WO2021169452A1 - 甲状腺结节图像的处理方法、装置和计算机设备 - Google Patents

甲状腺结节图像的处理方法、装置和计算机设备 Download PDF

Info

Publication number
WO2021169452A1
WO2021169452A1 PCT/CN2020/132595 CN2020132595W WO2021169452A1 WO 2021169452 A1 WO2021169452 A1 WO 2021169452A1 CN 2020132595 W CN2020132595 W CN 2020132595W WO 2021169452 A1 WO2021169452 A1 WO 2021169452A1
Authority
WO
WIPO (PCT)
Prior art keywords
area
nodule
image
cystic
thyroid
Prior art date
Application number
PCT/CN2020/132595
Other languages
English (en)
French (fr)
Inventor
陈超
卢沁阳
张璐
詹维伟
黄凌云
刘玉宇
Original Assignee
平安科技(深圳)有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 平安科技(深圳)有限公司 filed Critical 平安科技(深圳)有限公司
Publication of WO2021169452A1 publication Critical patent/WO2021169452A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/0002Inspection of images, e.g. flaw detection
    • G06T7/0012Biomedical image inspection
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/11Region-based segmentation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/10Image acquisition modality
    • G06T2207/10132Ultrasound image
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20024Filtering details
    • G06T2207/20032Median filtering
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20084Artificial neural networks [ANN]
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/20Special algorithmic details
    • G06T2207/20112Image segmentation details
    • G06T2207/20132Image cropping
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00Indexing scheme for image analysis or image enhancement
    • G06T2207/30Subject of image; Context of image processing
    • G06T2207/30004Biomedical image processing

