WO2024212136A1 - 一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法 - Google Patents

一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法 Download PDF

Info

Publication number
WO2024212136A1
WO2024212136A1 PCT/CN2023/087890 CN2023087890W WO2024212136A1 WO 2024212136 A1 WO2024212136 A1 WO 2024212136A1 CN 2023087890 W CN2023087890 W CN 2023087890W WO 2024212136 A1 WO2024212136 A1 WO 2024212136A1
Authority
WO
WIPO (PCT)
Prior art keywords
contour
voronoi
segmentation
convex difference
segmentation result
Prior art date
Legal status (The legal status 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 status listed.)
Ceased
Application number
PCT/CN2023/087890
Other languages
English (en)
French (fr)
Inventor
胡战利
耿孟晓
黄振兴
张娜
梁栋
郑海荣
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Institute of Advanced Technology of CAS
Original Assignee
Shenzhen Institute of Advanced Technology of CAS
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 Shenzhen Institute of Advanced Technology of CAS filed Critical Shenzhen Institute of Advanced Technology of CAS
Priority to PCT/CN2023/087890 priority Critical patent/WO2024212136A1/zh
Publication of WO2024212136A1 publication Critical patent/WO2024212136A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/12—Edge-based segmentation

Definitions

  • the invention belongs to the technical field of image segmentation, and particularly relates to an interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference.
  • Cervical cancer is a common gynecological malignancy with high morbidity and mortality.
  • the latest data released by the World Health Organization's International Agency for Research on Cancer in December 2020 showed that there are nearly 604,000 new cases of cervical cancer and more than 341,000 deaths each year, of which more than 80% of cases are from developing countries.
  • Conventional smears (CS) technology is one of the simplest and most important cytological screening methods for the prevention and early detection of cervical cancer.
  • pathologists manually adjust the field of view and size under a high-power microscope for diagnosis. This very repetitive observation and diagnosis process not only takes time and effort, but also often impairs vision and greatly reduces the efficiency of clinical diagnosis.
  • CAD computer-aided diagnosis
  • the purpose of the embodiments of this specification is to provide an interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference.
  • the present application provides an interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference, the method comprising:
  • the Voronoi contour is used as the initial contour to input into the pre-built weighted convex difference model to obtain the initial segmentation result;
  • the final segmentation results are filtered.
  • acquiring a pathological image includes:
  • the original pathological slices are cropped into pathological images of a preset pixel size.
  • the annotation point satisfies the following conditions:
  • the number of marking points meets the preset conditions.
  • the marking points are evenly distributed in the pathological image according to the positions of the cell nuclei in the pathological image.
  • the weighted convex difference model is:
  • ⁇ is a parameter
  • f is the pathological image
  • c 1 ,c 2 ⁇ R is the average value of the pixels inside and outside the segmentation boundary during two-phase segmentation
  • ⁇ is the entire area of the pathological image
  • u is the segmentation result
  • M() is the constructed optimization objective function
  • ⁇ (f) and ⁇ (f) are adaptive weights used to capture the edge information of the image.
  • the adaptive weights ⁇ (f) and ⁇ (f) are respectively:
  • solving a weighted convex difference model includes:
  • Hot kernel convolution is used to approximate the length regularization term:
  • ⁇ is the scale parameter
  • model (3) is approximately equivalent to the following minimization problem:
  • screening the final segmentation result according to the initial segmentation result includes:
  • a histogram is created, where the horizontal axis of the histogram is the serial number of each cell nucleus and the vertical axis is the area;
  • the connected areas with an area smaller than the threshold are integrated into the background area to obtain the final segmentation result.
  • the present application provides an interactive cell nucleus segmentation device combining Voronoi contour and weighted convex difference, the device comprising:
  • An acquisition and annotation module is used to acquire pathological images and annotate annotation points on the pathological images
  • the Voronoi contour construction module is used to construct the Voronoi contour using the annotated points as prior information
  • An initial segmentation module is used to input the Voronoi contour as an initial contour into a pre-built weighted convex difference model to obtain an initial segmentation result;
  • the screening module is used to screen the final segmentation results according to the initial segmentation results.
  • the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, an interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference as in the first aspect is implemented.
  • the present application provides a readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference as in the first aspect.
  • the solution adopts an interactive semi-automatic method in which the pathologist marks several points to obtain more information about the image, increases the pathologist's subjective guidance, which makes the definition clearer and accurately solves the problem of rough results previously obtained.
  • a new weighted convex difference mathematical model is proposed to improve the accuracy of segmentation.
  • the perimeter of the segmentation region is approximated by hot kernel convolution, and the proposed model is solved using a first-order linearization method to obtain an approximate solution of the model.
  • a post-processing stage was explored, which selected a suitable threshold according to the drawn histogram to remove unnecessary spots. It combined the advantages of the histogram and was not only simple and easy to operate, but also avoided the arbitrariness and blindness of the selection.
  • FIG1 is a flow chart of an interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference provided in the present application;
  • FIG2 is another schematic diagram of a flow chart of an interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference provided by the present application;
  • FIG3 is a schematic diagram showing a comparison of the segmentation of two pathological images from a cervical cancer dataset using the method of the present application and the existing method;
