Latest Google AI To Predict Lung Cancer Better Than Humans
Alphabets’ Google is all set to expand its presence in the healthcare sector. Scientists at Google have reported have novel research wherein AI – artificial intelligence has been trained to predict lung cancer risks. They claim the AI works far better than humans in diagnosing lung cancer.
Over 1.7 Million deaths per year occur globally due to lung cancer as per WHO reports. In 2018 alone 2.9 Million death cases were reported of people who died from lung cancer. Lung cancer, if diagnosed at an early stage has better chances of recovery.
Shravya Shetty, Technical Lead at Google said that lower dose CT screening has proven to reduce the death rates, yet till date, doctors face challenges to detect it at an early stage thus leading to intensive prolonged procedures with inflating financial costs.
She further added that in 2017 their team initiated this project of using AI to address some of these challenges. With the use of 3D volumetric modeling alongside datasets from Northwestern University and other partners, they have made significant progress in designing an AI for Lung cancer prediction that will lay the ground for clinical testing procedures in future.
The most common challenge faced by radiologists is the use of 2D mages scanned in a single CT scan, that makes it harder to spot cancer as it can be very minute in nature.
Google scientists have designed an AI model to overcome these difficulties. The AI generates an overall report of the cancer malignancy in 3D volume and also assists in locating infectious lung nodules. Google’s AI can also scan through a patient’s past records – compare & detect for any discrepancy.
As per reports published in Nature Medicine, the AI model comprises an automated imaging system that can detect lung cancer at a higher accuracy thus increasing possibilities for earlier treatment.
Google’s team claim to have used 45.8K de-identified chest CT screening datasets from NIH – National Lung Screening Trial study & Northwestern University in order to design this lung cancer predicting AI model.
Upon comparison between single CT scans, Google’s AI performed better than six radiologists. So far it has been able to detect 5% more lung cancer cases when compared to unassisted radiologists.
Google Cloud Healthcare and Life Sciences team will now help the research team to make this newly developed AI model be accessible through Cloud Healthcare API.
Prior to this Google had been experimentation with AI in healthcare with research being conducted in areas of eye diseases, etc.