Artificial Intelligence is already revolutionizing medicine and it’s going to change healthcare as we know it.
As we can see in almost every layer of our lives, Artificial Intelligence is already changing how we see and connect with the world. But the improvements that it’s making in the field of medicine and healthcare are a real game-changer.
In fact, A.I. is already making a difference in the areas of Radiology, Dermatology, Ophthalmology, Oncology, Cardiology, and several others. Especially when it comes to reading and interpreting images and scans.A recent study at the Medical University of Vienna led by dermatologist Herald Kittler has shown that software developed for classifying moles outperformed physicians when detecting skin lesions, especially one called Pigmented Actinic Keratosis - a rough, scaly patch on the skin that develops from years of exposure to the sun. This algorithm can distinguish between different kinds of cancerous and benign lesions after being fed 10,000 images labeled by doctors. When doing reverse engineering on the software to find out how it arrived at these conclusions, it showed that the system paid more than the usual attention to the skin that surrounded the colored areas, therefore the algorithm may have been detecting sun exposure in the skin around the lesions. Thanks to these discoveries, Kittler and his colleagues now teach future doctors to look for sun damage in the skin around a blemish.
“The US Food and Drug Administration has approved several AI imaging products in recent years, including AI tools for diagnosing eye conditions such as diabetic retinopathy, glaucoma, and age-related macular degeneration”
Another example of how A.I. is changing the healthcare system is currently underway at Duke University Hospital in Durham, North Carolina, where they are using a machine learning software called Sepsis Watch. This system raises an alarm when a patient is at risk for developing sepsis - a condition caused by the body's response to an infection, and the number 1 cause of death in US hospitals. Tuned with 32 million data points from past patients, the algorithm scans patients charts in real-time and alerts nurses by sending them notifications when patients are flagged with moderate to high risk for sepsis. The system was created by the researchers at the Duke Institute for Health Innovation, built with one of the tech companies’ favorite techniques called “deep learning” and trained on 50,000 patient records.
The US Food and Drug Administration has approved several AI imaging products in recent years, including AI tools for diagnosing eye conditions such as diabetic retinopathy, glaucoma, and age-related macular degeneration. These tools promise to help improve the standards of care that the eye doctors can offer their patients and many ophthalmologists have been showing a lot of interest in them. The giant Google has gone as far as setting up a system in Thailand that can detect eye disease in diabetics with 90 percent accuracy.
However, there’s still room to improve. Eye imaging algorithms might fail. In the case of Google’s system, more than 20 percent of images are rejected, mostly due to lighting quality. Whatsmore, according to research done in the UK there is a lack of diversity in the eye images data, as most of it comes from patients in North America, Europe, and China, making it less likely to work properly within the African, Latin American, or Southeast Asian population. But as this technology becomes more widely available, doctors are hoping to see an improvement in demographic machine learning training data.
Even though A.I. has been shown to outperform doctors in several areas, the idea of medical specialists being replaced by machines is far from happening. The algorithm can still make mistakes. In the real world, these tools act better as assistants to real doctors, who not only diagnose but also treat and build relationships with their patients. The key is collaboration, Artificial Intelligence elevates rather than eliminates doctors, it is already making a difference in real clinical settings and shows amazing promise for the years to come.