  • FIG4 is a schematic diagram showing a comparison of the segmentation of slices of different magnifications of the same original pathological image using the method provided by the present application and the existing method;
  • FIG5 is a schematic diagram of the structure of an interactive cell nucleus segmentation device combining Voronoi contour and weighted convex difference provided by the present application;
  • FIG6 is a schematic diagram of the structure of an electronic device provided in this application.
  • the pathology department of traditional hospitals takes a long time to complete the inspection report (usually about a week, CT, MRI, etc. are completed on the same day), and the pathology image size is large (tens of thousands * tens of thousands of pixels), and it needs to be viewed at different magnifications.
  • Manual inspection not only takes time and effort, but also damages vision and greatly reduces the efficiency of clinical diagnosis.
  • the present application provides an interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference, which makes full use of the prior information provided by pathologists and uses a new weighted convex difference model to make the segmentation result more accurate.
  • FIG. 1 and 2 show a schematic flow chart of an interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference provided in an embodiment of the present application.
  • an interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference may include:
  • S110 Acquire a pathological image, and mark the annotation points on the pathological image.
  • obtaining pathological images may include:
  • the original pathological slices are cropped into pathological images of a preset pixel size.
  • the number of marking points meets the preset conditions.
  • the marking points are evenly distributed in the pathological image according to the positions of the cell nuclei in the pathological image.
  • the above S110 and S120 are preprocessing stages (corresponding to the preprocessing stage in Figure 2), and the goal is to construct the Voronoi contour based on the key points marked by the pathologist as prior information.
  • the original pathological slices with larger pixels are first uniformly cropped into pathological images with a pixel size of 1024 ⁇ 1024.
  • the annotation points marked by doctors on the cell nucleus as prior information to build the Voronoi contour.
  • doctors manually mark the points we set these two ideal requirements to ensure that the constructed Voronoi contour is more suitable for cell nucleus segmentation: (1)
  • the density of the annotation points should be as appropriate as possible. Marking too many points will waste a lot of time and energy of pathologists and cause unnecessary waste of resources; if there are too few marking points, it will be detrimental to the next step of nuclear segmentation; Therefore, it is necessary to select an appropriate number of annotation points.
  • annotation points should be distributed as evenly as possible in the pathological image according to the position of the cell nucleus; if the distance between the annotation points is very close, it is likely that multiple points will be assimilated as the same point when constructing the Voronoi contour. In other words, points with too close annotation distances may become "invalid points". It is undeniable that due to the diversity of the density and position distribution of cell nuclei, it is difficult for us to avoid the above two problems in the annotation process.
  • a weighted convex difference model is a pre-built model.
  • the mathematical model building stage corresponds to the mathematical model building stage in FIG. 2
  • the weighted convex difference model is:
  • ⁇ is a parameter
  • f is the pathological image, i.e., the image to be segmented
  • c 1 ,c 2 ⁇ R is the average value of the pixels inside and outside the segmentation boundary during two-phase segmentation
  • ⁇ is the entire area of the pathological image
  • u is the segmentation result, which takes a value of 0 or 1.
  • 0 represents black, i.e., the background area
  • 1 represents white, i.e., the segmentation target area, i.e., the cell nucleus area
  • M() is the constructed optimization objective function
  • the first term ⁇ ⁇ ((fc 1 ) 2 u+(fc 2 ) 2 (1-u))dx is the fidelity term
  • the second term is the length regularization term, which is used to characterize the length of the segmented area
  • the third term The purpose of is to avoid excessive penalty caused by the second term, which leads to under-segmentation problem.
  • ⁇ (f) and ⁇ (f) are adaptive weights used to capture the edge information of the image.
  • ⁇ (f) and ⁇ (f) are specifically defined as:
  • ⁇ is the scale parameter
  • model (3) is approximately equivalent to the following minimization problem:
  • weighted convex difference model can also be replaced by other improved mathematical active contour models.
  • S140 Filter the final segmentation result according to the initial segmentation result.
  • step S130 we input the Voronoi contour obtained in step 1 as the initial contour into the optimized active contour model (weighted convex difference model) in step S130 for segmentation, and then determine the threshold by counting the area of each connected region in the form of a histogram, so as to eliminate some unnecessary areas in the segmentation result according to the threshold.
  • the post-processing stage corresponds to the post processing stage in Figure 2.
  • S140 screening the final segmentation result according to the initial segmentation result may include:
  • a histogram is created, where the horizontal axis of the histogram is the serial number of each cell nucleus and the vertical axis is the area;
  • the connected areas with an area smaller than the threshold are integrated into the background area to obtain the final segmentation result.
  • the interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference provided in the present application adopts an interactive semi-automatic method in which the pathologist marks several points to obtain more information about the image, increasing the subjective guidance of the pathologist, which makes the definition clearer and accurately solves the problem of the rough results obtained previously.
  • the interactive cell nucleus segmentation method provided by this application combines Voronoi contour and weighted convex difference, and proposes a new weighted convex difference mathematical model to improve the accuracy of segmentation.
  • the perimeter of the segmentation area is approximated by hot kernel convolution, and the proposed model is solved by a first-order linearization method to obtain an approximate solution of the model.
