一个结合形态学滤波和高斯-拉普拉斯滤波检测乳腺X线影像中钙化点的新方法

A New Approach to Mammogram Detection by Using Morphological and Laplacion-of-a-Gaussian Filter

微钙化点是乳腺X线影像上独立的亮点,它表征乳腺癌的早期症状。但由于乳腺图片本身规模大、复杂且噪声多,微钙化点体积很小,与乳腺背景对比度又低,所以很难被检出。形态学带通滤波(MBF)算法尽管能快速检出钙化点,但精度不足。高斯-拉普拉斯滤波(LoGF)尽管能比较精确地检出钙化点的位置但较费时。本文提出一种结合上述两者优点同时能克服其缺点的钙化点检测新方法。在南京中大医院乳腺癌数据集上所做实验结果表明,该方法的检测时间接近于MBF方法,同时准确度和LoGF检测方法相当。

Microcalcification is an early sign of breast cancer appearing as isolated bright spots in mammogram images.However,there is a difficulty in detecting the spots because they are small-sized and have noisy and big image background.Morphological bandpass filter（MBF） is a fast method for detecting microcalcifications,but the accuracy there-by is not satisfied.Though Laplacion-of-a-Gaussian（LoGF） method can achieve high accuracy in location,it is time consuming.For these reasons,a new detection method for combining the two above-mentioned methods is proposed in this paper.We conducted the experiments on the breast cancer database of Nanjing Zhongda Hospital.The experimental results confirm that the detecting speed for microcalcifications is comparable to that with the use of morphological filter method,and the detection precision is comparable to that with the use of LoGF method.