Definitions

  • This application relates to the field of artificial intelligence technology, and in particular to a method, device and computer equipment for processing images of thyroid nodules.
  • Thyroid nodule is a very common clinical disease of the thyroid. It is a mass that exists in the thyroid gland and can be benign or malignant. If a patient's thyroid nodule is diagnosed as malignant, then he has a high probability of developing thyroid cancer. Ultrasound imaging has become the preferred method of thyroid examination due to its low price, risk-free, and simple use. In ultrasound images, it is possible to find thyroid nodules and determine the nature of thyroid nodules based on the strength of the echo (that is, the gray value in the image). The composition of thyroid nodules is an important indicator of the thyroid imaging report and data system. It is divided into solid, solid-based, cystic-based, and cystic. When the ultrasound image of the thyroid nodule is generated, if the composition of the thyroid nodule can be quickly determined, it will be of great help in screening for malignant signs of the thyroid nodule.
  • the existing methods for judging the composition of thyroid nodules usually use the gray value as the judgment standard, and determine whether the thyroid nodule is cystic or solid by judging the gray distribution of the ultrasound image.
  • the HI value is used for judgment (the difference between the gray standard deviation of all points in the thyroid nodule and the mean value).
  • the probability that the thyroid nodule is solid is higher.
  • the inventor realizes that this composition judgment method only uses the gray value as the criterion for identifying the composition of thyroid nodules, and does not take into account other factors related to the composition of thyroid nodules, which may easily cause larger identification errors. As a result, the recognition accuracy of thyroid nodules is low.
  • the main purpose of this application is to provide a processing method, device, computer equipment and storage medium for thyroid nodule images, aiming to solve the technical problem of low recognition accuracy in the existing methods for judging the composition of thyroid nodules.
  • This application proposes a method for processing images of thyroid nodules, including:
  • a rough classification result of the composition of the thyroid nodule is generated according to a first preset rule, wherein the rough classification result of the composition is mainly cystic Or cystic, or the result of the rough classification of the composition is mainly substantial or substantial;
  • the result of the sub-classification of the composition includes: cystic, cystic, substantial, or substantial.
  • This application also provides an image processing device for thyroid nodules, including:
  • the first acquisition module is configured to acquire an input ultrasound image with thyroid nodules, mark the nodule area on the ultrasound image, and generate a nodule mask image corresponding to the nodule area;
  • the first generating module is used to intercept the image corresponding to the smallest circumscribed rectangle of the nodule mask image to obtain the nodule image;
  • the first processing module is configured to perform binary segmentation processing on the nodule image according to the preset OTSU segmentation algorithm, generate a processed first nodule image, and obtain sound and shadow candidates from the first nodule image area;
  • the first determining module is configured to determine the acoustic shadow area from the first nodule image according to the non-acoustic shadow part in the acoustic shadow candidate area and the nodule boundary of the nodule area;
  • the second processing module is used to crop the sound and shadow area in the nodule image to obtain a corresponding second nodule image, and perform binary segmentation on the second nodule image according to the OTSU segmentation algorithm Processing to obtain the processed third nodule image;
  • the second acquisition module is configured to acquire the black partial area inside the nodule area in the third nodule image, and use the black partial area as a cystic candidate area;
  • a calculation module configured to calculate the first area of the cystic candidate area, the second area of the nodule area, and the third area corresponding to the intersection of the nodule area and the acoustic shadow area;
  • the second generation module is configured to generate a rough classification result of the composition of the thyroid nodule according to the first preset rule according to the first area, the second area, and the third area, wherein the composition is rough
  • the classification result is mainly cystic or cystic, or the constituted rough classification result is mainly solid or solid;
  • a second determining module configured to determine a cystic area from the third nodule image according to the gradient of the third nodule image and the cystic candidate area;
  • the third generation module is configured to call the preset calculation rule corresponding to the result of the rough classification according to the fourth area, the second area and the third area of the cystic region to calculate and generate the thyroid nodule
  • the sub-classification results of the composition of the section wherein the sub-classification results of the composition include: cystic, cystic, substantial, or substantial.
  • the present application also provides a computer device, including a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, a method for processing an image of a thyroid nodule is realized, wherein the thyroid gland
  • the nodule image processing method includes the following steps:
  • a rough classification result of the composition of the thyroid nodule is generated according to a first preset rule, wherein the rough classification result of the composition is mainly cystic Or cystic, or the result of the rough classification of the composition is mainly substantial or substantial;
  • the result of the sub-classification of the composition includes: cystic, cystic, substantial, or substantial.
  • the present application also provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, a method for processing an image of a thyroid nodule is realized, wherein the method for processing an image of the thyroid nodule It includes the following steps:
  • a rough classification result of the composition of the thyroid nodule is generated according to a first preset rule, wherein the rough classification result of the composition is mainly cystic Or cystic, or the result of the rough classification of the composition is mainly substantial or substantial;
  • the result of the sub-classification of the composition includes: cystic, cystic, substantial, or substantial.
  • the thyroid nodule image processing method, device, computer equipment and storage medium provided in this application can accurately identify the composition of the thyroid nodule. On the one hand, it can further generate an analysis report related to the composition of the thyroid nodule. On the other hand, the composition of the thyroid nodule can also be input into the corresponding neural network model to make a judgment about the benign and malignant thyroid nodule.
  • FIG. 1 is a schematic flowchart of a method for processing an image of a thyroid nodule according to an embodiment of the present application
  • FIG. 2 is a schematic structural diagram of a thyroid nodule image processing device according to an embodiment of the present application.
  • Fig. 3 is a schematic structural diagram of a computer device according to an embodiment of the present application.
  • This solution can be applied to the digital medical field in smart cities, thereby promoting the construction of smart cities.
  • a method for processing an image of a thyroid nodule includes:
  • S1 Obtain an input ultrasound image with thyroid nodules, mark the nodule area on the ultrasound image, and generate a nodule mask image corresponding to the nodule area;
  • S3 Perform binary segmentation processing on the nodule image according to a preset OTSU segmentation algorithm, generate a processed first nodule image, and obtain a sound and shadow candidate region from the first nodule image;
  • S4 Determine an acoustic shadow area from the first nodule image according to the non-acoustic shadow part in the acoustic shadow candidate area and the nodule boundary of the nodule area;
  • S6 Obtain a black partial area inside the nodule area in the third nodule image, and use the black partial area as a cystic candidate area;
  • S7 Calculate the first area of the cystic candidate area, the second area of the nodule area, and the third area corresponding to the intersection of the nodule area and the acoustic shadow area;
  • S9 Determine a cystic area from the third nodule image according to the gradient of the third nodule image and the cystic candidate area;
  • the sub-classification results of the composition include: cystic, cystic, solid, or solid.
  • the execution subject of this method embodiment is a thyroid nodule image processing device.
  • the above-mentioned thyroid nodule image processing device can be realized by a virtual device, such as software code, or by a physical device written or integrated with relevant execution codes, and can communicate with the user through a keyboard, mouse, and remote control. , Touchpad or voice control device for human-computer interaction.
  • the thyroid nodule image processing device in this embodiment can quickly and accurately identify the composition of the thyroid nodule in the ultrasound image. Specifically, firstly, an input ultrasound image with thyroid nodules is acquired, the nodule area is marked on the ultrasound image, and a nodule mask image corresponding to the nodule area is generated.
  • the image corresponding to the smallest circumscribed rectangle of the nodule mask image is intercepted to obtain the nodule image.
  • the OTSU segmentation algorithm is an adaptive threshold calculation method.
  • the original slice image is grayed out. According to the gray value of the image, all the pixels of the image are divided into two different categories: foreground, target and background. , The difference between the two classes is reflected in the variance between the classes.
  • the OTSU segmentation algorithm is to find the appropriate threshold by traversing all the gray values , Which maximizes the variance between the two classes.
  • the OTSU segmentation algorithm may be used to determine the optimal first threshold for image binarization corresponding to the first nodule image, and perform binary segmentation processing on the first nodule image according to the first threshold to obtain The first nodule image, and the acoustic shadow candidate area refers to the foreground (white) part of the first nodule image.
  • M is the length of the nodule image
  • N is the width of the nodule image
  • M*N is the area of the nodule
  • N 0 is the area of the foreground pixel
  • N 1 is the area of the background pixel.
  • a certain threshold T can be used to A nodule image is divided into two parts: foreground and background. The proportion of foreground pixels is ⁇ 0 , and the proportion of background pixels is ⁇ 1.
  • the average gray values of foreground and background are ⁇ 0 , ⁇ 1 , Finally, the between-class variance ⁇ is obtained.
  • different thresholds T are calculated to find the threshold T that minimizes the variance ⁇ between classes. This is the result of the OTSU segmentation algorithm.
  • the final acoustic shadow area is determined from the first nodule image based on the non-acoustic shadow part in the acoustic shadow candidate area and the nodule boundary of the nodule area.
  • the physical characteristics of the sound and shadow it is possible to filter out the sound and shadow candidate areas that are not part of the sound and shadow to be eliminated, and then to eliminate the first connection that passes through the nodule boundary of the nodule area in the sound and shadow candidate area after the elimination processing.
  • the domain is the final sound and shadow area mentioned above. After that, the final sound and shadow area is cropped from the nodule image to obtain the corresponding second nodule image, and the second nodule image is subjected to binary segmentation processing according to the OTSU segmentation algorithm to obtain the processed third nodule image. Nodule image.
  • the process of performing binary segmentation on the second nodule image according to the above-mentioned OTSU segmentation algorithm can refer to the process of performing binary segmentation on the above-mentioned nodule image according to the OTSU segmentation algorithm.
  • a rough classification result of the composition of the thyroid nodules is generated according to a first preset rule, wherein the rough classification result of the composition is predominantly cystic or cystic, Or the result of the above-mentioned rough classification of the composition is mainly substantial or substantial.
  • the size of the designated nodule area can be compared by comparing the area of the cystic candidate area with the area of the nodule minus the area of the acoustic shadow area included in the nodule, and then compare the results according to the corresponding size. To generate a rough classification result corresponding to the structure of the thyroid nodule. Then, according to the gradient of the third nodule image and the cystic candidate area, the final cystic area is determined from the third nodule image.
  • the gray scale and gradient of the nodule image can be combined to accurately identify the cystic region in the nodule image, and the gradient map of the third nodule image can be generated, and the seed point that meets the preset conditions can be determined from the gradient map.
  • the sub-category results of the above composition include: cystic, cystic, substantial, or substantial.
  • two different calculation rules used to generate the detailed classification results will be correspondingly set in advance, and the two different calculation rules will have corresponding two ratio thresholds.
  • This embodiment includes a two-step recognition process for the formation of thyroid nodules from coarse to fine. In the coarse recognition process, the physical characteristics of the acoustic shadow during thyroid ultrasound scanning are first used to extract the acoustic shadow area in the nodule image and remove it.