  • the interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference provided in this application explores a post-processing stage, which selects a suitable threshold according to the drawn histogram to remove redundant spots. It combines the advantages of the histogram and is not only simple and easy to operate, but also avoids the arbitrariness and blindness of the selection.
  • the interactive cell nucleus segmentation device combining Voronoi contour and weighted convex difference may include:
  • the acquisition and annotation module 510 is used to acquire the pathological image and annotate the annotation points on the pathological image;
  • a Voronoi contour construction module 520 is used to construct a Voronoi contour using the marked points as prior information
  • An initial segmentation module 530 is used to input the Voronoi contour as an initial contour into a pre-built weighted convex difference model to obtain an initial segmentation result;
  • the screening module 540 is used to screen the final segmentation result according to the initial segmentation result.
  • the acquisition and marking module 510 is further used for:
  • the original pathological slices are cropped into pathological images of a preset pixel size.
  • annotation points meet the following conditions:
  • the number of marking points meets the preset conditions.
  • the marking points are evenly distributed in the pathological image according to the positions of the cell nuclei in the pathological image.
  • the weighted convex difference model is:
  • ⁇ is a parameter
  • f is the pathological image
  • c 1 ,c 2 ⁇ R is the average value of the pixels inside and outside the segmentation boundary during two-phase segmentation
  • ⁇ is the entire area of the pathological image
  • u is the segmentation result
  • M() is the constructed optimization objective function
  • ⁇ (f) and ⁇ (f) are adaptive weights used to capture the edge information of the image.
  • the adaptive weights ⁇ (f) and ⁇ (f) are:
  • a weighted convex difference model including:
  • Hot kernel convolution is used to approximate the length regularization term:
  • ⁇ is the scale parameter
  • model (3) is approximately equivalent to the following minimization problem:
  • the screening module 540 is further configured to:
  • a histogram is created, where the horizontal axis of the histogram is the serial number of each cell nucleus and the vertical axis is the area;
  • the connected areas with an area smaller than the threshold are integrated into the background area to obtain the final segmentation result.
  • This embodiment provides an interactive cell nucleus segmentation device combining Voronoi contour and weighted convex difference, which can execute the embodiment of the above method. Its implementation principle and technical effect are similar and will not be repeated here.
  • Fig. 6 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. As shown in Fig. 6, a schematic diagram of the structure of an electronic device 600 suitable for implementing the embodiment of the present application is shown.
  • the electronic device 600 includes a central processing unit (CPU) 601, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 602 or a program loaded from a storage part 608 into a random access memory (RAM) 603.
  • ROM read-only memory
  • RAM random access memory
  • various programs and data required for the operation of the device 600 are also stored.
  • the CPU 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604.
  • An input/output (I/O) interface 605 is also connected to the bus 604.
  • the following components are connected to the I/O interface 605: an input section 606 including a keyboard, a mouse, etc.; an output section 607 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 608 including a hard disk, etc.; and a communication section 609 including a network interface card such as a LAN card, a modem, etc.
  • the communication section 609 performs communication processing via a network such as the Internet.
  • a drive 610 is also connected to the I/O interface 606 as needed.
  • a removable medium 611 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 610 as needed, so that a computer program read therefrom is installed into the storage section 608 as needed.
  • an embodiment of the present disclosure includes a computer program product, which includes a computer program tangibly contained on a machine-readable medium, and the computer program includes a program code for executing the above-mentioned interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference.
  • the computer program can be downloaded and installed from a network through the communication part 609, and/or installed from a removable medium 611.
  • Each square frame can represent a part of a module, a program segment, or a code, and the aforementioned module, program segment, or a part of the code comprises one or more executable instructions for realizing the specified logical function.
  • the functions marked in the square frame can also occur in a sequence different from that marked in the accompanying drawings. For example, two square frames represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved.
  • each square frame in the block diagram and/or flow chart, and the combination of the square frames in the block diagram and/or flow chart can be realized by a dedicated hardware-based system that performs the specified function or operation, or can be realized by a combination of dedicated hardware and computer instructions.
  • the units or modules involved in the embodiments described in the present application may be implemented by software or hardware.
  • the units or modules described may also be arranged in a processor.
  • the names of these units or modules do not constitute limitations on the units or modules themselves in certain circumstances.
  • a typical implementation device is a computer.
  • the computer may be, for example, a personal computer, a laptop, a mobile phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
  • the present application further provides a storage medium, which may be the storage medium included in the aforementioned device in the above embodiment; or may be a storage medium that exists independently and is not assembled into the device.
  • the storage medium stores one or more programs, and the aforementioned programs are used by one or more processors to execute the interactive cell nucleus segmentation method combining Voronoi contour and weighted convex difference described in the present application.
  • Storage media include permanent and non-permanent, removable and non-removable media that can implement information storage by any method or technology.
  • Information can be computer-readable instructions, data structures, program modules or other data.
  • Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device.
  • a computing device The machine-readable medium does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)
  • Image Analysis (AREA)

