TWI511085B - Nipple image segmentation method - Google Patents

Description

Nipple image segmentation method

The present invention relates to an image processing technique, and aims to provide a nipple image segmentation method capable of quickly and accurately dividing a nipple image block among chest image images.

According to the woman, the woman is full of motherhood and feels warm and soft creatures. The reason for the difference between women and men is that in addition to sound and personality, there is a very important reason because women’s unique women symbolize breasts. It is a great sign that represents the birth of the next generation; women generally have this characteristic after maturity (unlike other primates only expand and bulge during breastfeeding), and have the charm of attracting the opposite sex, so the breast For women, it is not just a simple breast-feeding organ, but specialization has become one of the most important signs of attracting the opposite sex.

In principle, the breast is mainly composed of breast, milk duct, adipose tissue and fibrous tissue; among them, the gland is composed of 15-20 breast lobe, and each breast lobe contains many mammary lobules ( Lobule), each mammary lobules is composed of 10-100 acinus (acinus), which collects 15-20 lactiferous ducts, which are radially aggregated into breast sacs (ampulla) ) and open to the nipple.

According to statistics, the most common cancers in women with breast cancer (except skin cancer and lung cancer), although early detection leads to a higher cure rate, breast cancer is still the leading cause of cancer death in adult women under 54 years old, and for those over 54 years old Is second Bit. In women of all ages, breast cancer is the leading cause of cancer death in women after lung cancer. With the advancement of technology, in the process of breast examination, the breast tissue will be examined in detail through mammography, and even the relevant image processing software can be used to find the breast and the algorithm can accurately find the position of the tumor. For further testing or treatment.

Furthermore, when the tumor position is actually calibrated, the relative position of the nipple is used as the basis for calibrating the tumor position; however, the shape of the breast of the human body is not completely the same, and the shape of the left and right breasts of the same person is not completely identical. The position of the nipple also varies from person to person. Therefore, if a method can be provided to analyze or segment the image block belonging to the nipple in a relatively objective and accurate manner, such as a breast image, it will help to accurately calibrate the relative position of the tumor, which not only leads to a more accurate test. The data is more helpful to improve the cure rate of breast cancer.

In view of the above, the present invention provides a method for segmenting a nipple image which can quickly and accurately divide a nipple image block among chest image images, and is a main object thereof.

To achieve the above objective, the nipple image segmentation method disclosed by the present invention basically comprises the following steps: (a) loading a digitalized chest section image; (b) finding a part of the chest section that belongs to the chest cavity. (b) establishing a chest line fitting curve to remove the image of the chest object in the image area of the object; (d) ascertaining structural parameters of the image of the breast object belonging to the breast part among the image areas of the object; (e) obtaining the edge point of the breast for one of the breasts for which the image analysis is desired; (f) obtaining a rough estimate of the nipple of the breast; (g) collecting the breast edge point fitting curve centering on the rough estimate point of the nipple , the nipple image block outside the fitting curve can be segmented.

According to the above technical features, the nipple image segmentation method is taken In the step of estimating the position of the breast nipple, the breast is projected on the X and Y axes, and the position of the top and middle of the breast is roughly searched, and then the nipple is searched within a predetermined pixel range according to the position of the top of the breast. The image pixel of the brightest point is the seed pixel and its grayscale value is recorded, and thus the seed pixel is started, as long as the difference between the surrounding pixel grayscale value and the seed pixel grayscale value is within the preset value. So that a nipple drop point is obtained at the breast image, and then the centroid of the nipple drop point block is calculated, that is, the nipple rough estimate point of the breast.

According to the above technical feature, the nipple image segmentation method can further obtain the breast spot on both sides of the breast after obtaining the edge point of the breast for one of the breasts for which the image analysis is to be performed, and then the center of the breast with two bright points Calculate the position of the centerline, and then construct the cleavage baseline by linear regression from the edge of the cleavage, and then create an intermediate point with the intersection of the centerline, and gradually find the tangent point of the breast from the middle point, and find the distance from the middle point and The farthest edge point of the tangent point is regarded as the breast base point, and the image data between the left breast base point and the right breast base point is removed to obtain the image data of the more complete and simple breast.

According to the above technical feature, the nipple image segmentation method can firstly remove the background noise of the chest section image by the operating software first after loading a digitalized chest section image, and then find the chest cavity belonging to the chest cavity portion. Object image.

The nipple image segmentation method is characterized in that in the step of collecting the breast edge point fitting curve centered on the nipple rough estimation point, a horizontal line is drawn at the middle portion of the breast to find the closest horizontal line on both sides of the breast. The midpoint of the two breasts, and calculate the distance from the edge point of the middle part of the breast to the rough point of the nipple, and draw the distance from the edge point of the middle part of the breast to the rough point of the nipple as the center of the nipple rough point. A maximum circle and a minimum circle are drawn, and a reference circle between the maximum circle and the smallest circle is drawn with the average distance between the edge points of the two breasts and the rough point of the nipple, and then the edge point within the reference circle Fit the curve in an elliptical manner Line, you can segment the nipple image block outside the fitting curve.

Specifically, the nipple image segmentation method of the present invention can produce the following effects:

1. Quickly and accurately segment the nipple image block in the breast image in the chest profile image.

2. Help to accurately calibrate the relative position of the tumor.

3. Can still bring more accurate test data and improve the cure rate of breast cancer.

C1‧‧‧Maximum circle

C2‧‧‧Minimum circle

C3‧‧‧ benchmark circle

L1‧‧‧ center line

L2‧‧‧ cleavage baseline

L3‧‧‧ horizontal line

L4‧‧‧ fitting curve

P1‧‧‧ intermediate point

P2‧‧‧cut points

P3‧‧‧ Breast base point

P4‧‧‧ nipple rough estimate

P5‧‧‧ mid-edge point of the breast

P6‧‧‧ mid-edge point of the breast

Figure 1 is a basic flow chart of the present invention.

Figure 2 is a schematic diagram of a full body digitalized chest section image first loaded in the present invention.

Figure 3 is a schematic view of the image after removing the image of the thoracic object of the present invention.

Fig. 4 is a schematic view showing the state of the breast base point of the present invention.

Fig. 5 is a schematic view showing the state in which the invention obtains a rough estimate point.

Figure 6 is a schematic diagram showing the state of the present invention for obtaining a reference circle and fitting a curve from the collected breast edge points.

The features of the present invention can be clearly understood by referring to the drawings and the detailed description of the embodiments.

As shown in the basic flowchart of the present invention, the nipple image segmentation method of the present invention basically comprises the following steps:

(a) Load a digitalized chest section image (as shown in Figure 2) Full body digitalized chest section image).

(b) Find the chest cavity belonging to the chest part in the image of the chest section The image of the object; in the implementation, the object image area including the breast and the chest object is cut out in the chest profile image by setting a threshold, and the possible height of the chest cavity is estimated, and then the object is The image area is image grouped to find the image of the chest object belonging to the chest part; and, after loading a digital chest section image, the background of the chest section image is first cleared by the operating software. Noise, and then find the image of the chest object belonging to the chest.

(c) Establish a chest line fitting curve to remove the image of the thoracic object in the image area of the object (as shown in Figure 3).

(d) A structural parameter of the image of the breast object belonging to the breast portion among the image regions of the object is obtained.

(e) obtaining the edge of the breast for one of the breasts for which the image analysis is to be performed; in practice, as shown in Figure 5, the edge of the breast is obtained for one of the breasts for which the image analysis is desired After that, the breast spots on both sides of the breast are further obtained, and the position of the center line L1 is calculated by the centroids of the two bright points, and the decidal baseline L2 is constructed by linear regression of the edge points of the cleavage, and then the intersection with the center line L1 is generated. At the intermediate point P1, the tangent point P2 of the breast is gradually searched from the intermediate point P1, and the edge point farthest from the intermediate point P1 and the tangent point P2 is found as the breast base point P3, and then the left breast base point and the right breast base point are cleared. The middle image data is used to obtain image data of a relatively complete and simple breast.

(f) obtaining a rough estimate of the nipple of the breast; in the step of obtaining the rough position of the nipple of the breast during the implementation, the breast is projected on the X and Y axes, and the position of the top and middle of the breast is roughly searched, and then According to the position of the top of the breast, in the predetermined pixel range, the image pixel looking for the brightest point of the nipple is the seed pixel and the grayscale value is recorded, thereby starting from the seed pixel, as long as the surrounding pixel grayscale value is The difference between the gray level values of the seed pixels and the data points within the preset value are retained, so that a nipple drop point block is obtained at the breast image, and then the centroid of the nipple drop point block is calculated, that is, the breast is The nipple is roughly estimated to be P4.

