Artificial Intelligence Detects Breast Cancer with Near 100% Accuracy

By using artificial intelligence, a computer can find breast cancer 30 times faster than a single doctor can. AI scientists at Houston Methodist have programmed software to do almost all the work of accurately interpreting a patient's mammogram (breast x-ray) with 99 percent probability. This groundbreaking work was published in the journal Cancer, and may lead to breast biopsies becoming obsolete.

Breast Cancer, Mammograms, and the Cost of Diagnosis
According to the Centers for Disease Control and Prevention, mammograms in the US employ huge resources to decrease breast cancer mortality. Almost 67% of women over 40 have regular mammograms. The number of physician appointments due to mammogram procedures has grown to over 12 million per year, and outpatient visits to hospitals have grown to 3 million. Despite all this, there are many false positive results (looks cancerous, but is not). What happens then is an anxiety-causing ordeal for the patient, with more tests and a biopsy. This makes the whole procedure time-consuming and costly.

AI Does the Work of Hundreds of Clinicians
The computer software thinks intelligently like a medical professional with a trained eye. This is accomplished through machine learning, a new and powerful way to make computers think like humans. By sifting through millions of breast cancer records, the computer builds a knowledge base that gives it the ability to interpret a patient's mammogram with unparalleled accuracy and speed. When putting the software through its trial runs, researchers used 500 patients at risk of breast cancer. The AI algorithm assessed all patient reports, diagnostic data, and mammograms. Surprisingly, it took a few hours for the computer to make a cancer risk diagnosis for all 500 patients. Contrastingly, it took two physicians over 50 hours to access the cancer risk of just 50 patients.

If the math is correct, one computer armed with the new artificial intelligence could do the work of over a hundred clinicians. This software could save doctors hundreds of hours of time, and give breast cancer patients shorter waiting times and an accurate diagnosis. This gives doctors the ability to rapidly assess the risk of cancer in their patients, and whether or not further tests are necessary.