Cervical cytopathy degree scoring method based on TBS diagnostic standard quantification

Abstract

The invention discloses a cervical cytopathy degree scoring method based on TBS diagnostic standard quantification, and relates to the problem that the cytopathy degree is difficult to automatically detect in a cervical cytopathic auxiliary diagnosis technology. According to the cervical cell pathology auxiliary diagnosis technology, the overall lesion condition of the cervical cell smear is automatically evaluated through the image processing technology. However, a traditional scheme depends on implicit feature extraction based on an encoder, the interpretability is poor, and evaluation indexes and effective clinical evidences accepted by pathologists are difficult to provide. In order to improve the problem, the invention provides a cervical cell lesion degree scoring method based on TBS diagnostic standard quantification. Experiments prove that the method can be used for effectively extracting lesion characteristics such as nuclear cytoplasmic ratio, nuclear deep staining and nuclear roundness of cells, and effectively scoring the degree of cytopathy according to the characteristics. The method is applied to cervical cell lesion feature extraction and pathological interpretability lesion degree scoring.

Description

Technical Field

[0001] This invention relates to a scoring method for quantifying the severity of cervical cellular lesions using the TBS diagnostic criteria. Background Technology

[0002] Cervical cancer is a major malignant tumor threatening women's health, and it is the only cancer with a clearly defined cause that is preventable and controllable through vaccines and screening. According to the 2018 Global Cancer Report released by the World Health Organization / International Agency for Research on Cancer, there were 570,000 new cases of cervical cancer and 311,000 deaths in 2018, ranking fourth among the causes of cancer incidence and death in women. According to the 2022 China Cancer Report, the incidence and mortality rates of cervical cancer among Chinese women are showing a year-on-year increase and a trend towards younger ages. Since 2018, the WHO has repeatedly called for the global elimination of cervical cancer. In 2023, ten departments, including the National Health Commission, jointly released the "Action Plan for Accelerating the Elimination of Cervical Cancer (2023-2030)," proposing a Chinese goal of achieving a 50% cervical cancer screening rate among women of appropriate age by 2025 and 70% by 2030. The cure rate for early-stage cancer is as high as 100%, while the rate for late-stage cancer is less than 10%. Early detection and standardized treatment are key to reducing the incidence and mortality rates of cervical cancer.

[0003] Pathological examination is currently a clinically significant diagnostic method, including cytopathological diagnosis and histopathological diagnosis. Cytopathological diagnosis involves collecting exfoliated human cells, preparing stained slides, and then having a pathologist make a diagnosis by observing abnormal cells under a microscope according to the TBS standard. Cytopathological diagnosis is characterized by simple sample collection and wide application, and is particularly suitable for the early diagnosis and screening of cervical cancer, and has been widely promoted. However, traditional cytopathological diagnosis relies entirely on doctors' "manual operation and visual observation," which has three major drawbacks: (1) the accuracy rate of pathologists' diagnoses is low, resulting in a large number of misdiagnoses and missed diagnoses; (2) in frontline work, pathologists have subjective biases in their understanding of the diagnostic indicators described in the TBS standard, leading to different doctors having different diagnostic results for the same cell; (3) the workload of pathological diagnosis is large, and it is difficult to increase the number of diagnoses. This makes it difficult to achieve early detection and early treatment of cervical cancer.

[0004] This invention aims to solve the problem of quantifying the degree of cervical cell lesions in the TBS diagnostic standard. It designs cytopathological parameters by combining expert diagnostic experience to measure the degree of cervical cell lesions, and finally obtains reliable and accurate cervical cell lesion degree score results. Summary of the Invention

[0005] The purpose of this invention is to address the problem of pathologists' misunderstanding of the TBS standard by proposing a quantitative scoring method for the degree of cervical cell lesions based on the TBS diagnostic standard.

