Lung cancer is the disease with the highest mortality rate among cancer. Google is developing AI to detect these lung cancers. AI will perform cancer screening using LDCT, and it can detect cancer with higher precision than radiologists.
The Google research team used an anonymous chest CT scan data from the US National Institutes of Health and Northwestern University to provide end-to-end in-depth learning using data from AI and the results obtained. Six AI and human specialists who completed the study later conducted lung cancer screening using LDCT data.
A cancer screening test on a single CT scan shows that AI is thought to be cancerous rather than human, but in fact it reduces cancer by 11% and cancer detection rate by 5%. Even in the absence of cancer, AI can detect cancer with high accuracy.
There was little difference in the cancer detection rate of AI compared to human specialist when referring to past CT examination as well as single CT scan. In future predictions, AI outperformed humans. In an experiment that predicted the risk of lung cancer two years after the test, the detection rate of AI test was more than 9.5% higher than the result of human radiologist.
However, there is a potential risk for medical diagnosis using AI. There was only one radiologist misdiagnosed as a mistake in reading a CT scan, but AI misdiagnosis could affect many patients. This is because the learning process can lead to errors.
The researchers say that deep-run can increase specificity without sacrificing search speed, and it is necessary to conduct various studies and research on the role that AI can play in cost-effectiveness of cancer screening. For more information, please click here .