Abstract

本申请提供一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法，该方法包括：获取病理图像，并在所述病理图像上进行标注点的标注；利用所述标注点作为先验信息，构造维诺轮廓；将所述维诺轮廓作为初始轮廓输入预先构建的加权凸差分模型中，得到初始分割结果；根据所述初始分割结果，筛选最终分割结果。该分割方法分割更准确。

Description

一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法 技术领域
本发明属于图像分割技术领域，特别涉及一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法。
背景技术
宫颈癌是一种常见的妇科恶性肿瘤，发病率和死亡率都很高。世界卫生组织国际癌症研究所2020年12月发布的最新数据显示，每年宫颈癌新增病例近60.4万例，死亡病例超过34.1万例，其中80％以上的病例来自发展中国家。常规涂片(Conventional Smears，CS)技术是预防和早期发现宫颈癌的最简单和最重要的细胞学筛查方法之一。在传统的病理诊断过程中，病理学家在高倍显微镜下手动调整视野和大小以进行诊断。这种非常重复的观察和诊断过程不仅花费时间和精力，而且往往会损害视力并大大降低临床诊断的效率。因此，宫颈癌自动计算机辅助诊断(Computer-Aided Diagnosis，CAD)系统的开发是一个主要的临床问题。病理图像分割是宫颈癌分析的基础，它允许病理学家根据分割结果(如形状、大小和分布)观察到的细胞核特征做出诊断决定。然而，由于细胞的形状和染色的不均匀分布，细胞核的精确分割是一个巨大的挑战。
当前大多数方法解决细胞核分割问题时忽略了病理专家提供的先验信息。
发明内容
本说明书实施例的目的是提供一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法。
为解决上述技术问题，本申请实施例通过以下方式实现的：
第一方面，本申请提供一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法，该方法包括：
获取病理图像，并在病理图像上进行标注点的标注；
利用标注点作为先验信息，构造维诺轮廓；
将维诺轮廓作为初始轮廓输入预先构建的加权凸差分模型中，得到初始分割 结果；
根据初始分割结果，筛选最终分割结果。
在其中一个实施例中，获取病理图像，包括：
获取原始病理切片；
将原始病理切片裁剪成预设像素大小的病理图像。
在其中一个实施例中，标注点满足以下条件：
标记点的数量满足预设条件；且，
标记点依据病理图像中的细胞核的位置均匀地分布在病理图像中。
在其中一个实施例中，加权凸差分模型为：


其中，λ为参数，f为病理图像，c1,c2∈R为两相分割时分割边界内部和外部像素的平均值；Ω为病理图像的整个区域；u为分割结果；M()为构建的优化目标函数；α(f)和β(f)为自适应权重，用于捕捉图像的边缘信息。
在其中一个实施例中，自适应权重α(f)和β(f)分别为：

在其中一个实施例中，求解加权凸差分模型，包括：
热核卷积用于近似长度正则化项：
其中，*表示卷积，高斯函数Gτ的定义为：
其中，τ为尺度参数；
因此，模型(3)近似等于以下最小化问题：
对于问题(5)，由于第三项是凹的，通过一阶线性化在uk处展开目标函数ετ(u)；然后，将(5)中的问题u转化为以下问题：
其中且表示凹函数ετ(u)在uk处的次梯度，具体表示为：
其中Δ表示拉普拉斯算子，从方程(6)中，凸集上线性函数的最小点可以在其边界点处获得；最小值可以在以下条件下实现：
在其中一个实施例中，根据初始分割结果，筛选最终分割结果，包括：
对于初始分割结果，计算每个初始分割结果对应的连接区域的面积；
根据每个连接区域的面积，创建直方图，其中，直方图的横轴为每个细胞核的序号，纵轴为面积；
根据直方图的分布，选取阈值；
将面积小于阈值对应的连接区域整合为背景区域，得到最终分割结果。
第二方面，本申请提供一种结合维诺轮廓和加权凸差分的交互式细胞核分割装置，该装置包括：
获取及标注模块，用于获取病理图像，并在病理图像上进行标注点的标注；
维诺轮廓构造模块，用于利用标注点作为先验信息，构造维诺轮廓；
初始分割模块，用于将维诺轮廓作为初始轮廓输入预先构建的加权凸差分模型中，得到初始分割结果；
筛选模块，用于根据初始分割结果，筛选最终分割结果。
第三方面，本申请提供一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，处理器执行程序时实现如第一方面的结合维诺轮廓和加权凸差分的交互式细胞核分割方法。
第四方面，本申请提供一种可读存储介质，其上存储有计算机程序，该程序被处理器执行时实现如第一方面的结合维诺轮廓和加权凸差分的交互式细胞核分割方法。
由以上本说明书实施例提供的技术方案可见，该方案：采用了一种交互式的半自动方法，其中病理学家标记几个点以获得关于图像的更多信息，增加了病理学家的主观指导，这使得定义更清晰，并准确地解决了先前获得的粗略结果的问题。