  • the size of the nodule area So that the acoustic shadow area in the nodule image will not be misjudged as a cystic area, so according to the first area of the candidate cystic area in the nodule image after the acoustic shadow area is removed, the size of the nodule area The second area and the third area corresponding to the intersection of the nodule area and the acoustic shadow area are used to accurately generate a rough classification result of the thyroid nodules.
  • the gray scale and gradient information of the nodule image are combined to determine the cystic area in the nodule image, so that the fourth area of the cystic area, the second area and the third area are called
  • the preset calculation rules corresponding to the above-mentioned rough classification results are used to quickly and accurately calculate and generate the fine classification results of the thyroid nodules.
  • it is possible to accurately identify the composition of the thyroid nodule.
  • it can further generate an analysis report related to the composition of the thyroid nodule, and on the other hand, it can also input the composition of the thyroid nodule to the corresponding nerve.
  • Use the network model to make judgments about benign and malignant thyroid nodules.
  • This application can be applied to the fields of smart medical care and digital medical care through the optimization of medical image analysis.
  • the step S4 of determining the acoustic shadow area from the first nodule image based on the non-acoustic shadow part and the nodule boundary in the acoustic shadow candidate area includes:
  • S401 Perform elimination processing on the non-acoustic and shadow part to obtain a processed acoustic and shadow candidate area
  • S402 Acquire a first connected domain that passes through the nodule boundary of the nodule area in the processed acoustic shadow candidate area;
  • the step of determining the acoustic shadow area from the first nodule image based on the non-acoustic shadow part and the nodule boundary in the acoustic shadow candidate area may specifically include: first obtaining the above The non-acoustic shadow part in the acoustic shadow candidate area. Among them, since the starting point of the sound shadow must be inside the nodule, the sound shadow will continue a certain distance outside the boundary of the nodule, and the middle image of the sound shadow will not have the physical characteristics of a large number of faults.
  • the non-acoustic shadow part in the acoustic shadow candidate area can be filtered out by using a linear array probe to traverse each column in the first nodule image from top to bottom. Then, the above-mentioned non-acoustic shadow part is eliminated, and the processed acoustical shadow candidate area is obtained. Wherein, by eliminating the above-mentioned non-acoustic shadow part in the acoustic shadow candidate area, the accuracy of the subsequently generated processed acoustic shadow candidate area can be improved. Then, the first connected domain that passes through the nodule boundary of the nodule area in the acoustic shadow candidate area after the processing is obtained.
  • the processed sound and shadow candidate area may be expanded in advance to eliminate the processed sound and shadow candidate area.
  • the fine points in the acoustic shadow candidate area can prevent the fine points from affecting the first connected domain obtained subsequently, and improve the accuracy of the obtained first connected domain.
  • the first connected domain is used as the final sound and shadow area, so that the final sound and shadow area can be cropped from the nodule image to obtain the corresponding second nodule image.
  • the acoustic shadow area in the second nodule image has been eliminated, that is, the acoustic shadow area in the nodule image will not be misjudged as a cystic area, and the second nodule image can be used to accurately determine The recognition process of the thyroid nodule structure is carried out.
  • the step S8 of generating a rough classification result of the thyroid nodules according to the first area, the second area, and the third area according to the first preset rule include:
  • the step of generating a rough classification result of the thyroid nodules according to the first preset rule according to the first area, the second area, and the third area may specifically include: first Calculate the first difference of the second area minus the third area.
  • the second area is the area of the nodule area
  • the third area is the area corresponding to the intersection of the nodule area and the final acoustic shadow area.
  • the difference is to calculate the area of the nodule minus the area of the acoustic shadow area included in the nodule, so as to obtain an accurate designated nodule area. Then calculate the quotient obtained by dividing the above-mentioned first difference by 2.
  • the first area is the area of the cystic candidate area in the third nodule image.
  • the composition of the thyroid gland can be preliminarily determined Mainly cystic or cystic. If it is determined that the first area is not greater than the quotient value, then it is determined that the result of the rough classification of the composition of the thyroid nodules is mainly solid or solid.
  • the composition of the thyroid gland is mainly solid or solid.
  • This embodiment compares the area of the cystic candidate area with the area of the nodule minus the area of the acoustic shadow area included in the nodule, and compares the size of the designated nodule area, and then compares the results according to the corresponding size. To generate a rough classification result of the composition corresponding to the structural composition of the thyroid nodule, which is beneficial to further generate a fine classification result of the composition of the thyroid nodule based on the rough classification result of the structure.
  • the step S9 of determining the cystic area from the third nodule image according to the gradient of the third nodule image and the cystic candidate area includes:
  • S901 Search for seed points in the gradient map whose grayscale is less than a first preset threshold, and the gradient is less than a second preset threshold, and are located in the cystic candidate region;
  • S902 Perform region growth processing on the seed point according to a second preset rule to generate a corresponding designated connected domain
  • S906 Determine the second connected domain as the cystic area.
  • this embodiment is not limited to only using the gray information of the nodule image, but will combine the gray and gradient of the nodule image to accurately identify the cystic area in the nodule image .
  • the step of determining the cystic area from the third nodule image according to the gradient of the third nodule image and the cystic candidate area may include: first calculating the gradient of the third nodule image, and generating The corresponding gradient map. Among them, you can use the formula Calculate and generate a gradient map corresponding to the third nodule image, g x and g y are edge detection operators respectively, and calculate the gradient map g in the horizontal and vertical directions for the third nodule image.
  • the seed points in the above-mentioned gradient map whose gray scale is smaller than the first preset threshold value, the gradient is smaller than the second preset threshold value, and are located in the above-mentioned cystic candidate region are found.
  • the above-mentioned first preset threshold and second preset threshold are not specifically limited, and can be set according to actual needs, for example, they can be set to 2 and 4 respectively.
  • a region growth process is performed on the seed point according to a second preset rule to generate a corresponding designated connected domain.
  • the above-mentioned second preset rule may specifically be a seed filling method.
  • the process of the seed filling method includes: selecting a foreground pixel as a seed point, and then combining foreground pixels adjacent to the seed into the same In a set, the final set of pixels is a connected region.
  • the above-mentioned designated connected domain is obtained, the above-mentioned designated connected domain is corroded. Among them, by corroding the designated connected domain, the hole can be filled and the connection part with the non-nodular area in the designated connected domain can be removed to realize the refinement of the shape of the thyroid nodule. Then calculate the product of the second area and the specified value. Among them, there is no specific limitation on the above specified value, and it can be set according to actual needs, for example, it can be set to 0.001.
  • the second connected component whose connected component area is larger than the above product is selected.
  • the connected domains with too small area they will be regarded as meaningless regions.
  • the meaningless regions need to be eliminated from the cystic candidate regions.
  • the above-mentioned second connected domain is determined as the final above-mentioned cystic region, so that the final fourth area of the above-mentioned cystic region, as well as the above-mentioned second area and the above-mentioned third area can be called and roughly classified according to the above-mentioned composition.
  • the results correspond to the preset calculation rules to quickly and accurately calculate and generate the sub-classification results of the thyroid nodules.
  • the step S902 of performing region growth processing on the seed point according to the third preset rule to generate a corresponding designated connected domain includes:
  • S9020 Obtain a designated seed point, where the designated seed point is any one of all the seed points;
  • S9021 Combine the pixels adjacent to the specified seed point into the same set according to the constituent conditions of the connected area to generate a specified pixel set, where the constituent conditions of the connected area include that the pixel values are the same and the pixel values are adjacent;
  • the above step of performing region growth processing on the seed point according to the third preset rule to generate the corresponding designated connected domain may specifically include: first obtaining the designated seed point, wherein the designated seed point The point is any one of the above-mentioned seed points. Then, according to the constituent conditions of the connected area, the pixels adjacent to the specified seed point are merged into the same set to generate a specified pixel set; wherein the constituent conditions of the connected area include the same pixel value and adjacent pixel values. Finally, the generated specified set is used as the connected domain corresponding to the specified seed point.
  • this embodiment can quickly and conveniently obtain the designated connected domain of the seed point after the growth and filling process, so that the final cysticity in the third nodule image can be quickly determined based on the designated connected domain.
  • Region and according to the obtained final fourth area of the cystic region, as well as the above-mentioned second area and the above-mentioned third area, call the preset calculation rules corresponding to the above-mentioned rough classification results to quickly and accurately calculate and generate thyroid nodules The composition of the sub-category results.
  • the method includes:
  • the gradient map can be further subjected to median filtering processing to filter out the useless noise data.
  • the gradient map includes: firstly filtering out the abnormal gradients in the gradient map through a median filter algorithm. After the abnormal gradient is obtained, the abnormal gradient is deleted from the gradient map.
  • the median filter algorithm is used to remove the abnormal gradients in the gradient map, which can improve the accuracy of the seed points to be found subsequently, and further improve the accuracy of the final cystic region generated subsequently.
  • the preset calculation rule corresponding to the result of the constituent rough classification is called to calculate and generate
  • the step S10 of the sub-classification result of the composition of the thyroid nodules includes:
  • the step of calling a preset calculation rule corresponding to the result of the rough classification of the composition to calculate and generate the result of the fine classification of the composition of the thyroid nodule may be specifically It includes: first calculating the second difference of the second area minus the third area. And calculate the ratio between the fourth area and the second difference.
  • the second area is the area of the nodule area
  • the third area is the area corresponding to the intersection of the nodule area and the final acoustic shadow area.
  • the first area is calculated by subtracting the third area from the second area. The difference is to calculate the area of the nodule minus the area of the acoustic shadow area included in the nodule, so as to obtain an accurate designated nodule area.
  • the ratio between the fourth area and the second difference refers to the ratio between the final area of the cystic region and the area of the designated nodule.
  • the above-mentioned first ratio threshold is not specifically limited, and can be set according to actual needs, for example, it can be set to 0.95. If it is determined that the ratio is greater than the first ratio threshold, it is determined that the sub-classification result of the thyroid nodule is cystic; otherwise, it is determined that the sub-classification result of the thyroid nodule is mainly cystic. When the result of the rough classification of the composition is dominant or substantial, it is determined whether the ratio is smaller than the preset second ratio threshold.
  • the above-mentioned second ratio threshold is not specifically limited, and can be set according to actual needs, for example, it can be set to 0.08.
  • the ratio is less than the second ratio threshold, it is judged that the result of the sub-classification of the composition of the thyroid nodule is solid; otherwise, it is judged that the result of the sub-classification of the composition of the thyroid nodule is mainly solid.
  • the ratio between the designated nodule area obtained by calculating the final cystic area and the area of the nodule minus the area of the acoustic shadow area included in the nodule And compare the ratio with the preset ratio threshold, so that the thyroid nodules can be quickly and accurately generated according to the corresponding size comparison results.
  • the ⁇ can further generate an analysis report related to the composition of the thyroid nodule, and on the other hand, it can also input the composition of the thyroid nodule to the corresponding
  • the neural network model is used to make judgments about benign and malignant thyroid nodules.
