There is a new project in the fight against cancer. Google researchers that specialize in artificial intelligence (AI) are working with scientists from Northwestern Medicine to put together AI technology that can find lung cancer based on cancer screening tests.
Comparing the Technology to Human Experts
The new technology was better at finding lung cancer than human radiologists that had eight years of experience in the field. Based on one screening test, the AI found the lung cancer five percent more often than a group of six radiology experts. The technology also gave ten percent less “false positives.” A false positive is when the expert believes there is cancer in the body, but there actually is not. The AI and the experts had about the same success rate when the humans were allowed to look at prior screening tests.
The National Lung Screening Test (NLST) did a study where they compared the AI’s ability to predict if someone would get cancer two years after a screening to an expert’s ability to predict the same. The new technology gave an accurate prediction almost ten percent more often than the human experts were able to – a very significant increase.
How it Works
All of this research was published in a journal called Nature Medicine. The artificial intelligence was used to do three things. First, the scientists used it to predict if the patient had lung cancer. Second, the technology created a score that predicted a patient’s risk for developing lung cancer. Third, it found the exact place that the cancer tissue was in the lungs.
Google’s next step is to make the technology available to other researchers. They are going to make it available through the Google Healthcare Cloud. The company is also planning further testing on the AI and is going to partner with other organizations on these tests.
Google’s lead researchers, named Shravya Shetty and Daniel Tse, have been giving updates on the technology’s development through blog posts. They say that the AI works by creating a three-dimensional picture of the entire chest area. Once it has the whole picture, the AI can then find the areas that have the cancerous tissue.
The scientists “trained” the technology by showing it nearly 50,000 different CT scans from 15,000 patients. Unfortunately, almost 600 of those patients went on to get cancer within a year. This was done during a study by the National Institutes of Health in 2002 that was later approved by Northwestern Medicine.
The Fight Against Lung Cancer
The statistics on lung cancer are frightening. The disease kills more than two million people every year, about the same as breast cancer. The World Health Organization says that makes lung cancer one of the deadliest diseases on the planet. One of the reasons for this high death rate is that less than five percent of patients get screened for lung cancer, according to a 2015 study.
The Google blog post went on to say that they hope the AI technology can make cancer screening more effective and affordable. They hope the results inspire more research and conversation about lung cancer.
Other Google Research
This also is not the first time Google has done extensive cancer research. A different technology they made, called Google Inception v3, was used in lung cancer research at New York University last year. AI is also an important part of a technology Google uses to find diseases in patients’ eyes. The technology can recommend the right kind of treatment for 50 eye diseases with over 90% accuracy.