نبذة مختصرة : Breast density has assumed an increasing importance due to its relationship with breast cancer, but its measurement has limitations related to the subjectivity of visual assessment or the difficult access to automatic methods of volumetric measurement. Our study proposes the development and clinical validation of the Aguida® software program to determine this density, in an objective and reproducible way, from flat mammography images. When using the software, the operator positions the cursor that limits the region of interest on the pectoral muscle in oblique mediolateral view. Then the program performs the segmentation of the image, separating the background from the breast image. In craniocaudal view, glandular density is classified based on the optical density of the pectoral muscle and the proportion of dense tissue is calculated automatically. For standardization of the method, 291 mammograms were used and for clinical validation mammograms obtained at 2 different times in 282 women. The program was developed in MatLab. As results we obtained an intraclass correlation coefficient (ICC) between the measurements of breast density, by the program and by a radiologist, 0.96 (95% CI 0.95 to 0.97), with a bias of only 0.67 percentage points and limit of agreement 95% of 13.5 percentage points; in assessing interobserver reliability, with two radiologists with different experience, the CCI was 0.94; in assessing intraobserver reliability, the ICC was 0.98. The distribution between density classes is close to the values obtained with volumetric software. In conclusion, the determination of breast density by the Aguida® program in flat mammography images showed high agreement with the visual determination by a radiologist, high inter- and intra-observer reliability. ; A densidade mamária tem assumido importância crescente devido a sua relação com o câncer de mama, mas a sua mensuração tem limitações ligadas à subjetividade da avaliação visual ou no difícil acesso aos métodos automáticos de medição volumétrica. ...
No Comments.