Reports indicate that artificial ielligence can perform better than doctors in detecting ovarian cancer from ultrasound images.
According to a new study whose results have been published in the journal Nature Medicine, a specially trained artificial ielligence program has achieved an accuracy rate of over 86% in detecting ovarian cancer through ultrasound scans. Meanwhile, the accuracy rate of human experts was slightly less than 83% and the accuracy rate of non-specialist doctors was about 78%.
The researchers trained an artificial ielligence program to distinguish between benign and maligna ovarian lesions. For this purpose, more than 17,000 ultrasound images of nearly 3,700 paties in 20 hospitals located in eight couries were used.
According to Elizabeth Epstein, one of the senior doctors in the Departme of Clinical Sciences and Education at Stockholm South General Hospital, ovarian tumors are common and often detected by chance, and artificial ielligence can help diagnose ovarian cancer, especially in difficult-to-diagnose cases as well as in situations that there is a shortage of specialists in the field of ultrasound, to be a useful tool.
It is worth noting that this artificial ielligence program was also able to reduce the need to refer paties to specialists, as it acted as a backup for the evaluation of human doctors. In a simulated care scenario, AI assistance reduced the number of referrals to specialists by 63%. It also reduced the rate of diagnostic errors by 18%.