  • the image processing method of thyroid nodules in the embodiments of the present application can also be applied to the blockchain field, such as storing data such as the results of the detailed classification of the thyroid nodules and other data on the blockchain.
  • the blockchain By using the blockchain to store and manage the results of the detailed classification of the thyroid nodules, the safety and non-tampering of the results of the detailed classification of the thyroid nodules can be effectively guaranteed.
  • an embodiment of the present application also provides an image processing device for thyroid nodules, including:
  • the first acquisition module 1 is configured to acquire an input ultrasound image with thyroid nodules, mark the nodule area on the ultrasound image, and generate a nodule mask image corresponding to the nodule area;
  • the first generating module 2 is used to intercept the image corresponding to the smallest circumscribed rectangle of the nodule mask image to obtain the nodule image;
  • the first processing module 3 is configured to perform binary segmentation processing on the nodule image according to a preset OTSU segmentation algorithm, generate a processed first nodule image, and obtain sound and shadow from the first nodule image Candidate area
  • the first determining module 4 is configured to determine the acoustic shadow area from the first nodule image according to the non-acoustic shadow part in the acoustic shadow candidate area and the nodule boundary of the nodule area;
  • the second processing module 5 is configured to crop the sound and shadow area in the nodule image to obtain a corresponding second nodule image, and perform binary value on the second nodule image according to the OTSU segmentation algorithm Segmentation processing to obtain the processed third nodule image;
  • the second acquisition module 6 is configured to acquire the black partial area inside the nodule area in the third nodule image, and use the black partial area as a cystic candidate area;
  • the calculation module 7 is used to calculate the first area of the cystic candidate area, the second area of the nodule area, and the third area corresponding to the intersection of the nodule area and the acoustic shadow area;
  • the second generating module 8 is configured to generate a rough classification result of the composition of the thyroid nodule according to a first preset rule according to the first area, the second area, and the third area, wherein the composition
  • the rough classification result is mainly cystic or cystic, or the constituted rough classification result is mainly solid or solid;
  • the second determining module 9 is configured to determine the cystic area from the third nodule image according to the gradient of the third nodule image and the cystic candidate area;
  • the third generating module 10 is configured to calculate and generate the thyroid gland according to the fourth area, the second area, and the third area of the cystic region by calling a preset calculation rule corresponding to the result of the rough classification of the composition
  • the results of the sub-classification of nodules, wherein the sub-classification results of the nodules include: predominantly cystic, predominantly cystic, predominantly solid, or predominantly solid.
  • the implementation process of the functions and roles of the generation module, the second determination module and the third generation module refer to the implementation process corresponding to steps S1 to S10 in the above-mentioned thyroid nodule image processing method, which will not be repeated here.
  • the above-mentioned first determining module includes:
  • the first acquisition submodule is used to acquire the non-acoustic shadow part in the acoustic shadow candidate area
  • the first processing sub-module is used to perform elimination processing on the non-acoustic and shadow part to obtain the processed acoustic and shadow candidate area;
  • the second acquisition sub-module is configured to acquire the first connected domain that passes through the nodule boundary of the nodule area in the processed acoustic shadow candidate area;
  • the first determining sub-module is configured to use the first connected domain as the sound and shadow area.
  • the realization process of the functions and functions of the first acquisition sub-module, the first processing sub-module, the second acquisition sub-module, and the first determination sub-module in the above-mentioned thyroid nodule image processing device is detailed in the above-mentioned thyroid nodule image processing device.
  • the processing method of the nodule image corresponds to the implementation process of steps S400 to S403, which will not be repeated here.
  • the above-mentioned second generation module includes:
  • the first calculation sub-module is used to calculate the first difference of the second area minus the third area
  • the second calculation sub-module is used to calculate the quotient obtained by dividing the first difference by 2;
  • the first judging sub-module is used to judge whether the first area is greater than the quotient
  • the first determination sub-module is configured to determine whether the rough classification result of the composition of the thyroid nodule is predominantly cystic or cystic if the first area is greater than the quotient;
  • the second determination sub-module is configured to determine whether the rough classification result of the composition of the thyroid nodule is mainly solid or solid if the first area is not greater than the quotient value.
  • the functions and effects of the first calculation sub-module, the second calculation sub-module, the first judgment sub-module, the first judgment sub-module and the second judgment sub-module in the above-mentioned thyroid nodule image processing device are realized
  • the process refer to the implementation process corresponding to steps S800 to S804 in the foregoing thyroid nodule image processing method, which will not be repeated here.
  • the above-mentioned second determining module includes:
  • the third calculation sub-module is used to calculate the gradient of the third nodule image and generate a corresponding gradient map
  • the first processing sub-module is configured to perform region growth processing on the seed point according to a second preset rule to generate a corresponding designated connected domain;
  • the second processing sub-module is used to perform corrosion processing on the designated connected domain
  • the fourth calculation sub-module is used to calculate the product of the second area and a specified value
  • a screening sub-module which is used to screen out the second connected domains whose connected domain area is larger than the product from all the specified connected domains;
  • the determining sub-module is configured to determine the second connected domain as the cystic area.
  • the third calculation sub-module, the search sub-module, the first processing sub-module, the second processing sub-module, the fourth calculation sub-module, the screening sub-module, and the determination sub-module in the above-mentioned processing device for thyroid nodule images For the implementation process of the function and effect of the thyroid nodule, please refer to the implementation process of corresponding steps S900 to S906 in the above-mentioned thyroid nodule image processing method, which will not be repeated here.
  • the above-mentioned first processing submodule includes:
  • the generating unit is used to combine the pixels adjacent to the specified seed point into the same set according to the constituent condition of the connected area to generate a specified pixel set, wherein the constituent condition of the connected area includes the same pixel value and the pixel value Adjacent
  • the determining unit is configured to use the generated designated set as the connected domain corresponding to the designated seed point.
  • the realization process of the functions and functions of the acquisition unit, the generation unit, and the determination unit in the above-mentioned thyroid nodule image processing device is detailed in the realization process corresponding to steps S9020 to S9022 in the above-mentioned thyroid nodule image processing method. , I won’t repeat it here.
  • the above-mentioned second determining module includes:
  • a screening unit configured to screen out abnormal gradients in the gradient map through a median filtering algorithm
  • the deleting unit is used to delete the abnormal gradient from the gradient map.
  • the implementation process of the functions and effects of the screening unit and the deletion unit in the above-mentioned thyroid nodule image processing device is detailed in the implementation process corresponding to steps S9000 to S9001 in the above-mentioned thyroid nodule image processing method, here No longer.
  • the above-mentioned third generation module includes:
  • a fifth calculation sub-module configured to calculate a second difference of the second area minus the third area
  • a sixth calculation sub-module configured to calculate the ratio between the fourth area and the second difference
  • the second judgment sub-module is configured to judge whether the ratio is greater than a preset first ratio threshold when the rough classification result is cystic predominantly or cystic;
  • the third determination sub-module is configured to determine that the sub-classification result of the thyroid nodule is cystic if the ratio is greater than the first ratio threshold; otherwise, it is determined that the sub-classification result of the thyroid nodule is cystic.
  • the third judgment sub-module is configured to judge whether the ratio is less than a preset second ratio threshold when the rough classification result is substantial or substantial;
  • the fourth determination sub-module is used to determine that the sub-classification result of the thyroid nodule is solid if the ratio is less than the second ratio threshold; otherwise, it is determined that the sub-classification result of the thyroid nodule is solid. Sex-oriented.
  • the fifth calculation sub-module, the sixth calculation sub-module, the second judgment sub-module, the third judgment sub-module, the third judgment sub-module, and the fourth judgment sub-module in the apparatus for processing images of thyroid nodules For the implementation process of the functions and effects of, refer to the implementation process of corresponding steps S1000 to S1005 in the above-mentioned thyroid nodule image processing method, which will not be repeated here.
  • an embodiment of the present application also provides a computer device.
  • the computer device may be a server, and its internal structure may be as shown in FIG. 3.
  • the computer equipment includes a processor, a memory, a network interface, a display screen, an input device and a database connected by a system bus.
  • the processor designed for the computer equipment is used to provide computing and control capabilities.
  • the memory of the computer device includes a non-volatile storage medium and internal memory.
  • the non-volatile storage medium stores an operating system, a computer program, and a database.
  • the internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium.
  • the database of the computer equipment is used to store data such as nodule mask image, nodule image, nodule image, third nodule image, final cystic area, and thyroid nodule composition classification results.
  • the network interface of the computer device is used to communicate with an external terminal through a network connection.
  • the display screen of the computer equipment is an indispensable image and text output device in the computer, which is used to convert digital signals into optical signals so that text and graphics can be displayed on the screen of the display screen.
  • the input device of the computer equipment is the main device for information exchange between the computer and the user or other equipment, and is used to transfer data, instructions, and certain flag information to the computer.
  • the computer program is executed by the processor to realize a processing method of thyroid nodule images.
  • the above-mentioned processor executes the steps of the above-mentioned thyroid nodule image processing method:
  • a rough classification result of the composition of the thyroid nodule is generated according to a first preset rule, wherein the rough classification result of the composition is mainly cystic Or cystic, or the result of the rough classification of the composition is mainly substantial or substantial;
  • the result of the sub-classification of the composition includes: cystic, cystic, substantial, or substantial.
  • FIG. 3 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the devices and computer equipment to which the solution of the present application is applied.
  • An embodiment of the present application also provides a computer-readable storage medium.
  • the computer-readable storage medium may be non-volatile or volatile, and has a computer program stored thereon, which is realized when the computer program is executed by a processor.
  • the method for processing an image of thyroid nodules includes the following steps:
  • a rough classification result of the composition of the thyroid nodule is generated according to a first preset rule, wherein the rough classification result of the composition is mainly cystic Or cystic, or the result of the rough classification of the composition is mainly substantial or substantial;
  • the result of the sub-classification of the composition includes: cystic predominantly, cystic, solid predominant or substantive.
  • This solution can be applied to the digital medical field in smart cities, thereby promoting the construction of smart cities.
  • the thyroid nodule image processing method, device, computer equipment, and storage medium provided in the embodiments of this application can accurately identify the composition of the thyroid nodule, and on the one hand, it can further generate an exact connection with the thyroid nodule. Analysis report related to the composition of the thyroid nodule. On the other hand, the composition of the thyroid nodule can also be input into the corresponding neural network model to make a judgment about the benign and malignant thyroid nodule.
  • Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
  • Volatile memory may include random access memory (RAM) or external cache memory.
  • RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-rate data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Ultra Sonic Daignosis Equipment (AREA)