[0006] The aforementioned objectives are primarily achieved through the following technical solutions:

[0007] S1. Data Reading: Read cell bounding box image, cell edge pixel coordinate sequence, cell nucleus edge pixel coordinate sequence, and effective nucleus count. and the area of ​​normal middle layer cell nuclei;

[0008] S2. Calculation of basic cell features: Calculate cell area, cell nucleus area, cell nucleus perimeter, cell nucleus mask image, cytoplasmic pixel set, cell nucleus outer contour pixel set, and cell outer contour pixel set;

[0009] The method for calculating the cytoplasmic pixel set is as follows:

[0010]

[0011] in, Represents a set of cytoplasmic pixels. A polygon representing the sequence of pixel coordinates at the cell edges. A polygon representing the pixel coordinate sequence of the cell nucleus edge. Grayscale processing.

[0012] The method for calculating the pixel set of the outer contour of the cell nucleus is as follows:

[0013]

[0014] in, Represents the set of pixels representing the outer contour of the cell nucleus. This represents the equidistant indentation operation of a polygon. Find the intersection of polygons. Representing the indentation, implemented as follows:

[0015]

[0016] in, Calculate the diameter of a polygon. This is the scaling factor.

[0017] The method for calculating the pixel set of the outer contour of a cell is as follows:

[0018]

[0019] in, Represents the set of pixels representing the outer contour of a cell. The polygon represents the outline of the cell's bounding box image. To divide the tolerance.

[0020] S3. Calculate the nuclear enlargement ratio and describe the ratio of the cell nuclear area to the normal median cell nuclear area;

[0021] The larger the nuclear area, the greater the nuclear enlargement ratio. The method for calculating the nuclear enlargement ratio is as follows:

[0022]

[0023] in, This represents the increase in the proportion of the nucleus. Represents the area of ​​the cell nucleus in pixels. This represents the normal mid-layer kernel area in pixels.

[0024] S4. Calculate the nucleocytoplasmic ratio, describing the ratio of the cell's nuclear area to its cytoplasmic area;

[0025] The larger the area of ​​the cell nucleus in the cytoplasm, the larger the nucleocytosis ratio. The nucleocytosis ratio is calculated as follows:

[0026]

[0027] in, Represents the nucleus-to-cylinder ratio, Represents cell area in pixels.

[0028] S5. Calculate nuclear roundness, which describes the smoothness of the cell nuclear boundary;

[0029] The closer the cell nucleus is to a circle, the closer its roundness value is to 1; otherwise, it is close to 0. The method for calculating nuclear roundness is as follows:

[0030]

[0031] in, Represents the roundness of the kernel. This represents the perimeter of the cell nucleus, expressed in pixels.

[0032] S6. Calculate chromatin roughness to describe the uniformity of nuclear chromatin distribution as described in the TBS diagnostic criteria.

[0033] The coarser the chromatin, the higher its roughness value. The method for calculating chromatin roughness is as follows:

[0034]

[0035] in, Represents chromatin roughness. Represents a cell nuclear mask image. For the calculation of the gray-level co-occurrence matrix, For matrix standardization, To calculate entropy.

[0036] The gray-level co-occurrence matrix (GLCM) is a statistical method for describing the texture of an image. This matrix is ​​two-dimensional, where each element... This represents the grayscale value at a specific direction and distance. and grayscale value The number of times they occur simultaneously.

[0037] Matrix standardization is implemented as follows:

[0038]

[0039] in This is the matrix to be standardized.

[0040] The entropy calculation is implemented as follows:

[0041]

[0042] in Let be the entropy matrix to be determined.

[0043] S7. Calculate chromatin granularity and describe the degree of nuclear chromatin granularity as described in the TBS diagnostic criteria.

[0044] The larger the chromatin staining patch, the larger the chromatin granularity value. The method for calculating chromatin granularity is as follows:

[0045]

[0046] in, Represents chromatin granularity. A polygon representing the outline of the cell nucleus. Representative image Given the side length of the largest patch at the midpoint of the dividing line, the implementation is as follows:

[0047]

[0048] in, The difference in brightness is in the vertical direction. The difference in brightness in the horizontal direction is defined as follows:

[0049]

[0050]

[0051] in, For sampling images, The bounding box sampling function is calculated as follows:

[0052]

[0053] in, The coordinates of the sampling center are... This represents the side length of the sampling box.

