A dose distribution map analysis method and system based on image processing

Abstract

The application relates to the technical field of image analysis, and discloses a dose distribution map analysis method and system based on image processing, which comprises the following steps: acquiring a dose distribution map, and performing pixel analysis on the preprocessed dose distribution map to obtain a tissue density modulation coefficient representing the modulation degree of a dose-gray mapping relationship of a single pixel point corresponding to tissue density; the tissue density modulation coefficient generated in step S1 reflects the density modulation intensity of bone tissue, soft tissue, gas cavity and the like; in step S2, a dose-gray mapping matrix is established through a gray density responsivity and a heterogeneity correction value, so that the mapping deviation between gray and dose is eliminated; in step S3, a dynamic gray threshold value of each pixel point is generated; and in step S4, a target region, a normal tissue region and a boundary are accurately determined through comparison of a comparison offset and a neighborhood sign tendency comparison, so that the region attribution of each pixel point truly reflects the radiation dose distribution.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and specifically to a dose distribution map analysis method and system based on image processing. Background Technology

[0002] Currently, dose distribution map analysis mainly divides different dose regions through image segmentation. Specifically, it uses a globally fixed grayscale threshold to binarize the high-dose target area and the low-dose normal tissue area, and uses an edge detection algorithm to extract the contours of each dose gradient region, thereby achieving spatial segmentation of the dose distribution.

[0003] However, the above analysis method still has the following defects: the segmentation method with a fixed grayscale threshold ignores the modulation effect of tissue density differences between individual patients on the dose-grayscale mapping curve. In actual clinical images, the correspondence between the grayscale threshold and the true dose boundary will deviate from the calibration curve in the area where bone tissue and soft tissue meet or near air-containing cavities, resulting in frequent missegmentation and omissions in the segmentation results, and it is impossible to determine whether the pixel is a normal tissue area or a target area. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a dose distribution map analysis method and system based on image processing, which solves the aforementioned problems.

[0005] The above-mentioned technical objective of the present invention is achieved through the following technical solution:

[0006] A dose distribution map analysis method based on image processing includes:

[0007] Step S1: Obtain the dose distribution map and perform pixel analysis on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship of the tissue density corresponding to a single pixel.

[0008] Step S2: Based on the dose distribution map, the gray value of each pixel is correlated and calibrated with the corresponding tissue density modulation coefficient to eliminate the deviation between gray value and dose mapping in different tissue regions, and a dose-gray value mapping matrix representing the relationship between dose and gray value of each pixel is obtained.

[0009] Step S3: Calculate the dose-grayscale mapping matrix, analyze the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix, and obtain the dynamic grayscale threshold of each pixel;

[0010] Step S4: Determine the region to which the pixel belongs based on the dynamic grayscale threshold.

[0011] Furthermore, pixel analysis is performed on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship corresponding to the tissue density of a single pixel, including:

[0012] On the preprocessed dose distribution map, for each pixel, the gradient change in different directions is analyzed to obtain the local gray-level anisotropy index, which represents the difference in texture direction of the dose distribution map caused by the non-uniformity of tissue density.

[0013] Furthermore, pixel analysis is performed on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship corresponding to the tissue density of a single pixel. This also includes:

[0014] Based on the local gray-level anisotropy index, combined with the gray-level value of the pixel, the dispersion of the gray-level value is analyzed at multiple decreasing scales to generate a gray-level variation factor that represents the degree of influence of tissue density on the variation of gray-level distribution at different spatial resolutions.

[0015] By combining and mapping the local gray-level anisotropy index and the gray-level variation factor, a tissue density modulation coefficient is obtained, which represents the degree of modulation of the dose-gray-level mapping relationship of the tissue density corresponding to a single pixel.

[0016] Furthermore, based on the dose distribution map, the grayscale value of each pixel is correlated and calibrated with the corresponding tissue density modulation coefficient to eliminate the deviation in grayscale-dose mapping between different tissue regions, resulting in a dose-grayscale mapping matrix representing the correspondence between dose and grayscale for each pixel, including:

[0017] On the dose distribution map, for each pixel, its gray value and tissue density modulation coefficient are analyzed to determine the intensity of the interaction between the two, and a gray-density response representing the gray value of a single pixel after being modulated by tissue density is generated.

