Scientists train a computer to diagnose breast cancer
Breast cancer patients who worry about their finances may ease the burden on their checkbooks if they buy Canadian Tamoxifen at a fraction of what it costs in the U.S.
Breast cancer patients who worry about their finances may ease the burden on their checkbooks if they buy Canadian Tamoxifen at a fraction of what it costs in the U.S. Scientists are continuously looking for better ways to diagnose and treat this disease. One team of researchers from Stanford University may have found a better way to evaluate breast cancer by training a computer to help them.
The methods used to diagnose breast cancer today are very similar to how it was done in 1928, according to the Stanford team. These tests require pathologists to study tissue samples for at least three features: tube-like tumor cells, nuclei diversity among tumor surface cells and rate of cell division. The researchers used such measurements to teach their Computational Pathologist, or C-Path, how to analyze images of cancer tissue and determine survivability.
After learning how to evaluate samples from patients whose prognoses were known, C-Path eventually learned to assess more than 6,600 different factors. This acquired knowledge allowed C-Path to help diagnose a group of patients it had not previously seen.
Furthermore, the computer model identified influential factors in tissue samples that pathologists did not previously focus on, as reported Nov. 9 in the journal Science Translational Medicine.
The team believes this model can eventually help clinics that don't have specially trained professionals, as well as patients who can have various forms of breast cancer. However, they added that it probably won't fully replace pathologists.
|