[0054] S8. Calculate the perinuclear halo degree and describe the degree of significance of the perinuclear halo as described in the TBS diagnostic criteria;

[0055] The more pronounced the perinuclear halo, the greater the perinuclear halo value. The calculation method for perinuclear halo value is as follows:

[0056]

[0057] in, Represents the perinuclear halo degree, The threshold for truncating the brightness of pixels on the outer contour of the cell nucleus. The threshold for truncating the brightness of cytoplasmic pixels. The truncated mean operation is implemented as follows:

[0058]

[0059] in, For the set of operands of the operation, To truncate the upper bound proportion, To truncate the lower bound ratio, This represents a descending order of an ordered set.

[0060] S9. Calculate the voidness to describe the severity of hollowing-out cell formation as described in the TBS diagnostic criteria.

[0061] The larger the largest cavity in a cell, the greater the porosity. The porosity is calculated as follows:

[0062]

[0063] in, Represents the degree of voidness. The set of side lengths for cell cavities is calculated as follows:

[0064]

[0065] in, Image representing a cell bounding box. The cytoplasmic brightness threshold is calculated as follows:

[0066]

[0067] in, Represents the set of pixels representing the outer contour of a cell. The threshold for truncating the brightness of pixels on the outer contour of the cell.

[0068] S10. Calculate the cytopathic effect score and describe the degree of cytopathic effect based on the obtained cell parameters and the normal value of the item.

[0069] The formula for calculating the cytopathic effect score is as follows:

[0070]

[0071] in, The scoring weights corresponding to their respective pathological parameters are taken here. , The normalized bias calculated for each pathological parameter The calculation method is as follows ,the remaining The calculation method is consistent, with The calculation method for example is as follows:

[0072]

[0073] in, These represent the baseline values ​​for their respective pathological parameters. and These represent the lower and upper bounds of the normal range for their respective pathological parameters.

[0074] Invention Effects

[0075] This invention provides a method for scoring the degree of cervical cellular lesions using the TBS diagnostic standard. The method first receives basic cellular information, including cell bounding box images, cell edge pixel coordinate sequences, and cell nucleus edge pixel coordinate sequences, providing geometric information for further calculation of pathological parameters. Then, based on the basic cellular information, intermediate pathological parameters such as cell area, cell nucleus area, cell nucleus perimeter, cell nucleus mask image, and cytoplasmic pixel set are calculated. Next, pathological parameters are calculated for the cells. Finally, the cellular lesion score is calculated based on the weighted sum of the normalized deviations of the pathological parameters from normal values. Experiments show that this method can effectively distinguish cervical cells with different degrees of lesion, providing an effective auxiliary tool for pathologists. Attached Figure Description

[0076] Figure 1 Illustration of the cervical cell lesion severity scoring process;

[0077] Figure 2 SIPaKMeD dataset benchmark-lesion cell classification PR curve;

[0078] Figure 3 Distribution of cellular lesion scores for different lesion degrees in the SIPaKMeD dataset; Specific implementation methods Specific implementation method one:

[0080] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0081] like Figure 1 As shown, the TBS diagnostic criteria for quantifying the severity of cervical cellular lesions provided in this article includes the following steps:

[0082] S1. Data Reading: Read cell bounding box image, cell edge pixel coordinate sequence, cell nucleus edge pixel coordinate sequence, and effective nucleus count. and the area of ​​normal middle layer cell nuclei;

[0083] S2. Calculation of basic cell features: Calculate cell area, cell nucleus area, cell nucleus perimeter, cell nucleus mask image, cytoplasmic pixel set, cell nucleus outer contour pixel set, and cell outer contour pixel set;

[0084] S3. Calculate the nuclear enlargement ratio and describe the ratio of the cell nuclear area to the normal median cell nuclear area;

[0085] S4. Calculate the nucleocytoplasmic ratio, describing the ratio of the cell's nuclear area to its cytoplasmic area;

[0086] S5. Calculate nuclear roundness, which describes the smoothness of the cell nuclear boundary;

[0087] S6. Calculate chromatin roughness to describe the uniformity of nuclear chromatin distribution as described in the TBS diagnostic criteria.