提出了一个新的加权凸差数学模型，以提高分割的准确性。求解时采用热核卷积近似分割区域周长，并使用一阶线性化方法求解所提出的模型，以获得模型的近似解。
探索了一个后处理阶段，该阶段根据绘制的直方图选取合适的阈值，从而剔除多余的斑点，结合了直方图的优点，不仅简单、易于操作，而且避免了选择的任意性和盲目性。
附图说明
为了更清楚地说明本说明书实施例或现有技术中的技术方案，下面将对实施例或现有技术描述中所需要使用的附图作简单地介绍，显而易见地，下面描述中的附图仅仅是本说明书中记载的一些实施例，对于本领域普通技术人员来讲，在不付出创造性劳动性的前提下，还可以根据这些附图获得其他的附图。
图1为本申请提供的结合维诺轮廓和加权凸差分的交互式细胞核分割方法的流程示意图；
图2为本申请提供的结合维诺轮廓和加权凸差分的交互式细胞核分割方法的又一流程示意图；
图3为采用本申请方法与现有方法对来自宫颈癌数据集的两个病理图像的分割比较示意图；
图4为采用本申请提供的方法与现有方法对同一原始病理图像的不同放大倍数的切片的分割比较示意图；
图5为本申请提供的结合维诺轮廓和加权凸差分的交互式细胞核分割装置的结构示意图；
图6为本申请提供的电子设备的结构示意图。
具体实施方式
为了使本技术领域的人员更好地理解本说明书中的技术方案，下面将结合本说明书实施例中的附图，对本说明书实施例中的技术方案进行清楚、完整地描述，显然，所描述的实施例仅仅是本说明书一部分实施例，而不是全部的实施例。基于本说明书中的实施例，本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例，都应当属于本说明书保护的范围。
以下描述中，为了说明而不是为了限定，提出了诸如特定系统结构、技术之类的具体细节，以便透彻理解本申请实施例。然而，本领域的技术人员应当清楚，在没有这些具体细节的其它实施例中也可以实现本申请。在其它情况中，省略对众所周知的系统、装置、电路以及方法的详细说明，以免不必要的细节妨碍本申请的描述。
在不背离本申请的范围或精神的情况下，可对本申请说明书的具体实施方式做多种改进和变化，这对本领域技术人员而言是显而易见的。由本申请的说明书得到的其他实施方式对技术人员而言是显而易见得的。本申请说明书和实施例仅是示例性的。
关于本文中所使用的“包含”、“包括”、“具有”、“含有”等等，均为开放性的用语，即意指包含但不限于。
传统医院的病理科完成报告检查时间久(通常一周左右，CT、MRI等当天完成)，并且病理图像尺寸较大(几万*几万像素)，而且需要在不同放大倍数下 进行人工检查，不仅花费时间和精力，而且会损害视力并大大降低临床诊断的效率。
而基于深度学习的分割方法大多需要大量的人工标记样本，否则，分割结果不能正确识别到细胞核边缘。
另外，大部分的传统分割模型不能充分利用病理专家提供的先验信息。
基于上述缺陷，本申请提供一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法，充分利用病理专家提供的先验信息，并且使用新的加权凸差分模型，使得分割结果更准确。
下面结合附图和实施例对本发明进一步详细说明。
参照图1和图2，其示出了适用于本申请实施例提供的结合维诺轮廓和加权凸差分的交互式细胞核分割方法的流程示意图。
如图1和图2所示，一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法，可以包括：
S110、获取病理图像，并在病理图像上进行标注点的标注。
其中，获取病理图像，可以包括：
获取原始病理切片；
将原始病理切片裁剪成预设像素大小的病理图像。
其中，标注点满足以下条件：
标记点的数量满足预设条件；且，
标记点依据病理图像中的细胞核的位置均匀地分布在病理图像中。
S120、利用标注点作为先验信息，构造维诺轮廓。
具体的，上述S110、S120为预处理阶段(对应图2preprocessing stage)，目标是根据病理专家标记的关键点作为先验信息，构造维诺轮廓。
在预处理阶段，首先把较大像素的原始病理切片统一裁剪成像素大小为1024×1024的病理图像。为了充分利用病理图像中的信息，我们借助医生在细胞核上标注的标注点作为先验信息从而搭建出维诺轮廓。医生在进行人工标注点时我们设置了这两点理想的要求来确保搭建出的维诺轮廓更适合用于细胞核分割：(1)标注点的密度应尽可能适当。标记过多的点会浪费病理学家的大量时间和精力，并造成不必要的资源浪费；如果标记点太少，将不利于下一步的核分割； 因此，选取合适数量的标注点是十分有必要的。(2)标注点尽量依据细胞核的位置较为均匀地分布在病理图像中；如果标注点的距离十分接近，那么很可能会造成搭建维诺轮廓时多个点会被当作同一个点同化掉，也就是说，标注距离过于近的点可能会成为“无效点”。不可否认的是，由于细胞核的密度和位置分布的多样性，我们难以避免在标注过程中会有上述两点问题的存在。
S130、将维诺轮廓作为初始轮廓输入预先构建的加权凸差分模型中，得到初始分割结果。
具体的，加权凸差分模型是预先构建的模型。
一个实施例中，数学模型构建阶段对应图2中mathematical model building stage，加权凸差分模型为：