Abstract

甲状腺结节图像的处理方法、装置和计算机设备,方法包括:根据超声图像生成结节掩膜图像;生成与结节掩膜图像的最小外接矩形对应的结节图像;对结节图像进行处理生成第一结节图像,获取声影候选区域;从第一结节图像确定声影区域;在结节图像中裁剪声影区域得到第二结节图像,对第二结节图像进行处理得到第三结节图像;获取第三结节图像的囊性候选区域;计算囊性候选区域的第一面积、结节区域的第二面积及两者相关的第三面积,生成甲状腺结节的构成粗分类结果;从第三结节图像确定囊性区域;根据囊性区域的面积、第二面积与第三面积,生成甲状腺结节的构成细分类结果,准确识别出甲状腺结节的构成。

Description

甲状腺结节图像的处理方法、装置和计算机设备
本申请要求于2020年09月27日提交中国专利局、申请号为2020110357237,发明名称为“甲状腺结节图像的处理方法、装置和计算机设备”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及人工智能技术领域,具体涉及一种甲状腺结节图像的处理方法、装置和计算机设备。
背景技术
甲状腺结节是一种十分常见的甲状腺临床病症,它是一种存在于甲状腺腺体之内的肿块,有良性、恶性之分。如果一个患者的甲状腺结节被诊断为恶性,那么他有极大的概率患上了甲状腺癌。超声图像由于其价格低、无风险、使用简单等优点,成为了甲状腺检查的首选方法。在超声图像中,能够很好地根据回声的强弱(即图像中的灰度值)来寻找甲状腺结节并判断甲状腺结节的性质。甲状腺结节的构成是甲状腺影像报告和数据系统的一项重要指标,其分为实性、实性为主、囊性为主、囊性。当在甲状腺结节的超声图像生成时,如果能够快速地判断甲状腺结节的构成,这将对筛查甲状腺结节的恶性征象产生极大的帮助。
然而,现有对于甲状腺结节的构成判断方法,通常是将灰度值作为判断标准,通过判断超声图像的灰度分布来初步判断甲状腺结节是囊性还是实性。例如会使用HI值进行判断(甲状腺结节内所有点的灰度标准差与均值之差),当HI值越大时,甲状腺结节为实性的概率越高。但发明人意识到,这种构成判断方法仅仅是将灰度值作为识别甲状腺结节构成的判断标准,而没有考虑到与甲状腺结节构成相关的其他因素,从而容易造成较大的识别误差,使得甲状腺结节构成的识别准确性较低。
技术问题
本申请的主要目的为提供一种甲状腺结节图像的处理方法、装置、计算机设备和存储介质,旨在解决现有的对于甲状腺结节的构成判断方法存在识别准确性较低的技术问题。
技术解决方案
本申请提出一种甲状腺结节图像的处理方法,包括:
获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
本申请还提供一种甲状腺结节图像的处理装置,包括:
第一获取模块,用于获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
第一生成模块,用于截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
第一处理模块,用于按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
第一确定模块,用于根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
第二处理模块,用于在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
第二获取模块,用于获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
计算模块,用于计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
第二生成模块,用于根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
第二确定模块,用于根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
第三生成模块,用于根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
本申请还提供一种计算机设备,包括存储器和处理器,所述存储器中存储有计算机程序,所述处理器执行所述计算机程序时实现一种甲状腺结节图像的处理方法,其中,所述甲状腺结节图像的处理方法包括以下步骤:
获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
本申请还提供一种计算机可读存储介质,其上存储有计算机程序,所述计算机程序被处理器执行时实现一种甲状腺结节图像的处理方法,其中,所述甲状腺结节图像的处理方法包括以下步骤:
获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构 成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
有益效果
本申请中提供的甲状腺结节图像的处理方法、装置、计算机设备和存储介质,能够实现准确地识别出甲状腺结节的构成,一方面可以进一步生成确切的与甲状腺结节的构成相关的分析报告,另一方面还可以将该甲状腺结节的构成输入至相应的神经网络模型中来作出关于甲状腺结节结节良恶性的判断。
附图说明
图1是本申请一实施例的甲状腺结节图像的处理方法的流程示意图;
图2是本申请一实施例的甲状腺结节图像的处理装置的结构示意图;
图3是本申请一实施例的计算机设备的结构示意图。
本发明的最佳实施方式
应当理解,此处所描述的具体实施例仅仅用于解释本申请,并不用于限定本申请。
本方案可应用于智慧城市中的数字医疗领域,从而推动智慧城市的建设。
参照图1,本申请一实施例的甲状腺结节图像的处理方法,包括:
S1:获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
S2:截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
S3:按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
S4:根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
S5:在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
S6:获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
S7:计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
S8:根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
S9:根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
S10:根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
如上述步骤S1至S10所述,本方法实施例的执行主体为一种甲状腺结节图像的处理装置。在实际应用中,上述甲状腺结节图像的处理装置可以通过虚拟装置,例如软件代码实现,也可以通过写入或集成有相关执行代码的实体装置实现,且可以与用户通过键盘、鼠标、遥控器、触摸板或声控设备等方式进行人机交互。本实施例中的甲状腺结节图像的处理装置能够快速准确地识别出超声图像中甲状腺结节的构成。具体地,首先获取输入的带有甲状腺结节的超声图像,对上述超声图像标记结节区域,生成对应上述结节区域的结节掩膜图像。然后截取上述结节掩膜图像的最小外接矩形对应的图像,得到结节图像。在得到上述结节图像后,再按照预设的OTSU分割算法对上述结节图像进行二值分割处理,生成处理后的第一结节图像,并从上述第一结节图像中获取声影候选区域。其中,OTSU分割算法是一种自适应的计算阈值的方法,首先将原始切片图像灰度化,根据图像的灰度值,将图像所有的像素点分为前景即目标和背景两个不同的类别,两个类之间的差别体现在类间方差,前景和背景的区别越大,这两个类的类间的方差越大,OTSU分割算法就是通过遍历全部的灰度值,找到合适的阈值,使得这两个类的类间的方差最大。另外,可先通过上述OTSU分割算法确定与上述第一结节图像对应的图像二值化最优的第一阈值,并根据该第一阈值对上述第一结节图像进行二值分割处理,得到上述第一结节图像,且上述声影候选区域是指上述第一结节图像中的前景(白色)部分。具体的,OTSU分割算法的公式如下: ω 0=N 0/(M*N),ω 1=N 1/(M*N),σ=ω 0ω 101) 2。M为结节图像的长,N为结节图像宽,M*N表示结节面积,N 0为前景像素点的面积,N 1为后景像素点的面积,通过某一阈值T可将第一结节图像分为前景和后景两部分,前景像素点所占比例为ω 0,后景像素点所占比例为ω 1,前景和后景的灰度均值分别是μ 0、μ 1,最终得到类间方差σ。通过遍历的方式,对不同的阈值T进行计算,找到能使类间方差σ最小的阈值T,这就是OTSU分割算法分割的结果。在得到上述声影候选区域后,再根据上述声影候选区域中的非声影部分以及上述结节区域的结节边界,从上述第一结节图像中确定出最终的声影区域。其中,可以根据声影的物理特性从声影候选区域中筛选出并非声影部分消除掉,进而将经过消除处理后的声影候选区域中穿过上述结节区域的结节边界的第一连通域作为最终的上述声影区域。之后在上述结节图像中裁剪掉最终的上述声影区域,得到对应的第二结节图像,并按照上述OTSU分割算法对上述第二结节图像进行二值分割处理,得到处理后的第三结节图像。其中,由于该第二结节图像内的声影区域已经被消除,即不会出现将结节图像中的声影区域误判为囊性区域的情况。另外,按照上述OTSU分割算法对上述第二结节图像进行二值分割处理的处理过程可参照按照OTSU分割算法对上述结节图像进行二值分割处理的过程,只需将之前用到的结点面积(M*N)更改为指定结节面积,即使用上述结节区域的面积减去结节区域与上述最终的声影区域的交集对应的第三面积(M*N-(M*N)∩A,),A为上述最终的声影区域。在得到了上述第三结节图像后,再获取上述第三结节图像中处于结节区域内部的黑色部分区域(即处于结节区域内部的背景部分),并将上述黑色部分区域作为囊性候选区域。其中,处于结节区域内部的黑色部分区域即指处于结节区域内部的后景部分区域。并计算上述囊性候选区域的第一面积、上述结节区域的第二面积,以及上述结节区域与最终的上述声影区域的交集对应的第三面积。然后根据上述第一面积、上述第二面积以及上述第三面积,按照第一预设规则生成上述甲状腺结节的构成粗分类结果,其中,上述构成粗分类结果为囊性为主或囊性,或者上述构成粗分类结果为实性为主或实性。另外,可通过将囊性候选区域的面积与结点面积减去结节内包括的声影区域的面积后得到的指定结节面积的二分之一进行大小比较,进而根据对应的大小比较结果来生成与甲状腺结节的结构组成对应的构成粗分类结果。之后根据上述第三结节图像的梯度以及上述囊性候选区域,从上述第三结节图像确定出最终的囊性区域。其中,可结合结节图像的灰度和梯度来准确地识别出结节图像中的囊性区域,通过生成第三结节图像的梯度图,并从梯度图确定出满足预设条件的种子点,进而根据种子填充法来得到最终的上述囊性区域。最后根据最终的上述囊性区域的第四面积、上述第二面积与上述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成上述甲状腺结节的构成细分类结果,其中,上述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。另外,对于甲状腺结节两种不同的构成粗分类结果,会对应预先设置有用于生成构成细分类结果的两种不同的计算规则,且两种不同的计算规则会设置有对应的两个比值阈值,具体可通过计算最终的囊性区域与结点面积减去结节内包括的声影区域的面积后得到的指定结节面积之间的比值,并将该比值与预设的比值阈值进行大小比较,从而可以根据对应的大小比较结果来快速准确地生成甲状腺结节的构成细分类结果。本实施例包括了对于甲状腺结节构成由粗到细的两步识别过程,在粗识别过程中,先会利用甲状腺超声扫描时的声影物理特性提取出结节图像中的声影区域并剔除,使得不会出现将结节图像中的声影区域误判为囊性区域的情况,从而根据剔除掉声影区域后的结节图像中的候选囊性区域的第一面积,结节区域的第二面积,以及上述结节区域与上述声影区域的交集对应的第三面积来准确地生成甲状腺结节的构成粗分类结果。在得到构成粗分类结果后,再结合结节图像的灰度与梯度信息确定出结节图像中的囊性区域,从而根据囊性区域的第四面积以及上述第二面积、第三面积,调用与上述构成粗分类结果对应的预设计算规则来快速准确地计算生成甲状腺结节的构成细分类结果。通过本申请能够实现准确地识别出甲状腺结节的构成,一方面可以进一步生成确切的与甲状腺结节的构成相关的分析报告,另一方面还可以将该甲状腺结节的构成输入至相应的神经网络模型中来作出关于甲状腺结节结节良恶性的判断。本申请可适用于智慧医疗、数字医疗领域,通过对医学图像分析的优化。
进一步地,本申请一实施例中,上述根据所述声影候选区域中的非声影部分以及结节边界,从所述第一结节图像中确定出声影区域的步骤S4,包括:
S400:获取所述声影候选区域中的非声影部分;
S401:对所述非声影部分进行消除处理,得到处理后的声影候选区域;
S402:获取所述处理后的声影候选区域中穿过所述结节区域的结节边界的第一连通域;
S403:将所述第一连通域作为所述声影区域。
如上述步骤S400至S403所述,上述根据上述声影候选区域中的非声影部分以及结节边界,从上述第一结节图像中确定出声影区域的步骤,具体可包括:首先获取上述声影候选区域中的非声影部分。其中,由于声影具有起始点一定在结节内部,声影会在结节边界外再延续一段距离,且声影中图不会有大 量断层的物理特性。从而可以根据该物理特性,采用线阵探头从上至下遍历第一结节图像中每一列的方式来筛选出上述声影候选区域中的非声影部分。然后对上述非声影部分进行消除处理,得到处理后的声影候选区域。其中,通过在声影候选区域中消除上述非声影部分,能够提高后续生成的处理后的声影候选区域的准确性。之后获取上述处理后的声影候选区域中穿过上述结节区域的结节边界的第一连通域。其中,在获取上述处理后的声影候选区域中穿过上述结节区域的结节边界的第一连通域之前,还可预先对上述处理后的声影候选区域进行膨胀操作,以消除处理后的声影候选区域中的细微小点,避免细微小点对后续得到的第一连通域造成影响,提高获得的第一连通域的准确度。最后在得到上述第一连通域时,将上述第一连通域作为最终的上述声影区域,以便后续能够从上述结节图像中裁剪掉最终的上述声影区域来得到对应的第二结节图像,由于该第二结节图像内的声影区域已经被消除,即不会出现将结节图像中的声影区域误判为囊性区域的情况,进而可以根据该第二结节图像来准确地进行对于甲状腺结节构成的识别处理。
进一步地,本申请一实施例中,上述根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果的步骤S8,包括:
S800:计算所述第二面积减去所述第三面积的第一差值;
S801:计算所述第一差值除以2得到的商值;
S802:判断所述第一面积是否大于所述商值;
S803:若所述第一面积大于所述商值,则判定所述甲状腺结节的构成粗分类结果为囊性为主或囊性;
S804:若所述第一面积不大于所述商值,则判定所述甲状腺结节的构成粗分类结果为实性为主或实性。
如上述步骤S800至S804所述,上述根据上述第一面积、上述第二面积以及上述第三面积,按照第一预设规则生成上述甲状腺结节的构成粗分类结果的步骤,具体可包括:首先计算上述第二面积减去上述第三面积的第一差值。其中,上述第二面积为结节区域的面积,上述第三面积为上述结节区域与最终的上述声影区域的交集对应的面积,通过计算上述第二面积减去上述第三面积的第一差值,即计算结节面积减去结节内包括的声影区域的面积,从而能够得到精准的指定结节面积。然后计算上述第一差值除以2得到的商值。之后判断上述第一面积是否大于上述商值。如果判断出上述第一面积大于上述商值,则判定上述甲状腺结节的构成粗分类结果为囊性为主或囊性。其中,上述第一面积为第三结节图像中的囊性候选区域的面积,当该囊性候选区域的面积大于上述指定结节面积的二分之一时,则可初步判定上述甲状腺的构成为囊性为主或囊性。而如果判断出若上述第一面积不大于上述商值,则判定上述甲状腺结节的构成粗分类结果为实性为主或实性。其中,当上述囊性候选区域的面积不大于上述指定结节面积的二分之一时,则可初步判定上述甲状腺的构成为实性为主或实性。本实施例通过将囊性候选区域的面积与结点面积减去结节内包括的声影区域的面积后得到的指定结节面积的二分之一进行大小比较,进而根据对应的大小比较结果来生成与甲状腺结节的结构组成对应的构成粗分类结果,有利于后续根据该结构粗分类结果来进一步生成甲状腺结节的构成细分类结果。
进一步地,本申请一实施例中,上述根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域的步骤S9,包括:
S900:计算所述第三结节图像的梯度,并生成对应的梯度图;
S901:查找出所述梯度图中灰度小于第一预设阈值,梯度小于第二预设阈值,且处于所述囊性候选区域中的种子点;
S902:按照第二预设规则对所述种子点进行区域生长处理,生成对应的指定连通域;
S903:对所述指定连通域进行腐蚀处理;
S904:计算所述第二面积与指定数值的乘积;
