Method for automatically detecting position of nipple in molybdenum target image

Abstract

The invention discloses a method for automatically detecting the position of a nipple in a molybdenum target image. The method comprises the following steps that: S1, a mammary gland molybdenum targetimage is read and is preprocessed; S2, the size of a nipple in the mammary gland molybdenum target image is compared with a set standard value, if the size of the nipple in the mammary gland molybdenum target image is greater than or equal to the set standard value, the method enters a step S3, if the size of the nipple is smaller than the set standard value, the method enters a step S4; S3, themammary gland molybdenum target image is classified by using a rolling sphere method, and the position of the nipple is detected; and S4, a BBST (Rubber band straightening transform) algorithm is usedto straighten a breast edge in the mammary gland molybdenum target image into a rectangular RBST image, pixels along the breast boundary are mapped into the first row of the RBST image, and pixels along the normal of the object boundary are mapped into the RBST image according to columns. The method of the invention has the advantages of high accuracy, high efficiency and high robustness. With the method adopted, the multi-perspective analysis potential of the mammary gland molybdenum target image can be improved, the diagnostic accuracy of breast cancer can be improved, and the survival probability of patients with cancer can be improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method for automatically detecting the nipple position in a mammography target image. Background technique [0002] In the field of computer-aided diagnosis of breast cancer, the nipple is the only stable and consistent landmark in mammography images. The diagnosis of nipple position is an important research direction, and there are many diagnostic algorithms, and different diagnostic algorithms have their own advantages. It has advantages, but also disadvantages. [0003] In 2004, Zhou proposed a very strict algorithm for nipple detection. The algorithm is divided into two phases. In the first phase, nipple locations are detected based on significant changes in intensity values ​​along the breast border. In the second stage, the nipple position is detected based on the convergence of texture patterns around the nipple. The final nipple position is found based on a ...

AI Technical Summary

Problems solved by technology

However, dense mammography images may have structured noise and thus may produce false positives when using this method, so the results of this method are still not good enough

[0004] In 2008, Kinoshita et al. found that the position of the nipple is at the intersection of breast tissue components, and used Radon transform to detect the position of the nipple; although this is an excellent algorithm, its limitation is that it cannot accurately detect the breast boundary

But the disadvantage of this method is that detecting the connected region with the largest average Gaussian curvature as the nipple position is not true for all mammogram images, this method is accurate when detecting mammogram images with flat and small nipples low rate

Images