Computer Is More Accurate Than Humans At Detecting Cancer

Computers are advancing and are even seeming to surpass human intelligence.

A convolutional learning network, also called a CNN, was tested against 58 dermatologists.

These dermatologists had different levels of expertise in the field. More than half were experts with at least five years of experience. 48% of these dermatologists had less than five years of experience.

The CNN was shown over 100,000 images of skin lesions and was designed to distinguish benign lesions from malignant ones.

In this test, the machine scored nearly 10% higher than the dermatologists in detecting cancerous lesions.

The dermatologists performed better only after they were given more patient information.

The convolutional neural network did misdiagnose some benign lesions as malignant lesions. However, the amount of errors that the CNN made were drastically fewer than the amount that the dermatologists had made, leading experts to believe that this AI has the capability of preventing unnecessary surgeries.

However, computers are not more advanced than humans just yet. Melanoma is difficult to image in particular parts of the body, such as the toes, fingers, and scalp, so if a patient is unaware of a lesion and therefore unable to bring it to the machine’s attention, that lesion could go undetected.

Even though the computer performed very well in the test that it was given, it is still too soon to rely solely on computers for accurate medical results.

As experts Peter Soyer and Victoria Mar wrote, “There is no substitute for a thorough clinical examination.”

As things stand, computers are not yet advanced enough to catch every needed detail. Therefore, the human mind has not become obsolete and has not yet been advanced by computer science.

Leave a Reply

Your email address will not be published. Required fields are marked *