Thyroid gland detection device
By using adaptive median filters and Laplace filters in the thyroid gland detection device to process ultrasound images, combined with segmentation and sharding area detection techniques, the challenge of identifying thyroid nodules in ultrasound images is solved, and the challenge of realizing the More accurate diagnosis of thyroid cancer.
Patent Information
- Application Number
- CN202111181844.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-10-11
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2041-10-11
AI Technical Summary
The prior art has challenges in identifying thyroid nodules and detecting tumor signs in ultrasound images, relying on subjective analysis by physicians, and lacks automatic and accurate quantitative analysis methods to improve the diagnostic performance of thyroid cancer.
A thyroid gland detection device is designed, including an image acquisition module, an interest nodule image extraction module and a gland detection module, to remove spot noise through an adaptive median filter, and Laplace filter enhances image contrast and edges , and thyroid nodules were detected by segmentation and fragmentation areas.
Effectively remove spot noise in ultrasound images, enhance image quality, and facilitate doctors to more accurately judge the pathological characteristics of thyroid nodules, and improve the diagnostic accuracy of thyroid cancer.
Smart Images

Figure CN113940703B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of auxiliary medical diagnosis, and in particular to a thyroid gland detection device. Background Art
[0002] Thyroid cancer is one of the most common cancers in women worldwide, with an incidence rate three times higher in women than in men. In 2018, one in every 20 cancer diagnoses in women was thyroid cancer. Ultrasound imaging is a non-invasive, radiation-free, and low-cost tumor diagnostic technology. However, due to the low quality of ultrasound images, identifying thyroid nodules and detecting signs of tumors from ultrasound images is a challenging task.
[0003] Generally, the growth and development of malignant tumors can be reflected by their shape, appearance, texture, composition, and many other factors. As an effective tool, grayscale ultrasound (US) images can visualize these factors and help doctors better observe and understand the manifestations of pathological sites. However, in current clinical practice, the reading of ultrasound image features greatly depends on the subjective or semi-subjective analysis of images by ultrasound physicians, which limits the widespread application of ultrasound. Therefore, automatic and accurate nodule quantitative analysis is crucial for the accurate diagnosis of cancer.
[0004] The Thyroid Imaging Reporting and Data System (TI-RADS) is a scientific guide for measuring and reporting thyroid and breast nodules. However, there is no research on quantitative feature analysis of TI-RADS to improve the diagnostic performance of thyroid cancer classification. Summary of the invention
[0005] The technical problem to be solved by the present invention is to provide a thyroid gland detection device which can effectively detect the thyroid gland.
[0006] The technical solution adopted by the present invention to solve the technical problem is: to provide a thyroid gland detection device, comprising:
[0007] Image acquisition module: used to acquire ultrasound images with thyroid nodules;
[0008] The module for extracting the image of a nodule of interest is used to intercept the boundary of the thyroid nodule in the ultrasound image by selecting a number of coordinate points of interest to obtain an image of the nodule of interest;
[0009] Thyroid gland detection module: used to perform gland detection on the ultrasound image based on the nodules in the nodule image of interest.
[0010] It also includes an ultrasonic image preprocessing module: used to filter the ultrasonic image through an adaptive median filter with different window sizes to remove speckle noise in the ultrasonic image; and then filter the ultrasonic image after the speckle noise is removed through a Laplace filter to enhance the contrast and edge of the ultrasonic image.
[0011] The window size expression of the adaptive median filter in the ultrasonic image preprocessing module is: in, Denotes the maximum window size of the adaptive median filter, D min Indicates the minimum size of an ultrasound image;
[0012] When the minimum size of the ultrasound image is D min ≥250, the maximum window size F of the adaptive median filter Smax =13×13;
[0013] When the minimum size of the ultrasound image is 120 ≤ D min <250, the maximum window size of the adaptive median filter
[0014] When the minimum size of the ultrasound image is 0<D min <120, the maximum window size of the adaptive median filter
[0015] The expression of the Laplace filter in the ultrasonic image preprocessing module is: in, represents the remapping function of signal i, g represents the coefficient of the Gaussian pyramid where signal i is located, signal i represents the brightness information of the image, α is used to control the smoothness of the details, and σ is used to control the enhancement and attenuation of the edge.
[0016] The thyroid gland detection module uses the nodule in the nodule image of interest as a reference, and before performing gland detection on the ultrasound image, it includes: constructing a strip area outside the boundary of the thyroid nodule, and then dividing the strip area into R sub-strip areas according to a preset segmentation angle based on the center of the thyroid nodule, and dividing each sub-strip area into n sub-slice areas.
[0017] The thyroid gland detection module uses the nodules in the nodule image of interest as a reference to perform gland detection on the ultrasound image. If the i-th sub-band area R i satisfy Then it represents the sub-band area R i is a gland, wherein a1 represents a first preset value, a2 represents a second preset value, and a3 represents a third preset value. Indicates a logical AND operation from j=1 to n. represents the i-th sub-strip region R i The set of pixels in the jth sub-slice area in , express The average brightness of all pixels in the set and I x,y Represents the brightness value of the ultrasonic image I at the pixel (x, y).
[0018] Beneficial Effects
[0019] Due to the adoption of the above-mentioned technical scheme, the present invention has the following advantages and positive effects compared with the prior art: the present invention removes speckle noise of the ultrasound image through an adaptive median filter, and then enhances the ultrasound image through a Laplace filter, which is convenient for subsequent analysis; the present invention can effectively detect the thyroid gland, making it convenient for doctors to make better, faster and more accurate judgments on the relevant pathological characteristics of thyroid nodules. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 is a flow chart of an embodiment of the present invention;
[0021] Figure 2 is a comparison diagram of the ultrasonic image before and after preprocessing according to an embodiment of the present invention;
[0022] Figure 3 is a schematic diagram of dividing a strip region into R sub-strip regions according to an embodiment of the present invention;
[0023] Figure 4 Schematic diagram of gland detection results according to an embodiment of the present invention. DETAILED DESCRIPTION
[0024] The present invention will be further described below in conjunction with specific embodiments. It should be understood that these embodiments are only used to illustrate the present invention and are not intended to limit the scope of the present invention. In addition, it should be understood that after reading the content taught by the present invention, those skilled in the art can make various changes or modifications to the present invention, and these equivalent forms fall within the scope limited by the appended claims of the application equally.
[0025] The embodiment of the present invention relates to a thyroid gland detection device, see Figure 1 ,include:
[0026] Image acquisition module: used to acquire ultrasound images with thyroid nodules;
[0027] Ultrasonic image preprocessing module: used to filter the ultrasonic image through an adaptive median filter to remove speckle noise in the ultrasonic image; and then filter the ultrasonic image after the speckle noise is removed through a Laplace filter to enhance the contrast and edge of the ultrasonic image;
[0028] The module for extracting the image of a nodule of interest is used to intercept the boundary of the thyroid nodule in the ultrasound image by selecting a number of coordinate points of interest to obtain an image of the nodule of interest;
[0029] Thyroid gland detection module: used to perform gland detection on the ultrasound image based on the nodules in the nodule image of interest.
[0030] The present invention is described in detail below through a specific embodiment:
[0031] Speckle noise is a particle-like interference inherent in medical ultrasound images, which can cause deviations from the expected signal measured. The presence of speckle noise can greatly reduce the quality of ultrasound images and easily mislead further analysis. Therefore, in order to make image analysis more accurate, it is necessary to reduce the impact of speckle noise.
[0032] In the ultrasound image preprocessing module, this embodiment uses an adaptive median filter to preprocess the ultrasound image to eliminate the interference of speckle noise. The adaptive median filter can effectively reduce isolated bright or dark pixels (i.e., noise) in the image. Its working principle is to compare each pixel in the image with its neighboring pixels (within a predefined window). If the central pixel value is significantly different from the adjacent pixels, the central pixel value is regarded as noise, and then the central pixel value is replaced with the median value of the pixels in the neighborhood. Otherwise, the adaptive median filter will search in a larger window until the maximum window size is reached. This process is repeated until all noise pixels are removed from the image. This is different from traditional methods, which usually predefine the window to a fixed value, resulting in poor denoising results. Considering that thyroid nodules vary in size, the window needs to be adjusted to maximize filtering performance. The specific definitions are as follows:
[0033]
[0034] Among them, D min Indicates the minimum size of the ultrasound image; when the minimum size of the nodule of interest is D min ≥250, the maximum window size of the adaptive median filter When the minimum size of the ultrasound image is 120 ≤ D min <250, the maximum window size of the adaptive median filter When the minimum size of the ultrasound image is 0<D min <120, the maximum window size of the adaptive median filter
[0035] Although the adaptive median filter can significantly reduce the speckle noise contained in the image, the application of the adaptive median filter will distort the local information in the image. Therefore, the present embodiment further applies a Laplacian filter to the median filtered image to restore the potential loss in the image. The Laplacian filter is an edge-aware image filter based on direct correction of the Laplacian pyramid, which can greatly refine the image content without degrading the edge or generating halos. The Laplacian filter contains two basic parameters, one for controlling the enhancement and attenuation of the edge, expressed as σ, and the other for controlling the smoothing of the details, expressed as α, and the formula is as follows:
[0036]
[0037] in, represents the remapping function of signal i, g represents the coefficient of the Gaussian pyramid where signal i is located, signal i represents the brightness information of the image, α is used to control the smoothness of details, and σ is used to control the enhancement and attenuation of edges. Through experiments, it is found that the Laplace filter produces the best effect when σ = 1.2 and α = 0.8.
[0038] This embodiment combines an adaptive median filter and a Laplace filter to preprocess the ultrasound image. Figure 2 (a) in the figure represents the image before preprocessing. Figure 2 (b) in the figure represents the preprocessed image I filt It is not difficult to see that through preprocessing, this embodiment significantly reduces the speckle noise in the image while retaining the local texture information of the image, which is of great benefit to the further analysis of the ultrasound image.
