Spotting lung cancer – AI better than humans

May 20, 2019 (Medical News Today) 

The use of computed tomography (CT) scans is becoming more common to screen for lung cancer, and some research has shown that low-dose CT scans are more effective than using X-rays to screen for lung cancer, as they have reduced lung cancer deaths by as much as 20%.

However, a large amount of false positives and negatives have made low-dose CT scans less effective than they could be, so new research has set out to develop an artificial intelligence (AI) program to detect lung tumors in low-dose CT scans rather than only having radiologists examine the scans.

This research was led by Daniel Tse of the Google Health Research Group and the results were published in the journal Nature Medicine, with the study using over 40,000 different low-dose CT scans.

The scans were looked at by the AI program and six different radiologists with up to 20 years of clinical experience, and the AI program outperformed every radiologist. The program had 11% less false positives and 5% less false negatives than the radiologists.

Although lead researcher Tse believes this is a big step in improving the diagnosis a lung cancer, the AI program must be tested further in larger cohorts in order to validate the results.

Read more here.