Artificial Intelligence to read mammograms

Artificial intelligence researchers at Houston Methodist Hospital have developed computer software that could make the dreaded mammogram callback quicker and more reliable.

In a paper published in the journal Cancer, the researchers reported that the software interpreted results of breast X-rays 30 times faster than doctors and with 99 percent accuracy. Human review of mammograms results in unnecessary biopsies nearly 20 percent of the time.

"This software intelligently reviews millions of records in a short amount of time, enabling us to determine breast cancer risk more efficiently using a patient's mammogram," Stephen T. Wong, a Methodist Research Institute biomedical engineer and the study's co-senior investigator, said in a statement. "This has the potential to decrease unnecessary biopsies."
The authors noted artificial intelligence has previously been deployed to extract clinically useful information from mammograms, but said their work is the first to provide standardized guidelines that could aid clinical decisions.

Mammograms are considered one of oncology's best tools --12.1 million are performed annually in the U.S. -- because they can detect breast cancer at an early stage when it most treatable. But the test has its limitations: a false-negative mammogram looks normal even though cancer is present; a false-positive test looks abnormal even there's no cancer.
Biopsies, the removal of sample cells or tissue, are recommended when the radiologist reports anywhere from a 3 to 95 percent likelihood cancer is present. Wong said in the statement he hopes the software helps clinicians better define the exact risk that requires a biopsy.
The team used the artificial intelligence software to evaluate mammograms and pathology reports of 500 breast cancer patients. The software scanned patient charts, collected diagnostic features and correlated mammogram findings with a breast cancer subtype. Clinicians used results, like the expression of tumor proteins, to accurately predict each patient's probability of breast cancer diagnosis.
Without the software, manual review of 50 charts took two clinicians 50 to 70 hours. It took the software a few hours to return its results from all 500 patients.
More than 1.6 million breast biopsies are performed annually in the U.S. The cost ranges from $1,000 to $5,000 -- on the high end if it's a surgical biopsy performed in a hospital, on the low end if it's a needle biopsy performed in a doctor's office.

SOURCE:
The Houstonchronicle