INTRODUCTORY ALGORITHM FOR EDGE DETECTION IN MAMMOGRAM IMAGES: A DIDACTIC APPROACH

Resumo

This paper aims to present a didactic algorithm for contour detection in mammogram images. The algorithm uses image processing methods and receives an input image from a mammogram exam, returning a filtered output image, with the identification of existing nodules. Images from an academic repository were used for educational purposes. The algorithm was able to identify contours in the output image and presented high precision (0.9536), indicating few false positives. Accuracy was moderate (0.6023), but performance was limited by low recall (0.2270), F1-Score (0.0718) and IoU (0.0374) values. The data indicate that, although it is correct when detecting, the algorithm did not detect a portion of the real contours. Despite this, the algorithm can be improved as a didactic tool, with a view to learning concepts that are present in certain technologies.