S905:从所有所述指定连通域中筛选出连通域面积大于所述乘积的第二连通域;
S906:将所述第二连通域确定为所述囊性区域。
如上述步骤S900至S906所述,本实施例不局限于只利用结节图像的灰度信息,而是会结合结节图像的灰度和梯度来准确地识别出结节图像中的囊性区域。具体地,上述根据上述第三结节图像的梯度以及上述囊性候选区域,从上述第三结节图像确定出囊性区域的步骤可包括:首先计算上述第三结节图像的梯度,并生成对应的梯度图。其中,可以通过公式
Figure PCTCN2020132595-appb-000001
计算生成与第三结节图像对应的梯度图,g x与g y分别为边缘检测算子,对第三结节图像在水平方向和竖直方向上计算得到梯度图g。然后查找出上述梯度图中灰度小于第一预设阈值,梯度小于第二预设阈值,且处于上述囊性候选区域中的种子点。其中,对于上述第一预设阈值与第二预设阈值不作具体限定,可根据实际需求进行设置,例如可分别设置为2和4。在得到了上述种子点后,按照第二预设规则对上述种子点进行区域生长处理,生成对 应的指定连通域。其中,上述第二预设规则具体可为种子填充法,种子填充法的流程包括:选取一个前景像素点作为种子点,然后根据连通区域的构成条件,将与种子相邻的前景像素合并到同一个集合中,最终得到的像素集合则为一个连通区域。在得到了上述指定连通域后,再对上述指定连通域进行腐蚀处理。其中,通过对指定连通域进行腐蚀处理,可以进行填洞并去掉指定连通域中与非结节区域的连接部分,实现对于甲状腺结节形状的细化。之后计算上述第二面积与指定数值的乘积。其中,对于上述指定数值不作具体限定,可根据实际需求进行设置,例如可设置为0.001。并从所有上述指定连通域中筛选出连通域面积大于上述乘积的第二连通域。其中,对于面积过小的连通域,会将其视为无意义区域,为了避免这些无意义区域对囊性区域造成影响,则需要从囊性候选区域中剔除掉无意义区域。最后将上述第二连通域确定为最终的上述囊性区域,以便后续能够根据得到的最终的该囊性区域的第四面积,以及上述第二面积与上述第三面积,调用与上述构成粗分类结果对应的预设计算规则来快速准确地计算生成甲状腺结节的构成细分类结果。
进一步地,本申请一实施例中,上述按照第三预设规则对所述种子点进行区域生长处理,生成对应的指定连通域的步骤S902,包括:
S9020:获取指定种子点,其中,所述指定种子点为所有所述种子点中的任意一个种子点;
S9021:根据连通区域的构成条件,将与所述指定种子点相邻的像素合并到同一个集合中,生成指定像素集合,其中,连通区域的构成条件包括像素值相同,并且像素值相邻;
S9022:将生成的所述指定集合作为与所述指定种子点对应的连通域。
如上述步骤S9020至S9022所述,上述按照第三预设规则对所述种子点进行区域生长处理,生成对应的指定连通域的步骤,具体可包括:首先获取指定种子点,其中,上述指定种子点为所有上述种子点中的任意一个种子点。然后根据连通区域的构成条件,将与上述指定种子点相邻的像素合并到同一个集合中,生成指定像素集合;其中,连通区域的构成条件包括像素值相同,并且像素值相邻。最后将生成的上述指定集合作为与上述指定种子点对应的连通域。本实施例根据连通区域的构成条件,能够快速便捷地得到种子点经过生长填充处理后的指定连通域,从而后续能够根据该指定连通域来快速确定出第三结节图像中的最终的囊性区域,并根据得到的最终的该囊性区域的第四面积,以及上述第二面积与上述第三面积,调用与上述构成粗分类结果对应的预设计算规则来快速准确地计算生成甲状腺结节的构成细分类结果。
进一步地,本申请一实施例中,上述计算所述第三结节图像的梯度,并生成对应的梯度图的步骤S900之后,包括:
S9000:通过中值滤波算法筛选出所述梯度图中的异常梯度;
S9001:从所述梯度图中删除所述异常梯度。
如上述步骤S9000至S9001所述,在计算上述第三结节图像的梯度,并生成对应的梯度图后,还可进一步对上述梯度图进行中值滤波处理,以过滤掉无用的噪音数据。具体地,上述计算上述第三结节图像的梯度,并生成对应的梯度图的步骤之后,包括:首先通过中值滤波算法筛选出上述梯度图中的异常梯度。在得到了上述异常梯度后,再从上述梯度图中删除上述异常梯度。本实施例通过使用中值滤波算法去除掉梯度图中的异常梯度,能够提高后续查找的种子点的准确性,进而提高后续生成的最终的囊性区域的准确性。
进一步地,本申请一实施例中,上述根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果的步骤S10,包括:
S1000:计算所述第二面积减去所述第三面积的第二差值;
S1001:计算所述第四面积与所述第二差值之间的比值;
S1002:当所述构成粗分类结果为囊性为主或囊性时,判断所述比值是否大于预设的第一比值阈值;
S1003:若所述比值大于所述第一比值阈值,则判定所述甲状腺结节的构成细分类结果为囊性,否则判定所述甲状腺结节的构成细分类结果为囊性为主;
S1004:当所述构成粗分类结果为实性为主或实性时,判断所述比值是否小于预设的第二比值阈值;
S1005:若所述比值小于所述第二比值阈值,则判定所述甲状腺结节的构成细分类结果为实性,否则判定所述甲状腺结节的构成细分类结果为实性为主。
如上述步骤S1000至S1005所述,对于甲状腺结节两种不同的构成粗分类结果,会对应预先设置有用于生成构成细分类结果的两种不同的计算规则,且两种不同的计算规则会设置有对应的两个比值阈值。上述根据上述囊性区域的第四面积、上述第二面积与上述第三面积,调用与上述构成粗分类结果对应的预设计算规则计算生成上述甲状腺结节的构成细分类结果的步骤,具体可包括:首先计算上述第二面积减去上述第三面积的第二差值。并计算上述第四面积与上述第二差值之间的比值。其中,上述第二面积为结节区域的面积,上述第三面积为上述结节区域与最终的上述声影区域的交集对应的面积,通过计算 上述第二面积减去上述第三面积的第一差值,即计算结节面积减去结节内包括的声影区域的面积,从而能够得到精准的指定结节面积。另外,上述第四面积与上述第二差值之间的比值是指最终的上述囊性区域的面积与上述指定结节面积之间的比值。当上述构成粗分类结果为囊性为主或囊性时,判断上述比值是否大于预设的第一比值阈值。其中,对于上述第一比值阈值不作具体限定,可根据实际需求进行设置,例如可设置为0.95。如果判断出上述比值大于上述第一比值阈值,则判定上述甲状腺结节的构成细分类结果为囊性,否则判定上述甲状腺结节的构成细分类结果为囊性为主。而当上述构成粗分类结果为实性为主或实性时,判断上述比值是否小于预设的第二比值阈值。其中,对于上述第二比值阈值不作具体限定,可根据实际需求进行设置,例如可设置为0.08。如果判断出上述比值小于上述第二比值阈值,则判定上述甲状腺结节的构成细分类结果为实性,否则判定上述甲状腺结节的构成细分类结果为实性为主。本实施例在得到了甲状腺结节的构成粗分类结果后,通过计算最终的囊性区域与结点面积减去结节内包括的声影区域的面积后得到的指定结节面积之间的比值,并将该比值与预设的比值阈值进行大小比较,从而可以根据对应的大小比较结果来快速准确地生成甲状腺结节的构成细分类结果。通过本实施例能够实现准确地识别出甲状腺结节的构成,一方面可以进一步生成确切的与甲状腺结节的构成相关的分析报告,另一方面还可以将该甲状腺结节的构成输入至相应的神经网络模型中来作出关于甲状腺结节结节良恶性的判断。
本申请实施例中的甲状腺结节图像的处理方法还可以应用于区块链领域,如将上述甲状腺结节的构成细分类结果等数据存储于区块链上。通过使用区块链来对上述甲状腺结节的构成细分类结果进行存储和管理,能够有效地保证上述甲状腺结节的构成细分类结果的安全性与不可篡改性。
参照图2,本申请一实施例中还提供了一种甲状腺结节图像的处理装置,包括:
第一获取模块1,用于获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
第一生成模块2,用于截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
第一处理模块3,用于按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
第一确定模块4,用于根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
第二处理模块5,用于在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
第二获取模块6,用于获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
计算模块7,用于计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
第二生成模块8,用于根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
第二确定模块9,用于根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
第三生成模块10,用于根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
本实施例中,上述甲状腺结节图像的处理装置中的第一获取模块、第一生成模块、第一处理模块、第一确定模块、第二处理模块、第二获取模块、计算模块、第二生成模块、第二确定模块与第三生成模块的功能和作用的实现过程具体详见上述甲状腺结节图像的处理方法中对应步骤S1至S10的实现过程,在此不再赘述。
进一步地,本申请一实施例中,上述第一确定模块,包括:
第一获取子模块,用于获取所述声影候选区域中的非声影部分;
第一处理子模块,用于对所述非声影部分进行消除处理,得到处理后的声影候选区域;
第二获取子模块,用于获取所述处理后的声影候选区域中穿过所述结节区域的结节边界的第一连通域;
第一确定子模块,用于将所述第一连通域作为所述声影区域。
本实施例中,上述甲状腺结节图像的处理装置中的第一获取子模块、第一处理子模块、第二获取子模块与第一确定子模块的功能和作用的实现过程具体详见上述甲状腺结节图像的处理方法中对应步骤S400至S403的实现过程,在此不再赘述。
进一步地,本申请一实施例中,上述第二生成模块,包括:
第一计算子模块,用于计算所述第二面积减去所述第三面积的第一差值;
第二计算子模块,用于计算所述第一差值除以2得到的商值;
第一判断子模块,用于判断所述第一面积是否大于所述商值;
第一判定子模块,用于若所述第一面积大于所述商值,则判定所述甲状腺结节的构成粗分类结果为囊性为主或囊性;
第二判定子模块,用于若所述第一面积不大于所述商值,则判定所述甲状腺结节的构成粗分类结果为实性为主或实性。
本实施例中,上述甲状腺结节图像的处理装置中的第一计算子模块、第二计算子模块、第一判断子模块、第一判定子模块与第二判定子模块的功能和作用的实现过程具体详见上述甲状腺结节图像的处理方法中对应步骤S800至S804的实现过程,在此不再赘述。
进一步地,本申请一实施例中,上述第二确定模块,包括:
第三计算子模块,用于计算所述第三结节图像的梯度,并生成对应的梯度图;
查找子模块,用于查找出所述梯度图中灰度小于第一预设阈值,梯度小于第二预设阈值,且处于所述囊性候选区域中的种子点;
第一处理子模块,用于按照第二预设规则对所述种子点进行区域生长处理,生成对应的指定连通域;
第二处理子模块,用于对所述指定连通域进行腐蚀处理;
第四计算子模块,用于计算所述第二面积与指定数值的乘积;
筛选子模块,用于从所有所述指定连通域中筛选出连通域面积大于所述乘积的第二连通域;
确定子模块,用于将所述第二连通域确定为所述囊性区域。
本实施例中,上述甲状腺结节图像的处理装置中的第三计算子模块、查找子模块、第一处理子模块、第二处理子模块、第四计算子模块、筛选子模块与确定子模块的功能和作用的实现过程具体详见上述甲状腺结节图像的处理方法中对应步骤S900至S906的实现过程,在此不再赘述。
进一步地,本申请一实施例中,上述第一处理子模块,包括:
获取单元,用于获取指定种子点,其中,所述指定种子点为所有所述种子点中的任意一个种子点;
生成单元,用于根据连通区域的构成条件,将与所述指定种子点相邻的像素合并到同一个集合中,生成指定像素集合,其中,连通区域的构成条件包括像素值相同,并且像素值相邻;
确定单元,用于将生成的所述指定集合作为与所述指定种子点对应的连通域。
本实施例中,上述甲状腺结节图像的处理装置中的获取单元、生成单元与确定单元的功能和作用的实现过程具体详见上述甲状腺结节图像的处理方法中对应步骤S9020至S9022的实现过程,在此不再赘述。
进一步地,本申请一实施例中,上述第二确定模块,包括:
筛选单元,用于通过中值滤波算法筛选出所述梯度图中的异常梯度;
删除单元,用于从所述梯度图中删除所述异常梯度。
本实施例中,上述甲状腺结节图像的处理装置中的筛选单元与删除单元的功能和作用的实现过程具体详见上述甲状腺结节图像的处理方法中对应步骤S9000至S9001的实现过程,在此不再赘述。
进一步地,本申请一实施例中,上述第三生成模块,包括:
第五计算子模块,用于计算所述第二面积减去所述第三面积的第二差值;
第六计算子模块,用于计算所述第四面积与所述第二差值之间的比值;
第二判断子模块,用于当所述粗分类结果为囊性为主或囊性时,判断所述比值是否大于预设的第一比值阈值;
第三判定子模块,用于若所述比值大于所述第一比值阈值,则判定所述甲状腺结节的构成细分类结果为囊性,否则判定所述甲状腺结节的构成细分类结果为囊性为主;
第三判断子模块,用于当所述粗分类结果为实性为主或实性时,判断所述比值是否小于预设的第二比值阈值;
第四判定子模块,用于若所述比值小于所述第二比值阈值,则判定所述甲状腺结节的构成细分类结果为实性,否则判定所述甲状腺结节的构成细分类结果为实性为主。
本实施例中,上述甲状腺结节图像的处理装置中的第五计算子模块、第六计算子模块、第二判断子模块、第三判定子模块、第三判断子模块与第四判定子模块的功能和作用的实现过程具体详见上述甲状腺结节图像的处理方法中对应步骤S1000至S1005的实现过程,在此不再赘述。
参照图3,本申请实施例中还提供一种计算机设备,该计算机设备可以是服务器,其内部结构可以如图3所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口、显示屏、输入装置和数据库。其中,该计算机设备设计的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易 失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存储器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的数据库用于存储结节掩膜图像、结节图像、结节图像、第三结节图像、最终的囊性区域以及甲状腺结节的构成细分类结果等数据。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机设备的显示屏是计算机中必不可少的一种图文输出设备,用于将数字信号转换为光信号,使文字与图形在显示屏的屏幕上显示出来。该计算机设备的输入装置是计算机与用户或其他设备之间进行信息交换的主要装置,用于把数据、指令及某些标志信息等输送到计算机中去。该计算机程序被处理器执行时以实现一种甲状腺结节图像的处理方法。
上述处理器执行上述甲状腺结节图像的处理方法的步骤:
获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
本领域技术人员可以理解,图3中示出的结构,仅仅是与本申请方案相关的部分结构的框图,并不构成对本申请方案所应用于其上的装置、计算机设备的限定。
本申请一实施例还提供一种计算机可读存储介质,所述计算机可读存储介质可以是非易失性,也可以是易失性,其上存储有计算机程序,计算机程序被处理器执行时实现上述任一个示例性实施例所示出的甲状腺结节图像的处理方法,所述甲状腺结节图像的处理方法包括以下步骤:
获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、 囊性、实性为主或者实性。
本方案可应用于智慧城市中的数字医疗领域,从而推动智慧城市的建设。
综上所述,本申请实施例中提供的甲状腺结节图像的处理方法、装置、计算机设备和存储介质,能够实现准确地识别出甲状腺结节的构成,一方面可以进一步生成确切的与甲状腺结节的构成相关的分析报告,另一方面还可以将该甲状腺结节的构成输入至相应的神经网络模型中来作出关于甲状腺结节结节良恶性的判断。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,所述的计算机程序可存储与一非易失性计算机可读取存储介质中,该计算机程序在执行时,可包括如上述各方法的实施例的流程。其中,本申请所提供的和实施例中所使用的对存储器、存储、数据库或其它介质的任何引用,均可包括非易失性和/或易失性存储器。非易失性存储器可以包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限,RAM通过多种形式可得,诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双速据率SDRAM(SSRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
以上所述仅为本申请的优选实施例,并非因此限制本申请的专利范围,凡是利用本申请说明书及附图内容所作的等效结构或等效流程变换,或直接或间接运用在其他相关的技术领域,均同理包括在本申请的专利保护范围内。