[0088] S7. Calculate chromatin granularity and describe the degree of nuclear chromatin granularity as described in the TBS diagnostic criteria.

[0089] S8. Calculate the perinuclear halo degree and describe the degree of significance of the perinuclear halo as described in the TBS diagnostic criteria;

[0090] S9. Calculate the voidness to describe the severity of hollowing-out cell formation as described in the TBS diagnostic criteria.

[0091] S10. Calculate the cytopathic effect score and describe the degree of cytopathic effect based on the obtained cell parameters and the normal value of the item.

[0092] This invention first receives basic cellular information, then calculates intermediate pathological parameters based on the basic cellular information, and then calculates the pathological parameters of individual cells. Finally, it calculates the cellular lesion score based on the weighted sum of the normalized deviations of the pathological parameters from normal values.

[0093] The embodiments of the present invention will be described in detail below:

[0094] This invention uses a cell bounding box image, its nucleoplasmic contour information, and the nuclear area of ​​a normal mid-layer cell to apply the algorithm of this invention to achieve a quantitative cervical cell lesion score based on the TBS diagnostic standard, as follows.

[0095] like Figure 1 The training of the model shown includes the following steps:

[0096] S1. Data Reading: Read cell bounding box image, cell edge pixel coordinate sequence, cell nucleus edge pixel coordinate sequence, and effective nucleus count. and the area of ​​normal middle layer cell nuclei;

[0097] S2. Calculation of basic cell features: Calculate cell area, cell nucleus area, cell nucleus perimeter, cell nucleus mask image, cytoplasmic pixel set, cell nucleus outer contour pixel set, and cell outer contour pixel set;

[0098] The method for calculating the cytoplasmic pixel set is as follows:

[0099]

[0100] in, Represents a set of cytoplasmic pixels. A polygon representing the sequence of pixel coordinates at the cell edges. A polygon representing the pixel coordinate sequence of the cell nucleus edge. Grayscale processing.

[0101] The method for calculating the pixel set of the outer contour of the cell nucleus is as follows:

[0102]

[0103] in, Represents the set of pixels representing the outer contour of the cell nucleus. This represents the equidistant indentation operation of a polygon. Find the intersection of polygons. Representing the indentation, implemented as follows:

[0104]

[0105] in, Calculate the diameter of a polygon. This is the scaling factor, taken here. .

[0106] The method for calculating the pixel set of the outer contour of a cell is as follows:

[0107]

[0108] in, Represents the set of pixels representing the outer contour of a cell. The polygon represents the outline of the cell's bounding box image. To divide the tolerance, we take here. .

[0109] S3. Calculate the nuclear enlargement ratio and describe the ratio of the cell nuclear area to the normal median cell nuclear area;

[0110] The method for calculating the nuclear enlargement ratio is as follows:

[0111]

[0112] in, This represents the increase in the proportion of the nucleus. Represents the area of ​​the cell nucleus in pixels. This represents the normal mid-layer kernel area in pixels.

[0113] S4. Calculate the nucleocytoplasmic ratio, describing the ratio of the cell's nuclear area to its cytoplasmic area;

[0114] The nucleus-to-mass ratio is calculated as follows:

[0115]

[0116] in, Represents the nucleus-to-cylinder ratio, Represents cell area in pixels.

[0117] S5. Calculate nuclear roundness, which describes the smoothness of the cell nuclear boundary;

[0118] The method for calculating the roundness of the kernel is as follows:

[0119]

[0120] in, Represents the roundness of the kernel. This represents the perimeter of the cell nucleus, expressed in pixels.

[0121] S6. Calculate chromatin roughness to describe the uniformity of nuclear chromatin distribution as described in the TBS diagnostic criteria.

[0122] The method for calculating chromatin roughness is as follows:

[0123]

[0124] in, Represents chromatin roughness. Represents a cell nuclear mask image. For the calculation of the gray-level co-occurrence matrix, For matrix standardization, To calculate entropy.

[0125] Matrix standardization is implemented as follows:

[0126]

[0127] in This is the matrix to be standardized.

[0128] The entropy calculation is implemented as follows:

[0129]

[0130] in Let be the entropy matrix to be determined.

[0131] S7. Calculate chromatin granularity and describe the degree of nuclear chromatin granularity as described in the TBS diagnostic criteria.

[0132] The method for calculating chromatin granularity is as follows:

[0133]

[0134] in, Represents chromatin granularity. A polygon representing the outline of the cell nucleus. Representative image Given the side length of the largest patch at the midpoint of the dividing line, the implementation is as follows:

[0135]

[0136] in, The difference in brightness is in the vertical direction. The difference in brightness in the horizontal direction is defined as follows:

[0137]

[0138]

[0139] in, For sampling images, The bounding box sampling function is calculated as follows:

[0140]

[0141] in, The coordinates of the sampling center are... This represents the side length of the sampling box.

[0142] S8. Calculate the perinuclear halo degree and describe the degree of significance of the perinuclear halo as described in the TBS diagnostic criteria;

[0143] The method for calculating the perinuclear halo intensity is as follows:

[0144]

[0145] in, Represents the perinuclear halo degree, The threshold for truncating the brightness of pixels on the outer contour of the cell nucleus. The threshold for truncating the brightness of cytoplasmic pixels. The truncated mean operation is implemented as follows:

[0146]

[0147] in, For the set of operands of the operation, To truncate the upper bound proportion, To truncate the lower bound ratio, This represents a descending order of an ordered set.

[0148] S9. Calculate the voidness to describe the severity of hollowing-out cell formation as described in the TBS diagnostic criteria.

[0149] The method for calculating voidness is as follows:

[0150]

[0151] in, Represents the degree of voidness. The set of side lengths for cell cavities is calculated as follows:

[0152]

[0153] in, Image representing a cell bounding box. The cytoplasmic brightness threshold is calculated as follows:

[0154]

[0155] in, Represents the set of pixels representing the outer contour of a cell. The threshold for truncating the brightness of pixels on the outer contour of the cell.

[0156] S10. Calculate the cytopathic effect score and describe the degree of cytopathic effect based on the obtained cell parameters and the normal value of the item.

[0157] The formula for calculating the cytopathic effect score is as follows:

[0158]

[0159] in, The scoring weights corresponding to their respective pathological parameters are taken here. , The normalized bias calculated for each pathological parameter The calculation method is as follows ,the remaining The calculation method is consistent, with The calculation method for example is as follows:

[0160]

[0161] in, This indicates the baseline value of the pathological parameter under its subscript. and It indicates the lower and upper bounds of the normal range for the pathological parameters under its subscript.

[0162] This invention applies to each cell in the open-source dataset SIPaKMeD, using the Superficial-Intermediate class as the baseline cell, the Parabasal and Metaplastic classes as Class I lesion cells, the Koilocytotic class as Class II lesion cells, and the Dyskeratotic class as Class III lesion cells. With lesion cells as positive examples and the baseline cell as negative examples, the resulting lesion score (PR) curve is shown below. Figure 2 As shown, the distribution maps of scores for various cytopathic effects are as follows: Figure 3 As shown in the figure, we observed that the area under the curve (AUC-PR) reached 0.9948, with the baseline cells concentrated in the low lesion score region. This demonstrates that our method can effectively distinguish baseline cells from lesion cells based on cytopathic score.

[0163] This invention may have other embodiments. Without departing from the spirit and essence of this invention, those skilled in the art can make various corresponding changes and modifications according to this invention, but these corresponding changes and modifications should all fall within the scope of this invention.

Claims

1. A quantitative scoring method for the degree of cervical cellular lesions based on the TBS diagnostic criteria, characterized by: The scoring method includes the following steps: S1. Data Reading: Read cell bounding box image, cell edge pixel coordinate sequence, cell nucleus edge pixel coordinate sequence, and effective nucleus count. and the area of ​​normal middle layer cell nuclei; S2. Calculation of basic cell features: Calculate cell area, cell nucleus area, cell nucleus perimeter, cell nucleus mask image, cytoplasmic pixel set, cell nucleus outer contour pixel set, and cell outer contour pixel set; S3. Calculate the nuclear enlargement ratio and describe the ratio of the cell nuclear area to the normal median cell nuclear area; S4. Calculate the nucleocytoplasmic ratio, describing the ratio of the cell's nuclear area to its cytoplasmic area; S5. Calculate nuclear roundness, which describes the smoothness of the cell nuclear boundary; S6. Calculate chromatin roughness to describe the uniformity of nuclear chromatin distribution as described in the TBS diagnostic criteria. S7. Calculate chromatin granularity and describe the degree of nuclear chromatin granularity as described in the TBS diagnostic criteria. S8. Calculate the perinuclear halo degree and describe the degree of significance of the perinuclear halo as described in the TBS diagnostic criteria; S9. Calculate the voidness to describe the severity of hollowing-out cell formation as described in the TBS diagnostic criteria. S10. Calculate the cytopathic effect score and describe the degree of cytopathic effect based on the obtained cell parameters and normal values.

2. The cervical cellular lesion severity scoring method quantified by the TBS diagnostic criteria as described in claim 1, characterized in that, The method for calculating chromatin roughness in step S6 is as follows: Chromatin roughness describes the uniformity of nuclear chromatin distribution as described in the TBS diagnostic criteria, and the calculation formula is as follows: in, Represents chromatin roughness. Represents a cell nuclear mask image. For grayscale processing, For the calculation of the gray-level co-occurrence matrix, For matrix standardization, To calculate entropy.

3. The cervical cellular lesion severity scoring method quantified by the TBS diagnostic criteria as described in claim 1, characterized in that, The method for calculating chromatin granularity in step S7 is as follows: Chromatin granularity describes the degree of nuclear chromatin granulation as described in the TBS diagnostic criteria, and the calculation formula is as follows: in, Represents chromatin granularity. Represents the area of ​​the cell nucleus in pixels. Calculate the diameter of a polygon. A polygon representing the outline of the cell nucleus. Representative image Given the side length of the largest patch at the midpoint of the dividing line, the implementation is as follows: in, The difference in brightness is in the vertical direction. The difference in brightness in the horizontal direction is defined as follows: in, For sampling images, The bounding box sampling function is calculated as follows: in, The coordinates of the sampling center are... This represents the side length of the sampling box.

4. The cervical cellular lesion severity scoring method quantified by the TBS diagnostic criteria as described in claim 1, characterized in that, The method for calculating the perinuclear halo in step S8 is as follows: The perinuclear halo intensity describes the degree of prominence of the perinuclear halo as described in the TBS diagnostic criteria, and is calculated using the following formula: in, Represents the perinuclear halo degree, Represents the set of pixels representing the outer contour of the cell nucleus. Represents a set of cytoplasmic pixels. The threshold for truncating the brightness of pixels along the outer contour of the cell nucleus. The threshold for truncating the brightness of cytoplasmic pixels. The truncated mean operation is implemented as follows: in, For the set of operands of the operation, To truncate the upper bound proportion, To truncate the lower bound ratio, This represents a descending order of an ordered set.

5. The cervical cellular lesion severity scoring method quantified by the TBS diagnostic criteria as described in claim 1, characterized in that... The method for calculating voidness in step S9 is as follows: Voidness describes the severity of vacuolation as described in the TBS diagnostic criteria, and is calculated as follows: in, Represents the degree of voidness. Represents cell area in pixels. The set of side lengths for cell cavities is calculated as follows: in, Image representing a cell bounding box. The polygon representing the outline of the cell. The cytoplasmic brightness threshold is calculated as follows: in, Represents the set of pixels representing the outer contour of a cell. The threshold for truncating the brightness of pixels on the outer contour of the cell.

Images