| dc.description.abstract | Breast cancer is the most common type in women and the second most common worldwide. There is a frequent need to diagnose this cancer at an early stage, giving the patient a greater chance of cure. Among the preventive methods to help early detection of breast cancer, there is mammography, which is simple and noninvasive, but demands highly complex analysis of the images generated. The radiologist must have great experience in mammography findings to detect a possible cancer. CAD (Computer-Aided Detection) is a tool created to aid in this difficult diagnosis task. CAD is a software that analyzes mammographyimages rapidly and separates regions that may be considered pathological findings. Verification of the sensitivity of CAD as a diagnostic tool is very important to validate its use together with the doctor in the early detection of breast cancer. False-positive results generate a stresson the patient and, in many cases, can lead him/her to perform unnecessary procedures. On the other hand, false-negative results are worse, since they may result in high levels of risk to the patient health. This work investigates the sensitivity of the CAD tool to mammography, focusing on false negative results. From a database of 1930 mammograms categorized by a medical specialist according to the BI-RADS (Breast Imaging Reporting and Data System), 45 mammograms were randomly selected for analysis in this study, 30 of them considered to be a cancer with high probability (Category 4 BI-RADS). A second medical expert evaluated these images and the mammographic findings found by the CAD were compared with thosefound by the two doctors, focusing on the false-negative tests with the highest probability of containing cancer. The results suggest that CAD is an effective tool for second reading at diagnosis, with sensitivity superior to 90% for the tested cases. | |