其中，λ为参数，f为病理图像即待分割图像，c1,c2∈R为两相分割时分割边界内部和外部像素的平均值；Ω为病理图像的整个区域；u为分割结果，取值为0或1，示例性的，0代表黑色，即背景区域，1代表白色，即分割目标区域，也即细胞核区域；M()为构建的优化目标函数；在式(3)中，第一项λ∫Ω((f-c1)2u+(f-c2)2(1-u))dx为保真项；第二项为长度正则项，用来刻画分割区域的长度；第三项的作用是为了避免第二项造成的过度惩罚，导致欠分割问题的产生。α(f)和β(f)为自适应权重，用于捕捉图像的边缘信息。
其中，α(f)和β(f)具体定义为：

求解上述模型即可得到整个分割框架中的分割结果u。
但是，对于问题(3)，很难找到其最优解，因此我们可以使用凸松弛方法u∈[0，1]来替换二进制值约束。为了降低模型(3)的计算复杂性，热核卷积用于近似长度正则化项，即：
其中，*表示卷积，高斯函数Gτ的定义为：
其中，τ为尺度参数；
因此，模型(3)近似等于以下最小化问题：
对于问题(5)，由于第三项是凹的，通过一阶线性化在uk处展开目标函数ετ(u)；然后，将(5)中的问题u转化为以下问题：
其中且表示凹函数ετ(u)在uk处的次梯度，具体表示为：
其中Δ表示拉普拉斯算子，从方程(6)中，凸集上线性函数的最小点可以在其边界点处获得；最小值可以在以下条件下实现：
基于模型(1)-(3)的图像分割算法如下所示。
可以理解的，上述加权凸差分模型还可以更换为其他改进的数学活动轮廓模型。
S140、根据初始分割结果，筛选最终分割结果。
具体的，为了得到更准确地分割结果，我们将步骤1中得到的维诺轮廓作为初始轮廓输入步骤S130中的优化的活动轮廓模型(加权凸差分模型)中进行分割后，通过直方图的形式统计出每个连通区域的面积来确定阈值，从而根据阈值来剔除分割结果中一些不必要的区域，后处理阶段对应图2中post processing stage。
一个实施例中，S140根据初始分割结果，筛选最终分割结果可以包括：
对于初始分割结果，计算每个初始分割结果对应的连接区域的面积；
根据每个连接区域的面积，创建直方图，其中，直方图的横轴为每个细胞核的序号，纵轴为面积；
根据直方图的分布，选取阈值；
将面积小于阈值对应的连接区域整合为背景区域，得到最终分割结果。
具体的，首先进行执行区域的统计，对于初始分割结果u，计算每个连接区域的面积，然后创建直方图，其中，直方图以每个细胞核的序号为横轴，面积为纵轴，最后确定阈值δ，通过观察直方图的分布，选择合适的阈值，以便将小于δ的所有区域整合到背景区域中。
遵循上述过程，根据阈值δ去除分割结果中的一些不必要区域。该方法不仅简单易行，而且避免了阈值选择的随意性和盲目性。
本申请提供的结合维诺轮廓和加权凸差分的交互式细胞核分割方法，采用了一种交互式的半自动方法，其中病理学家标记几个点以获得关于图像的更多信息，增加了病理学家的主观指导，这使得定义更清晰，并准确地解决了先前获得的粗略结果的问题。
本申请提供的结合维诺轮廓和加权凸差分的交互式细胞核分割方法，提出了一个新的加权凸差数学模型，以提高分割的准确性。求解时采用热核卷积近似分割区域周长，并使用一阶线性化方法求解所提出的模型，以获得模型的近似解。
本申请提供的结合维诺轮廓和加权凸差分的交互式细胞核分割方法，探索了一个后处理阶段，该阶段根据绘制的直方图选取合适的阈值，从而剔除多余的斑点，结合了直方图的优点，不仅简单、易于操作，而且避免了选择的任意性和盲目性。
实验对比
如图3所示为采用本申请提供的方法与现有方法(包括OSTU、CV、WBHV、ICTM)对来自宫颈癌数据集的两个病理图像的分割比较。其中，ROI由橙色和青色框标记，从分割结果可以看出，采用本申请方法分割更准确。
如图4所示为采用本申请提供的方法与现有方法(包括OSTU、CV、WBHV、ICTM)对同一原始病理图像的不同放大倍数的切片的分割比较。其中，ROI由橙色框标记，从分割结果可以看出，采用本申请方法分割更准确。
参照图5，其示出了根据本申请一个实施例描述的结合维诺轮廓和加权凸差 分的交互式细胞核分割装置的结构示意图。
如图5所示，结合维诺轮廓和加权凸差分的交互式细胞核分割装置，可以包括：
获取及标注模块510，用于获取病理图像，并在病理图像上进行标注点的标注；
维诺轮廓构造模块520，用于利用标注点作为先验信息，构造维诺轮廓；
初始分割模块530，用于将维诺轮廓作为初始轮廓输入预先构建的加权凸差分模型中，得到初始分割结果；
筛选模块540，用于根据初始分割结果，筛选最终分割结果。
可选的，获取及标注模块510还用于：
获取原始病理切片；
将原始病理切片裁剪成预设像素大小的病理图像。
可选的，标注点满足以下条件：
标记点的数量满足预设条件；且，
标记点依据病理图像中的细胞核的位置均匀地分布在病理图像中。
可选的，加权凸差分模型为：


其中，λ为参数，f为病理图像，c1,c2∈R为两相分割时分割边界内部和外部像素的平均值；Ω为病理图像的整个区域；u为分割结果；M()为构建的优化目标函数；α(f)和β(f)为自适应权重，用于捕捉图像的边缘信息。
可选的，自适应权重α(f)和β(f)分别为：