[0018] Based on the gray-density responsivity, the changes in the gray-density responsivity of the neighborhood around the pixel are analyzed, and a heterogeneity correction value representing the reverse compensation of local mapping deviation is generated.

[0019] Furthermore, based on the dose distribution map, the grayscale value of each pixel is correlated and calibrated with the corresponding tissue density modulation coefficient to eliminate the deviation in grayscale-dose mapping between different tissue regions, resulting in a dose-grayscale mapping matrix representing the correspondence between dose and grayscale for each pixel. This also includes:

[0020] For gray-density responsivity and heterogeneity correction values, at each pixel, the gray value is used as the basis, and the gray-density responsivity of that pixel is superimposed to obtain a dose-gray-level mapping matrix that represents the relationship between dose and gray value for each pixel.

[0021] Furthermore, the dose-grayscale mapping matrix is ​​calculated, and the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix are analyzed to obtain the dynamic grayscale threshold of each pixel, including:

[0022] For each pixel in the dose-grayscale mapping matrix, the slope of the mapping value as grayscale changes within the neighborhood of that pixel is analyzed to generate a local gradient sensitivity that represents the sensitivity of the dose-grayscale mapping relationship at each pixel to grayscale changes.

[0023] Based on local gradient sensitivity, the consistency of the bending direction of the mapping curve between a pixel and its surrounding pixels is analyzed, and a mapping curvature anomaly index representing the distortion of the mapping trend of each pixel in the mapping matrix is ​​generated.

[0024] Furthermore, the dose-grayscale mapping matrix is ​​calculated, and the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix are analyzed to obtain the dynamic grayscale threshold of each pixel. This also includes:

[0025] The coordinated variation of local gradient sensitivity and mapping curvature anomaly index in the dose-grayscale mapping matrix is ​​analyzed, the sorting consistency of each pixel in the mapping is analyzed, and dose transition feature values ​​representing the mapping inflection critical points of each pixel are generated.

[0026] Furthermore, the dose-grayscale mapping matrix is ​​calculated, and the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix are analyzed to obtain the dynamic grayscale threshold of each pixel. This also includes:

[0027] Based on the dose transition feature value, and combined with the distribution of dose transition feature values ​​of other pixels in the neighborhood around the pixel, a mapping consistency analysis is performed to generate a neighborhood mapping coordination degree that represents the correction of the dose transition feature value deviation.

[0028] The dose transition feature value is adjusted based on the neighborhood mapping coordination degree to generate a dynamic grayscale threshold for each pixel.

[0029] Furthermore, based on the dynamic grayscale threshold, the region to which a pixel belongs is determined, including:

[0030] On the preprocessed dose distribution map, for each pixel, the gray value of the pixel is compared with the dynamic gray value threshold, and the gray value is modulated according to the dispersion of the gray value in the local neighborhood of the pixel to generate a comparison offset.

[0031] Based on the comparison offset, and combined with the spatial distribution of the comparison offsets of other pixels in the surrounding neighborhood, the region to which the pixel belongs is determined.

[0032] Furthermore, an image processing-based dose distribution map analysis system, applied to the above analysis method, includes:

[0033] The image analysis unit is used to acquire the dose distribution map and perform pixel analysis on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship of the tissue density corresponding to a single pixel.

[0034] The influence analysis unit is used to correlate and calibrate the gray value of each pixel with the corresponding tissue density modulation coefficient according to the dose distribution map, eliminate the deviation of gray value and dose mapping in different tissue regions, and obtain the dose-gray value mapping matrix representing the relationship between dose and gray value of each pixel.

[0035] The threshold calculation unit is used to calculate the dose-grayscale mapping matrix, analyze the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix, and obtain the dynamic grayscale threshold of each pixel.

[0036] The region determination unit is used to determine the region to which a pixel belongs based on a dynamic grayscale threshold.

[0037] In summary, the present invention has the following main beneficial effects:

[0038] In step S1, the tissue density modulation coefficient is generated by fusing the local gray-level anisotropy index and the gray-level variation factor. This accurately reflects the density modulation intensity of areas such as bone tissue, soft tissue junctions, and air-containing cavities, avoiding the omission of oblique texture and gray-level discrete information in single-direction or single-scale analysis. In step S2, a dose-gray-level mapping matrix is ​​established by using gray-level-density responsivity and heterogeneity correction values ​​to effectively eliminate mapping deviations between gray-level and dose. In step S3, a dynamic gray-level threshold for each pixel is generated based on the collaborative analysis of local gradient sensitivity, mapping curvature anomaly index, dose transition characteristic value, and neighborhood mapping coordination. This can adaptively capture the turning point caused by abnormal bending direction of the mapping curve and inconsistent sensitivity sorting. Finally, in step S4, the target area, normal tissue area, and boundary are accurately determined by comparing the contrast offset and neighborhood sign tendency. This scheme reduces the missegmentation rate of density abrupt change areas and ensures that the region assignment of each pixel truly reflects the radiation dose distribution. Attached Figure Description

[0039] Figure 1 This is a flowchart illustrating the steps of a dose distribution map analysis method based on image processing according to the present invention;

[0040] Figure 2 This is a schematic diagram of a dose distribution map analysis system based on image processing according to the present invention. Detailed Implementation

[0041] 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, and 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.

[0042] refer to Figure 1 and Figure 2 A dose distribution map analysis method based on image processing, comprising:

[0043] Step S1: Obtain the dose distribution map and perform pixel analysis on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship corresponding to the tissue density of a single pixel. The dose distribution map is the dedicated visualization image atlas in radiotherapy planning.

[0044] Preprocessing of dose distribution maps includes at least image denoising, grayscale normalization, and cropping of invalid edge regions;

[0045] Step S2: Based on the dose distribution map, the gray value of each pixel is correlated and calibrated with the corresponding tissue density modulation coefficient to eliminate the deviation between gray value and dose mapping in different tissue regions, and a dose-gray value mapping matrix representing the relationship between dose and gray value of each pixel is obtained.

[0046] Step S3: Calculate the dose-grayscale mapping matrix, analyze the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix, and obtain the dynamic grayscale threshold of each pixel;

[0047] Step S4: Determine the region to which the pixel belongs based on the dynamic grayscale threshold.

[0048] In one embodiment, pixel analysis is performed on the preprocessed dose distribution map to obtain a tissue density modulation coefficient representing the degree of modulation of the dose-grayscale mapping relationship corresponding to a single pixel, including:

[0049] On the preprocessed dose distribution map, for each pixel, the gradient change in different directions is analyzed to obtain the local gray-level anisotropy index, which represents the difference in texture direction of the dose distribution map caused by the non-uniformity of tissue density. Specifically, on the preprocessed dose distribution map, for each pixel, a 3×3 neighborhood window is constructed with the pixel as the center, and the gray-level difference values ​​in the four directions of 0 degrees, 45 degrees, 90 degrees, and 135 degrees are calculated respectively. That is, along the 0-degree direction, the absolute value of the gray-level difference between the left and right adjacent pixels of the pixel is calculated; along the 45-degree direction, the absolute value of the gray-level difference between the upper left and lower right pixels of the pixel is calculated; along the 90-degree direction, the absolute value of the gray-level difference between the pixels above and below the pixel is calculated; and along the 135-degree direction, the absolute value of the gray-level difference between the upper right and lower left pixels of the pixel is calculated.

[0050] Divide the absolute value in each direction by the Euclidean distance between the two pixels in that direction. The distance in the 0-degree and 90-degree directions is 2 pixel units, and the distance in the 45-degree and 135-degree directions is 2√2 pixel units. This gives the gradient magnitude in each of the four directions. Then, calculate the ratio of the standard deviation to the mean of the gradient magnitude in each of the four directions to obtain the local gray-level anisotropy index of the pixel. The larger the local gray-level anisotropy index, the more significant the anisotropy caused by the non-uniformity of tissue density. It can independently reflect the degree of modulation strength of a single pixel position by the difference in the spatial distribution of tissue density.

[0051] In the dose distribution map, uneven tissue density leads to local anisotropy in dose deposition, manifested as different rates of change in grayscale texture in different directions. The four directions of 0 degrees, 45 degrees, 90 degrees and 135 degrees are selected because 0 degrees and 90 degrees, i.e., the horizontal and vertical directions, can capture grayscale changes aligned with the beam incident direction and the conventional dose grid, while 45 degrees and 135 degrees, i.e., the diagonal directions, can reflect the non-orthogonal texture differences caused by oblique tissue boundaries or density gradients. The four directions together constitute a complete directional base in the plane, avoiding the omission of oblique modulation information when only two directions are used, and are more suitable for complex grayscale modulation caused by irregular tissue density fields in the dose distribution map.

[0052] In one embodiment, pixel analysis is performed on the preprocessed dose distribution map to obtain a tissue density modulation coefficient representing the degree of modulation of the dose-grayscale mapping relationship corresponding to a single pixel, and the method further includes:

[0053] Based on the local gray-level anisotropy index, combined with the gray-level value of the pixel, the dispersion of the gray-level value is analyzed at multiple decreasing scales to generate a gray-level variation factor that represents the degree of influence of tissue density on gray-level distribution variation under different spatial resolutions. Specifically, this includes: constructing neighborhood windows with side lengths of 5 pixels and 3 pixels respectively, centered on the current pixel. The 5-pixel window is a large-scale window, mainly used to cover a wider spatial range, thereby capturing the overall pattern of slow changes in density gradient and large-scale gray-level dispersion in continuous tissue regions. The 3-pixel window is a small-scale window, used to reflect short-term gray-level fluctuations in a local fine range, adapting to local gray-level abrupt changes caused by small tissue density differences.

[0054] Within each neighborhood window, the ratio of the standard deviation to the mean of all pixel grayscale values ​​is calculated to obtain the coefficient of variation at that scale. The coefficient of variation at each scale is multiplied by the local grayscale anisotropy index of that pixel to obtain two weighted values. The mean of these two weighted values ​​is calculated, and the result is normalized to the 0-1 interval to generate a grayscale variation factor that represents the degree of influence of tissue density on grayscale distribution variation at different spatial resolutions. The larger the value of the grayscale variation factor, the more significant the grayscale dispersion caused by tissue density inhomogeneity at multiple scales. Unlike the single-pixel neighborhood orientation feature detection of the local grayscale anisotropy index, the grayscale variation factor focuses on representing the grayscale dispersion characteristics caused by large-scale tissue density inhomogeneity. The two feature dimensions are complementary.

[0055] By combining and mapping the local gray-level anisotropy index and the gray-level variation factor, a tissue density modulation coefficient is obtained, which represents the degree of modulation of the dose-gray-level mapping relationship by the tissue density corresponding to a single pixel. Specifically, the square root of the product of the local gray-level anisotropy index and the gray-level variation factor is obtained. When the values ​​of the local gray-level anisotropy index and the gray-level variation factor are simultaneously large, it indicates that the gray-level dispersion of the pixel position is high and the directional anisotropy is significant, and the modulation effect of tissue density on the dose-gray-level mapping is the strongest. If the value of either parameter is zero, the tissue density modulation coefficient is zero, indicating that the pixel is not affected by tissue density modulation.

[0056] By employing gradient amplitude statistics in four directions (0°, 45°, 90°, and 135°), this method accurately captures anisotropic textures caused by density abrupt changes in areas such as the junction of bone and soft tissue and air-filled cavities. This avoids missing oblique modulation information in a single direction. Furthermore, the grayscale variation factor analyzes the degree of grayscale dispersion at 5-pixel and 3-pixel scales, taking into account both large-scale density gradients and local short-term fluctuations. The resulting tissue density modulation coefficient reflects the modulation strength of each pixel on the dose-grayscale mapping relationship of tissue density, thereby adaptively correcting the segmentation boundary. Compared with fixed threshold methods, this approach reduces the probability of missegmentation and missed segmentation in complex density regions, ensuring accurate division between the target area and normal tissue areas.