[0039] In the thyroid gland detection module, the thyroid gland is a large group of cells that secrete hormones and is commonly found in the thyroid gland. The gland is often used as an isoechoic reference in echo analysis. Since thyroid nodules often grow on the gland, a reference for the gland can be obtained from the area around the outside of the nodule.
[0040] See also Figure 3 In this embodiment, when searching for the reference of the gland, a strip region (composed of LB and OL) is first constructed outside the boundary LB of the thyroid nodule, and then the strip region is divided into R sub-strip regions. For simplicity, this embodiment regards the strip region as a circular ring, and divides the strip region 360 degrees based on the center of the nodule. The segmentation formula is as follows:
[0041]
[0042] Where θ represents the splitting angle, Figure 3The segmentation angle in is 45 degrees, dividing the nodule into 8 sub-band regions, which can provide a more robust and accurate estimation of the appearance of local echoes.
[0043] When actually detecting the gland, this embodiment divides the band area into 36 sub-band areas, that is, |R|=36. When studying the appearance of the nodule, this embodiment specifically constructs 5 continuous sub-slice areas with a width of 5 pixels in the sub-band area (the sub-slice areas are gradually constructed outward along the edge of the nodule), and these sub-slice areas cover a total of 1-25 pixels outside the thyroid nodule. This embodiment represents the sub-band area around the nodule as R i (i=1,...,36), and define the sub-strip region R i The set of pixels in the jth sub-slice area is If the sub-band region R i If the following two conditions are met at the same time, the sub-band region R i is the effective area of the gland reference, and the formula is as follows:
[0044]
[0045] in, Indicates logical AND operation from j=1 to 5, represents the i-th sub-strip region R i The set of pixels in the jth sub-slice area in , express The average brightness of all pixels in the set and I x,y Represents the brightness value of the ultrasonic image I at the pixel (x, y).
[0046] See also Figure 4 , Figure 4 The white circled part in (a) is the nodule. Figure 4 In (b), several approximately square areas represent the detected glands.
[0047] After checking all sub-band areas, the median brightness of each valid sub-band area is calculated. The median of these median brightness values is finally used as the reference value for detecting glands and is regarded as the reference of isoechoicity, denoted as G Ref .
[0048] It can be seen that the present invention can effectively detect the thyroid gland, making it easier for doctors to make better, faster and more accurate judgments on the relevant pathological characteristics of thyroid nodules.
Claims
1. A thyroid gland detection device, characterized in that: include: Image acquisition module: used to acquire ultrasound images with thyroid nodules; The module for extracting the image of a nodule of interest is used to intercept the boundary of the thyroid nodule in the ultrasound image by selecting a number of coordinate points of interest to obtain an image of the nodule of interest; Thyroid gland detection module: used to perform gland detection on the ultrasound image based on the nodule in the nodule image of interest; before performing gland detection on the ultrasound image based on the nodule in the nodule image of interest, a strip area is constructed outside the boundary of the thyroid nodule, and then the strip area is divided into R sub-strip areas according to a preset segmentation angle based on the center of the thyroid nodule, and each sub-strip area is divided into n sub-slice areas; when performing gland detection on the ultrasound image based on the nodule in the nodule image of interest, if the i-th sub-strip area R i satisfy Then it represents the sub-band area R i is a gland, wherein a1 represents a first preset value, a2 represents a second preset value, and a3 represents a third preset value. Indicates logical AND operation from j = 1 to n, sec Ri,j represents the i-th sub-strip region R i The set of pixels in the jth sub-slice area in , Indicates sec Ri,j The average brightness of all pixels in the set and I x,y Represents the brightness value of the ultrasonic image I at the pixel (x, y).
2. The thyroid gland detection device according to claim 1, characterized in that: It also includes an ultrasonic image preprocessing module: used to filter the ultrasonic image through an adaptive median filter with different window sizes to remove speckle noise in the ultrasonic image; and then filter the ultrasonic image after the speckle noise is removed through a Laplace filter to enhance the contrast and edge of the ultrasonic image.
3. The thyroid gland detection device according to claim 2, characterized in that: The window size expression of the adaptive median filter in the ultrasonic image preprocessing module is: in, Denotes the maximum window size of the adaptive median filter, D min Indicates the minimum size of an ultrasound image; When the minimum size of the ultrasound image is D min ≥250, the maximum window size of the adaptive median filter When the minimum size of the ultrasound image is 120 ≤ D min <250, the maximum window size of the adaptive median filter When the minimum size of the ultrasound image is 0<D min <120, the maximum window size of the adaptive median filter 4. The thyroid gland detection device according to claim 2, characterized in that: The expression of the Laplace filter in the ultrasonic image preprocessing module is: in, represents the remapping function of signal i, g represents the coefficient of the Gaussian pyramid where signal i is located, signal i represents the brightness information of the image, α is used to control the smoothness of the details, and σ is used to control the enhancement and attenuation of the edge.