可选的，求解加权凸差分模型，包括：
热核卷积用于近似长度正则化项：
其中，*表示卷积，高斯函数Gτ的定义为：
其中，τ为尺度参数；
因此，模型(3)近似等于以下最小化问题：
对于问题(5)，由于第三项是凹的，通过一阶线性化在uk处展开目标函数ετ(u)；然后，将(5)中的问题u转化为以下问题：
其中且表示凹函数ετ(u)在uk处的次梯度，具体表示为：
其中Δ表示拉普拉斯算子，从方程(6)中，凸集上线性函数的最小点可以在其边界点处获得；最小值可以在以下条件下实现：
可选的，筛选模块540，还用于：
对于初始分割结果，计算每个初始分割结果对应的连接区域的面积；
根据每个连接区域的面积，创建直方图，其中，直方图的横轴为每个细胞核的序号，纵轴为面积；
根据直方图的分布，选取阈值；
将面积小于阈值对应的连接区域整合为背景区域，得到最终分割结果。
本实施例提供的一种结合维诺轮廓和加权凸差分的交互式细胞核分割装置，可以执行上述方法的实施例，其实现原理和技术效果类似，在此不再赘述。
图6为本发明实施例提供的一种电子设备的结构示意图。如图6所示，示出了适于用来实现本申请实施例的电子设备600的结构示意图。
如图6所示，电子设备600包括中央处理单元(CPU)601，其可以根据存储在只读存储器(ROM)602中的程序或者从存储部分608加载到随机访问存储器(RAM)603中的程序而执行各种适当的动作和处理。在RAM 603中，还存储有设备600操作所需的各种程序和数据。CPU 601、ROM 602以及RAM 603通过总线604彼此相连。输入/输出(I/O)接口605也连接至总线604。
以下部件连接至I/O接口605：包括键盘、鼠标等的输入部分606；包括诸如阴极射线管(CRT)、液晶显示器(LCD)等以及扬声器等的输出部分607；包括硬盘等的存储部分608；以及包括诸如LAN卡、调制解调器等的网络接口卡的通信部分609。通信部分609经由诸如因特网的网络执行通信处理。驱动器610也根据需要连接至I/O接口606。可拆卸介质611，诸如磁盘、光盘、磁光盘、半导体存储器等等，根据需要安装在驱动器610上，以便于从其上读出的计算机程序根据需要被安装入存储部分608。
特别地，根据本公开的实施例，上文参考图1描述的过程可以被实现为计算机软件程序。例如，本公开的实施例包括一种计算机程序产品，其包括有形地包含在机器可读介质上的计算机程序，计算机程序包含用于执行上述结合维诺轮廓和加权凸差分的交互式细胞核分割方法的程序代码。在这样的实施例中，该计算机程序可以通过通信部分609从网络上被下载和安装，和/或从可拆卸介质611被安装。
附图中的流程图和框图，图示了按照本发明各种实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的 每个方框可以代表一个模块、程序段、或代码的一部分，前述模块、程序段、或代码的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。也应当注意，在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个接连地表示的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和/或流程图中的每个方框、以及框图和/或流程图中的方框的组合，可以用执行规定的功能或操作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。
描述于本申请实施例中所涉及到的单元或模块可以通过软件的方式实现，也可以通过硬件的方式来实现。所描述的单元或模块也可以设置在处理器中。这些单元或模块的名称在某种情况下并不构成对该单元或模块本身的限定。
上述实施例阐明的系统、装置、模块或单元，具体可以由计算机芯片或实体实现，或者由具有某种功能的产品来实现。一种典型的实现设备为计算机。具体的，计算机例如可以为个人计算机、笔记本电脑、行动电话、智能电话、个人数字助理、媒体播放器、导航设备、电子邮件设备、游戏控制台、平板计算机、可穿戴设备或者这些设备中的任何设备的组合。
作为另一方面，本申请还提供了一种存储介质，该存储介质可以是上述实施例中前述装置中所包含的存储介质；也可以是单独存在，未装配入设备中的存储介质。存储介质存储有一个或者一个以上程序，前述程序被一个或者一个以上的处理器用来执行描述于本申请的结合维诺轮廓和加权凸差分的交互式细胞核分割方法。
存储介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括，但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带，磁带磁磁盘存储或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。按照本文中的界定，计算 机可读介质不包括暂存电脑可读媒体(transitory media)，如调制的数据信号和载波。
需要说明的是，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、商品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、商品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括要素的过程、方法、商品或者设备中还存在另外的相同要素。
本说明书中的各个实施例均采用递进的方式描述，各个实施例之间相同相似的部分互相参见即可，每个实施例重点说明的都是与其他实施例的不同之处。尤其，对于系统实施例而言，由于其基本相似于方法实施例，所以描述的比较简单，相关之处参见方法实施例的部分说明即可。

Claims (10)

  1. 一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法，其特征在于，所述方法包括：
    获取病理图像，并在所述病理图像上进行标注点的标注；
    利用所述标注点作为先验信息，构造维诺轮廓；
    将所述维诺轮廓作为初始轮廓输入预先构建的加权凸差分模型中，得到初始分割结果；
    根据所述初始分割结果，筛选最终分割结果。
  2. 根据权利要求1所述的方法，其特征在于，所述获取病理图像，包括：
    获取原始病理切片；
    将所述原始病理切片裁剪成预设像素大小的病理图像。
  3. 根据权利要求1所述的方法，其特征在于，所述标注点满足以下条件：
    所述标记点的数量满足预设条件；且，
    所述标记点依据所述病理图像中的细胞核的位置均匀地分布在所述病理图像中。
  4. 根据权利要求1所述的方法，其特征在于，所述加权凸差分模型为：


    其中，λ为参数，f为病理图像，c1,c2∈R为两相分割时分割边界内部和外部像素的平均值；Ω为病理图像的整个区域；u为分割结果；M()为构建的优化目标函数；α(f)和β(f)为自适应权重，用于捕捉图像的边缘信息。
  5. 根据权利要求4所述的方法，其特征在于，所述自适应权重α(f)和β(f)分别为：

