Google's AI is better at detecting cancer than doctors!
"We know that when cases are diagnosed early, patients have a higher chance of survival[...]But unfortunately, over 80% of lung cancers are not caught early." – Dr Lily Peng
Lung cancer is the leading cause of cancer death with over 200,000 people being diagnosed and over 150,000 people dying each year because of it. Google's AI researchers have created an AI model capable of detecting signs of lung cancer earlier for faster diagnosis. Usually, pathologists have to review pathology slides in lymph node biopsies to try and detect tumours. This process often proves to be very time consuming, fatiguing and substantially relies on the pathologist’s experience and expertise, therefore many early signs of cancerous tissue could sometimes be missed by pathologists. This results in the diagnosis being delayed, which catastrophically decreases the patient’s chances of long-term survival since advanced cancers are more difficult to treat.
‘I think it has been very effective at identifying areas of concern and getting a second opinion’ — An expert pathologist
Google’s AI model can also detect early signs of breast and prostate cancer too! This is incredible news since the survival rates for these cancers are very high when detected at an early stage. This means that not only people's lives will be extended but this machine learning model could save many people’s lives. It uses a camera enabling it to see exactly what the pathologist sees; it then processes this data through a computer to check for any recognisable tumours. It was trained with over 42000 chest CT scans from nearly 15000 thousand patients. Many experienced pathologists taught the machine learning model what cancer cells look like. Google wants to enable almost all pathologists to have access to this machine learning model regardless of their location or funding. The model will be made available through Google Cloud Healthcare API while Google continues to improve it.
This model is so accurate that it outperformed humans on a CT scan analysis with up to 5% more cancer detection(false negatives) as well as reducing false positives by 11%!
This technology will benefit the patient at the end of the day.