Automatic Edge Correction for Nodule Cancer Segmentation using Fast Scanning Algorithm

Abstract

Nodule detection is one of the goals in the field of medical imaging. The problem that often occurs from the detection
of nodules in medical images such as mammographic image is the result of improper segmentation, which results in
low segmentation accuracy. The low accuracy of segmentation can affect the area and shape of nodules. These two
features can be used for the classification of the cancer stage. Therefore, a method that can be used to improve
segmentation accuracy is proposed, namely edge correction. The initial edge nodules are obtained using a quantum
canny enhancement method. Edge correction is done adaptively by looking at the pixel values around the edge pixels
so that the actual edge pixels are obtained. Edge correction can also connect broken edges by looking at the edge
pattern and direction. The results of this proposed method increase the average segmentation accuracy from 87%,
which is the result of segmentation without using edge correction, to 95.1%. Accuracy is assessed using the Jaccard
similarity method.