While mammography is currently the best method available for the early detection of breast cancer, up to 30% of mammograms in women who do have breast cancer are unfortunately diagnosed as being normal. Furthermore, because these cancers grow at various rates, missed cancers are subsequently detected years later, but by then may have grown substantially, thereby reducing the patient's successful prognosis (or life expectancy).

At the University of Chicago, a world class team has been developing also known as artificial intelligence, for detecting breast cancer automatically. Over the past seven years, we have identified approximately 230 patients in whom a breast lesion was missed. In such cases, a breast cancer is detected in a mammogram; the lesion is then retrospectively identified in the previous mammogram, which was obtained at least one year earlier and was read as a normal study at that time. A lesion is missed either because the radiologist did not see it (observation error) or because the radiologist saw the lesion, but thought it was a benign the radiologist saw the lesion, but thought it was a benign finding (interpretation error). Patients in whom one or both types of errors have occurred will be analyzed using these computerized detection programs. This unique study will allow the potential of our computer programs to reduce the number of missed breast cancers.

We believe that when implemented clinically, our computer programs will be able to reduce the number of missed cancers by 50%. This translates into a lower patient mortality and morbidity. Furthermore, because cancers will be discovered earlier, there should be a significant saving of health care dollars. Evidence indicated that removing small cancers greatly improves successful cure rates.