Claims (20)

  1. 一种甲状腺结节图像的处理方法,其中,包括:
    获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
    截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
    按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
    根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
    在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
    获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
    计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
    根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
    根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
    根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
  2. 根据权利要求1所述的甲状腺结节图像的处理方法,其中,所述根据所述声影候选区域中的非声影部分以及结节边界,从所述第一结节图像中确定出声影区域的步骤,包括:
    获取所述声影候选区域中的非声影部分;
    对所述非声影部分进行消除处理,得到处理后的声影候选区域;
    获取所述处理后的声影候选区域中穿过所述结节区域的结节边界的第一连通域;
    将所述第一连通域作为所述声影区域。
  3. 根据权利要求1所述的甲状腺结节图像的处理方法,其中,所述根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果的步骤,包括:
    计算所述第二面积减去所述第三面积的第一差值;
    计算所述第一差值除以2得到的商值;
    判断所述第一面积是否大于所述商值;
    若所述第一面积大于所述商值,则判定所述甲状腺结节的构成粗分类结果为囊性为主或囊性;
    若所述第一面积不大于所述商值,则判定所述甲状腺结节的构成粗分类结果为实性为主或实性。
  4. 根据权利要求1所述的甲状腺结节图像的处理方法,其中,所述根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域的步骤,包括:
    计算所述第三结节图像的梯度,并生成对应的梯度图;
    查找出所述梯度图中灰度小于第一预设阈值,梯度小于第二预设阈值,且处于所述囊性候选区域中的种子点;
    按照第二预设规则对所述种子点进行区域生长处理,生成对应的指定连通域;
    对所述指定连通域进行腐蚀处理;
    计算所述第二面积与指定数值的乘积;
    从所有所述指定连通域中筛选出连通域面积大于所述乘积的第二连通域;
    将所述第二连通域确定为所述囊性区域。
  5. 根据权利要求4所述的甲状腺结节图像的处理方法,其中,所述按照第三预设规则对所述种子点进行区域生长处理,生成对应的指定连通域的步骤,包括:
    获取指定种子点,其中,所述指定种子点为所有所述种子点中的任意一个种子点;
    根据连通区域的构成条件,将与所述指定种子点相邻的像素合并到同一个集合中,生成指定像素集合,其中,连通区域的构成条件包括像素值相同,并且像素值相邻;
    将生成的所述指定集合作为与所述指定种子点对应的连通域。
  6. 根据权利要求4所述的甲状腺结节图像的处理方法,其中,所述计算所述第三结节图像的梯度,并生成对应的梯度图的步骤之后,包括:
    通过中值滤波算法筛选出所述梯度图中的异常梯度;
    从所述梯度图中删除所述异常梯度。
  7. 根据权利要求1所述的甲状腺结节图像的处理方法,其中,所述根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果的步骤,包括:
    计算所述第二面积减去所述第三面积的第二差值;
    计算所述第四面积与所述第二差值之间的比值;
    当所述构成粗分类结果为囊性为主或囊性时,判断所述比值是否大于预设的第一比值阈值;
    若所述比值大于所述第一比值阈值,则判定所述甲状腺结节的构成细分类结果为囊性,否则判定所述甲状腺结节的构成细分类结果为囊性为主;
    当所述构成粗分类结果为实性为主或实性时,判断所述比值是否小于预设的第二比值阈值;
    若所述比值小于所述第二比值阈值,则判定所述甲状腺结节的构成细分类结果为实性,否则判定所述甲状腺结节的构成细分类结果为实性为主。
  8. 一种甲状腺结节图像的处理装置,其中,包括:
    第一获取模块,用于获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
    第一生成模块,用于截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
    第一处理模块,用于按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
    第一确定模块,用于根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
    第二处理模块,用于在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
    第二获取模块,用于获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
    计算模块,用于计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
    第二生成模块,用于根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
    第二确定模块,用于根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
    第三生成模块,用于根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
  9. 一种计算机设备,包括存储器和处理器,所述存储器中存储有计算机程序,其中,所述处理器执行所述计算机程序时实现一种甲状腺结节图像的处理方法:
    其中,所述甲状腺结节图像的处理方法包括:
    获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
    截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
    按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
    根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
    在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
    获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
    计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
    根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
    根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
    根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
  10. 根据权利要求9所述的计算机设备,其中,所述根据所述声影候选区域中的非声影部分以及结节边界,从所述第一结节图像中确定出声影区域的步骤,包括:
    获取所述声影候选区域中的非声影部分;
    对所述非声影部分进行消除处理,得到处理后的声影候选区域;
    获取所述处理后的声影候选区域中穿过所述结节区域的结节边界的第一连通域;
    将所述第一连通域作为所述声影区域。
  11. 根据权利要求9所述的计算机设备,其中,所述根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果的步骤,包括:
    计算所述第二面积减去所述第三面积的第一差值;
    计算所述第一差值除以2得到的商值;
    判断所述第一面积是否大于所述商值;
    若所述第一面积大于所述商值,则判定所述甲状腺结节的构成粗分类结果为囊性为主或囊性;
    若所述第一面积不大于所述商值,则判定所述甲状腺结节的构成粗分类结果为实性为主或实性。
  12. 根据权利要求9所述的计算机设备,其中,所述根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域的步骤,包括:
    计算所述第三结节图像的梯度,并生成对应的梯度图;
    查找出所述梯度图中灰度小于第一预设阈值,梯度小于第二预设阈值,且处于所述囊性候选区域中的种子点;
    按照第二预设规则对所述种子点进行区域生长处理,生成对应的指定连通域;
    对所述指定连通域进行腐蚀处理;
    计算所述第二面积与指定数值的乘积;
    从所有所述指定连通域中筛选出连通域面积大于所述乘积的第二连通域;
    将所述第二连通域确定为所述囊性区域。
  13. 根据权利要求12所述的计算机设备,其中,所述按照第三预设规则对所述种子点进行区域生长处理,生成对应的指定连通域的步骤,包括:
    获取指定种子点,其中,所述指定种子点为所有所述种子点中的任意一个种子点;
    根据连通区域的构成条件,将与所述指定种子点相邻的像素合并到同一个集合中,生成指定像素集合,其中,连通区域的构成条件包括像素值相同,并且像素值相邻;
    将生成的所述指定集合作为与所述指定种子点对应的连通域。
  14. 根据权利要求12所述的计算机设备,其中,所述计算所述第三结节图像的梯度,并生成对应的梯度图的步骤之后,包括:
    通过中值滤波算法筛选出所述梯度图中的异常梯度;
    从所述梯度图中删除所述异常梯度。
  15. 根据权利要求9所述的计算机设备,其中,所述根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果的步骤,包括:
    计算所述第二面积减去所述第三面积的第二差值;
    计算所述第四面积与所述第二差值之间的比值;
    当所述构成粗分类结果为囊性为主或囊性时,判断所述比值是否大于预设的第一比值阈值;
    若所述比值大于所述第一比值阈值,则判定所述甲状腺结节的构成细分类结果为囊性,否则判定所述甲状腺结节的构成细分类结果为囊性为主;
    当所述构成粗分类结果为实性为主或实性时,判断所述比值是否小于预设的第二比值阈值;
    若所述比值小于所述第二比值阈值,则判定所述甲状腺结节的构成细分类结果为实性,否则判定所述甲状腺结节的构成细分类结果为实性为主。
  16. 一种计算机可读存储介质,其上存储有计算机程序,其中,所述计算机程序被处理器执行时实现一种甲状腺结节图像的处理方法,其中,所述甲状腺结节图像的处理方法包括以下步骤:
    获取输入的带有甲状腺结节的超声图像,对所述超声图像标记结节区域,生成对应所述结节区域的结节掩膜图像;
    截取所述结节掩膜图像的最小外接矩形对应的图像,得到结节图像;
    按照预设的OTSU分割算法对所述结节图像进行二值分割处理,生成处理后的第一结节图像,并从所述第一结节图像中获取声影候选区域;
    根据所述声影候选区域中的非声影部分以及所述结节区域的结节边界,从所述第一结节图像中确定出声影区域;
    在所述结节图像中裁剪掉所述声影区域,得到对应的第二结节图像,并按照所述OTSU分割算法对所述第二结节图像进行二值分割处理,得到处理后的第三结节图像;
    获取所述第三结节图像中处于结节区域内部的黑色部分区域,并将所述黑色部分区域作为囊性候选区域;
    计算所述囊性候选区域的第一面积、所述结节区域的第二面积,以及所述结节区域与所述声影区域的交集对应的第三面积;
    根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果,其中,所述构成粗分类结果为囊性为主或囊性,或者所述构成粗分类结果为实性为主或实性;
    根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域;
    根据所述囊性区域的第四面积、所述第二面积与所述第三面积,调用与所述构成粗分类结果对应的预设计算规则计算生成所述甲状腺结节的构成细分类结果,其中,所述构成细分类结果包括:囊性为主、囊性、实性为主或者实性。
  17. 根据权利要求16所述的计算机可读存储介质,其中,所述根据所述声影候选区域中的非声影部分以及结节边界,从所述第一结节图像中确定出声影区域的步骤,包括:
    获取所述声影候选区域中的非声影部分;
    对所述非声影部分进行消除处理,得到处理后的声影候选区域;
    获取所述处理后的声影候选区域中穿过所述结节区域的结节边界的第一连通域;
    将所述第一连通域作为所述声影区域。
  18. 根据权利要求16所述的计算机可读存储介质,其中,所述根据所述第一面积、所述第二面积以及所述第三面积,按照第一预设规则生成所述甲状腺结节的构成粗分类结果的步骤,包括:
    计算所述第二面积减去所述第三面积的第一差值;
    计算所述第一差值除以2得到的商值;
    判断所述第一面积是否大于所述商值;
    若所述第一面积大于所述商值,则判定所述甲状腺结节的构成粗分类结果为囊性为主或囊性;
    若所述第一面积不大于所述商值,则判定所述甲状腺结节的构成粗分类结果为实性为主或实性。
  19. 根据权利要求16所述的计算机可读存储介质,其中,所述根据所述第三结节图像的梯度以及所述囊性候选区域,从所述第三结节图像确定出囊性区域的步骤,包括:
    计算所述第三结节图像的梯度,并生成对应的梯度图;
    查找出所述梯度图中灰度小于第一预设阈值,梯度小于第二预设阈值,且处于所述囊性候选区域中的种子点;
    按照第二预设规则对所述种子点进行区域生长处理,生成对应的指定连通域;
    对所述指定连通域进行腐蚀处理;
    计算所述第二面积与指定数值的乘积;
    从所有所述指定连通域中筛选出连通域面积大于所述乘积的第二连通域;
    将所述第二连通域确定为所述囊性区域。
  20. 根据权利要求19所述的计算机可读存储介质,其中,所述计算所述第三结节图像的梯度,并生成对应的梯度图的步骤之后,包括:
    通过中值滤波算法筛选出所述梯度图中的异常梯度;
    从所述梯度图中删除所述异常梯度。
PCT/CN2020/132595 2020-09-27 2020-11-30 甲状腺结节图像的处理方法、装置和计算机设备 WO2021169452A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202011035723.7A CN112102311B (zh) 2020-09-27 2020-09-27 甲状腺结节图像的处理方法、装置和计算机设备
CN202011035723.7 2020-09-27