  6. 根据权利要求5所述的方法，其特征在于，求解所述加权凸差分模型，包括：
    热核卷积用于近似长度正则化项：
    其中，*表示卷积，高斯函数Gτ的定义为：
    其中，τ为尺度参数；
    因此，模型(3)近似等于以下最小化问题：
    对于问题(5)，由于第三项是凹的，通过一阶线性化在uk处展开目标函数ετ(u)；然后，将(5)中的问题u转化为以下问题：
    其中且表示凹函数ετ(u)在uk处的次梯度，具体表示为：
    其中Δ表示拉普拉斯算子，从方程(6)中，凸集上线性函数的最小点可以在其边界点处获得；最小值可以在以下条件下实现：
  7. 根据权利要求1所述的方法，其特征在于，所述根据所述初始分割结果，筛选最终分割结果，包括：
    对于所述初始分割结果，计算每个所述初始分割结果对应的连接区域的面积；
    根据每个连接区域的面积，创建直方图，其中，所述直方图的横轴为每个细胞核的序号，纵轴为面积；
    根据所述直方图的分布，选取阈值；
    将面积小于所述阈值对应的连接区域整合为背景区域，得到最终分割结果。
  8. 一种结合维诺轮廓和加权凸差分的交互式细胞核分割装置，其特征在于，所述装置包括：
    获取及标注模块，用于获取病理图像，并在所述病理图像上进行标注点的标注；
    维诺轮廓构造模块，用于利用所述标注点作为先验信息，构造维诺轮廓；
    初始分割模块，用于将所述维诺轮廓作为初始轮廓输入预先构建的加权凸差分模型中，得到初始分割结果；
    筛选模块，用于根据所述初始分割结果，筛选最终分割结果。
  9. 一种电子设备，包括存储器、处理器及存储在存储器上并可在处理器上运行的计算机程序，其特征在于，所述处理器执行所述程序时实现如权利要求1-7中任一所述的结合维诺轮廓和加权凸差分的交互式细胞核分割方法。
  10. 一种可读存储介质，其上存储有计算机程序，其特征在于，该程序被处理器执行时实现如权利要求1-7中任一所述的结合维诺轮廓和加权凸差分的交互式细胞核分割方法。
PCT/CN2023/087890 2023-04-12 2023-04-12 一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法 Ceased WO2024212136A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2023/087890 WO2024212136A1 (zh) 2023-04-12 2023-04-12 一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2023/087890 WO2024212136A1 (zh) 2023-04-12 2023-04-12 一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法

Publications (1)

Publication Number Publication Date
WO2024212136A1 true WO2024212136A1 (zh) 2024-10-17

Family

ID=93058559

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2023/087890 Ceased WO2024212136A1 (zh) 2023-04-12 2023-04-12 一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法

Country Status (1)

Country Link
WO (1) WO2024212136A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110274336A1 (en) * 2010-03-12 2011-11-10 Institute For Medical Informatics Optimizing the initialization and convergence of active contours for segmentation of cell nuclei in histological sections
CN107256558A (zh) * 2017-05-18 2017-10-17 深思考人工智能机器人科技（北京）有限公司 一种无监督式的宫颈细胞图像自动分割方法及系统
CN108364288A (zh) * 2018-03-01 2018-08-03 北京航空航天大学 用于乳腺癌病理图像的分割方法和装置
CN113570628A (zh) * 2021-07-30 2021-10-29 西安科技大学 一种基于活动轮廓模型的白细胞分割方法
CN115457054A (zh) * 2022-09-13 2022-12-09 深圳先进技术研究院 图像分割方法、装置、设备及可读存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110274336A1 (en) * 2010-03-12 2011-11-10 Institute For Medical Informatics Optimizing the initialization and convergence of active contours for segmentation of cell nuclei in histological sections
CN107256558A (zh) * 2017-05-18 2017-10-17 深思考人工智能机器人科技（北京）有限公司 一种无监督式的宫颈细胞图像自动分割方法及系统
CN108364288A (zh) * 2018-03-01 2018-08-03 北京航空航天大学 用于乳腺癌病理图像的分割方法和装置
CN113570628A (zh) * 2021-07-30 2021-10-29 西安科技大学 一种基于活动轮廓模型的白细胞分割方法
CN115457054A (zh) * 2022-09-13 2022-12-09 深圳先进技术研究院 图像分割方法、装置、设备及可读存储介质

Similar Documents

Publication Publication Date Title
Mohanakurup et al. Breast cancer detection on histopathological images using a composite dilated Backbone Network
US12354387B2 (en) Identifying regions of interest from whole slide images
Salvi et al. Multi-tissue and multi-scale approach for nuclei segmentation in H&E stained images
CN114155365B (zh) 模型训练方法、图像处理方法及相关装置
WO2020114421A1 (zh) 对象检测模型训练方法、目标对象检测方法
De Bel et al. Automatic segmentation of histopathological slides of renal tissue using deep learning
Zhang et al. DCCL: A benchmark for cervical cytology analysis
Gadermayr et al. CNN cascades for segmenting whole slide images of the kidney
WO2020253508A1 (zh) 异常细胞检测方法、装置及计算机可读存储介质
CN113724267B (zh) 一种乳腺超声图像肿瘤分割方法及装置
CN116012332B (zh) 一种基于层次图的病理图像中原发肿瘤分期多示例学习方法、系统、设备及介质
Koyuncu et al. Object‐oriented segmentation of cell nuclei in fluorescence microscopy images
CN116580397A (zh) 病理图像识别方法、装置、设备及存储介质
CN115564949B (zh) 密集场景下的图像检测方法、装置、终端及介质
WO2024212136A1 (zh) 一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法
CN115937065A (zh) 显示模组的异物检测方法、装置、设备及存储介质
Suganya et al. Performance analysis of various filters for denoising breast cancer histopathology images
CN116664482A (zh) 一种结合维诺轮廓和加权凸差分的交互式细胞核分割方法
CN111861916A (zh) 病理图片的处理方法
Sakhaei et al. Segmentation of digital images with region growing algorithm
Yan et al. Hsdet: A representative sampling based object detector in cervical cancer cell images
CN116758068A (zh) 一种基于人工智能的骨髓图片细胞形态分析方法
Bisson et al. A high-precision hierarchical registration approach for stain-and scanner-independent colocalization on whole slide images in histopathology
Yang et al. An interactive nuclei segmentation framework with Voronoi diagrams and weighted convex difference for cervical cancer pathology images
Yang et al. Cervical nuclei segmentation in whole slide histopathology images using convolution neural network

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: 23932428

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE