovation Laboratory at Harvard with the idea of helping people make the best decisions when they are sick.
“While the internet can be a great resource for a lot of things, the medical information provided by simplistic web symptom checkers is often risky and unreliable,” said Co-founder and CEO of Buoy, Andrew Le.
“Buoy solves this all too familiar problem by leveraging advanced machine learning algorithms to deliver consumers a more personalised and accurate analysis of what could be ailing them, so they can quickly and easily feel in control of their care,” continued Le.
Through an “interview” the user is prompted to enter their age, gender and symptoms and then answer follow up questions, such as the severity of their symptom and the duration of it. Buoy’s advanced machine learning algorithm then analyses the user’s inputs in real-time, measuring them against millions of medical records to intelligently decide the most important question to ask next. Within two to three minutes, Buoy gains a detailed and accurate understanding of the user’s case, and is able to point the to the appropriate care options. If immediate care is needed, Buoy will present the user with different ways to connect with healthcare providers nearby.
“When people are sick, they naturally ask questions and want an expert opinion for what to do next. So giving them an accurate analysis is incredibly important. That’s why prior to coming to market, we put Buoy through a battery of quality control tests. In every instance, the tool has delivered unprecedented results,” said Le.
“Buoy has accurately analysed symptoms from the common cold to abdominal pain to how a change in running shoes has created a muscular or skeletal issue. It is the intelligent answer the internet has been looking for,” continued Le.
Buoy was recently used in a study to determine how the tool interpreted what a cough might mean as compared to the top five web-based symptom checkers. The study examined 100 standardised cases written by physicians, involving 33 different diagnoses, with severity ranging from life-threatening (pulmonary embolism) to benign (normal cough). It also looked at prevalence, from rare (histoplasmosis) to common (cold).
The study then ran all 100 test cases through the competitive set to examine accuracy rates against the actual test case diagnoses. The study found that Buoy’s analysis was 92% accurate as compared to WebMD at 56%, Healthline at 53%, Mayo Clinic at 38% and Isabel at 28%.