(g) collecting the breast edge point fitting curve centered on the rough point of the nipple, and then dividing the nipple image block outside the fitting curve; in the implementation, collecting the nipple roughing point as the center In the step of fitting the curve of the breast edge point, as shown in Fig. 6, a horizontal line L3 is drawn at the middle of the breast to find the middle edge points of the two breasts which are closest to the horizontal line L3 on both sides of the breast. P5, P6, and calculate the distance between the two breast midpoint points P5, P6 to the nipple rough estimate point P4, and the nipple rough estimate point P4 as the center with two breast mid-segment edge points P5, P6 to the nipple rough estimate point P4 The distance is a maximum circle C1 and a minimum circle C2 for the radius, and a radius of the largest circle C1 and the smallest circle C2 is drawn by the average distance between the two breast midpoint edge points P5 and P6 to the nipple rough evaluation point P4. The reference circle C3 in the middle is fitted to the curve L4 in an elliptical manner by the edge points in the range of the reference circle C3, so that the nipple image block outside the fitting curve L4 can be segmented.

Incidentally, the nipple image segmentation method disclosed in the present invention is used to improve the accuracy of setting the base point of the breast image, except that the chest profile can be cleared by the operating software before the threshold is set for the chest profile image. Background noise; and, after cutting the image area of the object containing the breast and thoracic objects in the image of the chest, and estimating the possible height of the chest, if the estimated height of the chest is not ideal, The threshold can be reset, the image area of the object containing the breast and thoracic objects is cut again in the image of the chest section, and the possible height of the chest cavity is re-estimated.

Compared with the conventional techniques, the nipple image segmentation method of the present invention can produce the following effects:

1. Quickly and accurately segment the breast image in the chest profile image Nipple image block.

2. Help to accurately calibrate the relative position of the tumor.

3. Can still bring more accurate test data and improve the cure rate of breast cancer.

In summary, the present invention provides a preferred and feasible nipple image segmentation method, and the invention patent application is provided according to the law; the technical content and technical features of the present invention have been disclosed above, but those skilled in the art may still be based on the present invention. The invention is to be construed as being limited and modified by the spirit of the invention. Therefore, the scope of the present invention should be construed as being limited by the scope of the appended claims

Claims (3)

A method for segmenting a nipple image, comprising the steps of: (a) loading a digitalized chest section image; (b) finding an image of a chest object belonging to a chest site in the chest section image; (c) establishing a chest Line fitting curve to remove the image of the chest object in the image area of the object; (d) obtaining the structural parameters of the image of the breast object belonging to the breast part in the image area of the object; (e) for the image analysis desired After a breast obtains the edge point of the breast, further obtains the breast highlights on both sides of the breast, and then calculates the position of the center line with the centroids of the two bright points, and then constructs the cleavage baseline by linear regression of the edge points of the cleavage, and then The intersection of the center line produces an intermediate point, and the tangent point of the breast is gradually searched from the middle point to the sides, and the edge point farthest from the midpoint and the tangent point is found as the breast base point, and then the left breast base point and the right breast base point are cleared. Imagery for obtaining a more complete and simple image of the breast; (f) obtaining a rough estimate of the nipple of the breast; (g) collecting the breast edge centered on the rough estimate of the nipple Point fitting curve, draw a horizontal line at the middle of the breast, find the two breast midpoint edges closest to the horizontal line on both sides of the breast, and calculate the midpoint edge points of the two breasts to the nipple rough estimate point The distance between the two breasts from the midpoint of the breast to the rough point of the nipple is the maximum circle and a minimum circle, and the edge points of the two breasts to the nipple rough point. Draw a reference circle between the largest circle and the smallest circle from the average radius, and then fit the curve in an elliptical manner with the edge points in the reference circle to segment the nipple image block outside the fitting curve. . The nipple image segmentation method according to claim 1, wherein the nipple image segmentation method is configured to perform the breast pair X and the step of obtaining the breast nipple rough estimation position The Y-axis is projected, roughly searching for the position of the top and middle of the breast, and according to the position of the top of the breast, in the predetermined pixel range, the image pixel of the glare of the nipple is searched for as a seed pixel and the grayscale value is recorded, thereby the seed The pixel is the start, as long as the difference between the surrounding pixel grayscale value and the seed pixel grayscale value is retained within the preset value, so that a nipple drop point is obtained at the breast image, and then the nipple is calculated. The centroid of the falling block is the rough estimate of the nipple of the breast. The nipple image segmentation method according to claim 1 or 2, wherein the nipple image segmentation method, after loading a digitalized chest profile image, is first performed by the operating software to first clear the background image of the chest profile image. Then, find out the image of the chest object belonging to the chest.