Publications (1)

Publication Number Publication Date
WO2021169452A1 true WO2021169452A1 (zh) 2021-09-02

Family

ID=73782679

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/132595 WO2021169452A1 (zh) 2020-09-27 2020-11-30 甲状腺结节图像的处理方法、装置和计算机设备

Country Status (2)

Country Link
CN (1) CN112102311B (zh)
WO (1) WO2021169452A1 (zh)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023173545A1 (zh) * 2022-03-15 2023-09-21 平安科技(深圳)有限公司 超声图像的基准图像的生成方法、装置、设备及存储介质

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030161513A1 (en) * 2002-02-22 2003-08-28 The University Of Chicago Computerized schemes for detecting and/or diagnosing lesions on ultrasound images using analysis of lesion shadows
CN107680678A (zh) * 2017-10-18 2018-02-09 北京航空航天大学 基于多尺度卷积神经网络甲状腺超声图像结节自动诊断系统
CN108734694A (zh) * 2018-04-09 2018-11-02 华南农业大学 基于faster r-cnn的甲状腺肿瘤超声图像自动识别方法
US20200005460A1 (en) * 2018-06-28 2020-01-02 Shenzhen Imsight Medical Technology Co. Ltd. Method and device for detecting pulmonary nodule in computed tomography image, and computer-readable storage medium
CN111243042A (zh) * 2020-02-28 2020-06-05 浙江德尚韵兴医疗科技有限公司 基于深度学习的超声甲状腺结节良恶性特征可视化的方法
CN111461202A (zh) * 2020-03-30 2020-07-28 上海尽星生物科技有限责任公司 甲状腺结节超声图像实时识别方法及装置
CN111553919A (zh) * 2020-05-12 2020-08-18 上海深至信息科技有限公司 一种基于弹性超声成像的甲状腺结节分析系统
CN111612752A (zh) * 2020-05-15 2020-09-01 江苏省人民医院(南京医科大学第一附属医院) 基于Faster-RCNN的超声图像甲状腺结节智能检测系统

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2005530138A (ja) * 2002-06-18 2005-10-06 ライフスパン バイオサイエンス,インク. 組織標本の中の重要な構造のコンピュータを利用した画像捕捉
KR101656582B1 (ko) * 2014-11-25 2016-09-12 재단법인 아산사회복지재단 갑상선암 자동 판별 방법 및 시스템
CA3010052C (en) * 2015-12-30 2023-12-05 Technische Universitat Dresden Method and means for diagnosing tumours

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030161513A1 (en) * 2002-02-22 2003-08-28 The University Of Chicago Computerized schemes for detecting and/or diagnosing lesions on ultrasound images using analysis of lesion shadows
CN107680678A (zh) * 2017-10-18 2018-02-09 北京航空航天大学 基于多尺度卷积神经网络甲状腺超声图像结节自动诊断系统
CN108734694A (zh) * 2018-04-09 2018-11-02 华南农业大学 基于faster r-cnn的甲状腺肿瘤超声图像自动识别方法
US20200005460A1 (en) * 2018-06-28 2020-01-02 Shenzhen Imsight Medical Technology Co. Ltd. Method and device for detecting pulmonary nodule in computed tomography image, and computer-readable storage medium
CN111243042A (zh) * 2020-02-28 2020-06-05 浙江德尚韵兴医疗科技有限公司 基于深度学习的超声甲状腺结节良恶性特征可视化的方法
CN111461202A (zh) * 2020-03-30 2020-07-28 上海尽星生物科技有限责任公司 甲状腺结节超声图像实时识别方法及装置
CN111553919A (zh) * 2020-05-12 2020-08-18 上海深至信息科技有限公司 一种基于弹性超声成像的甲状腺结节分析系统
CN111612752A (zh) * 2020-05-15 2020-09-01 江苏省人民医院(南京医科大学第一附属医院) 基于Faster-RCNN的超声图像甲状腺结节智能检测系统

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023173545A1 (zh) * 2022-03-15 2023-09-21 平安科技(深圳)有限公司 超声图像的基准图像的生成方法、装置、设备及存储介质

Also Published As

Publication number Publication date
CN112102311B (zh) 2023-07-18
CN112102311A (zh) 2020-12-18

Similar Documents

Publication Publication Date Title
Kazanskiy et al. Machine vision system for singularity detection in monitoring the long process
US20030152272A1 (en) Detecting overlapping images in an automatic image segmentation device with the presence of severe bleeding
CN110781885A (zh) 基于图像处理的文本检测方法、装置、介质及电子设备
JP2017521779A (ja) 画像解析を用いた核のエッジの検出
CN111899246B (zh) 玻片数字化信息质量检测方法、装置、设备及介质
WO2022141145A1 (zh) 面向对象的高分辨率遥感影像多尺度分割方法及系统
Tsai et al. Comprehensive, quantitative crack detection algorithm performance evaluation system
Chen et al. Deep learning for the detection and recognition of rail defects in ultrasound B-scan images
Koprowski et al. Assessment of significance of features acquired from thyroid ultrasonograms in Hashimoto's disease
CN114329034A (zh) 基于细粒度语义特征差异的图像文本匹配判别方法及系统
WO2021169452A1 (zh) 甲状腺结节图像的处理方法、装置和计算机设备
CN115861715A (zh) 基于知识表示增强的图像目标关系识别算法
Lang et al. Pavement cracking detection and classification based on 3d image using multiscale clustering model
CN112258532B (zh) 一种超声图像中胼胝体的定位及分割方法
CN116433695B (zh) 一种乳腺钼靶图像的乳腺区域提取方法及系统
CN113724267A (zh) 一种乳腺超声图像肿瘤分割方法及装置
Wilm et al. Fast whole-slide cartography in colon cancer histology using superpixels and CNN classification
WO2023220913A1 (zh) 细胞图像处理方法、电子设备以及存储介质
CN109460768B (zh) 一种用于组织病理学显微图像的文本检测与去除方法
CN113822839A (zh) 医学图像的处理方法、装置、计算机设备和存储介质
Wang et al. A TMA de-arraying method for high throughput biomarker discovery in tissue research
CN113537026B (zh) 建筑平面图中的图元检测方法、装置、设备及介质
Jamshidi et al. Image analysis method for heterogeneity and porosity characterization of biomimetic hydrogels
JP2015156894A (ja) 医用画像処理装置、その医用目的領域抽出方法及び医用目的領域抽出処理プログラム
CN115249358A (zh) 一种巨噬细胞内碳颗粒定量检测方法及系统、计算机设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20921752

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20921752

Country of ref document: EP

Kind code of ref document: A1