[0057] In one embodiment, based on the dose distribution map, the grayscale value of each pixel is correlated and calibrated with the corresponding tissue density modulation coefficient to eliminate the deviation in grayscale-dose mapping between different tissue regions, resulting in a dose-grayscale mapping matrix representing the dose-grayscale correspondence of each pixel, including:

[0058] On the dose distribution map, for each pixel, its gray value and tissue density modulation coefficient are analyzed to determine the intensity of their mutual influence, generating a gray-density responsivity representing the gray value of a single pixel after being modulated by tissue density. The specific calculation formula is as follows: ;

[0059] in, This represents the gray-density response of a single pixel after its gray value has been modulated by tissue density. This represents the grayscale value of the center pixel. Represents the tissue density modulation coefficient. Represents an exponential function. This represents the standard deviation of the grayscale values ​​of 9 pixels within a 3×3 neighborhood. This represents the global standard deviation of the grayscale values ​​of all pixels in the dose distribution map. This represents the arithmetic mean of the gray values ​​of 8 pixels within a 3×3 neighborhood (excluding the center pixel). Represents the hyperbolic tangent function. This represents the center pixel offset rate, i.e. , This represents the normalized value of the local gray-level anisotropy index. Indicates the grayscale variation factor;

[0060] Based on the gray-density responsivity, the changes in the gray-density responsivity of the neighborhood around the pixel are analyzed, and a heterogeneity correction value representing the reverse compensation of local mapping deviation is generated. Specifically, this includes: taking each pixel as the center, selecting a square neighborhood with a side length of 3 pixels, and calculating the mean and standard deviation of the gray-density responsivity of all pixels in the neighborhood.

[0061] Subtract the neighborhood mean from the gray-density responsivity of the center pixel, and then divide the difference by the sum of the neighborhood mean and 1 to obtain the center deviation magnitude.

[0062] Within the same neighborhood, the gray-density response of each pixel is regarded as a gray value. The four-directional gradient calculation method, which is exactly the same as the local gray-level anisotropy index, is used to calculate the gradient magnitude in the four directions: 0 degrees, 45 degrees, 90 degrees, and 135 degrees. Then, the standard deviation of these four magnitudes is calculated.

[0063] Finally, the neighborhood standard deviation is multiplied by the center deviation magnitude, and then multiplied by the ratio of the standard deviation of the four gradient magnitudes to the mean gray-density responsivity in that neighborhood. This is the heterogeneity correction value representing the reverse compensation of local mapping bias. The larger the value of the heterogeneity correction value, the more severe the spatial variation of the responsivity around that point, and the stronger the reverse compensation weight is required.

[0064] In one embodiment, based on the dose distribution map, the grayscale value of each pixel is correlated and calibrated with the corresponding tissue density modulation coefficient to eliminate the deviation in grayscale-dose mapping between different tissue regions, resulting in a dose-grayscale mapping matrix representing the dose-grayscale correspondence of each pixel. The method further includes:

[0065] For gray-density responsivity and heterogeneity correction values, at each pixel, based on the gray value, the gray-density responsivity of that pixel is superimposed to obtain a dose-gray-level mapping matrix representing the relationship between dose and gray value for each pixel. Specifically, at each pixel, based on its gray value, the gray-density responsivity of that pixel is directly added to the gray value to obtain an intermediate mapping value. Then, the heterogeneity correction value is multiplied by the tissue density modulation coefficient of that pixel, and this product result is added to the intermediate mapping value to obtain the final mapping value of that pixel.

[0066] The mapping value is arranged according to the row and column position of the pixel in the dose distribution map, that is, the row number corresponds to the vertical coordinate of the image and the column number corresponds to the horizontal coordinate. The mapping values ​​of all pixels together form a two-dimensional matrix, which is the dose-grayscale mapping matrix. The position of each pixel in the matrix corresponds one-to-one with the original pixel position.

[0067] Different tissue densities in the human body directly cause differences in local dose deposition. The pixel gray values ​​of the gray dose distribution map indirectly reflect the distribution level of radiation dose. Therefore, after gray calibration is completed through tissue density modulation and heterogeneity correction, the mapping value can indirectly represent the dose distribution change pattern at the corresponding location, realizing the correlation mapping between gray features and dose distribution.

[0068] By correlating and calibrating the grayscale value of each pixel with the corresponding tissue density modulation coefficient, the deviation between grayscale and dose mapping in different tissue regions is eliminated. Furthermore, through multi-parameter collaborative calculation, a dose-grayscale mapping matrix is ​​obtained. This matrix can adaptively adjust the mapping relationship in areas of density abrupt change, such as the junction of bone and soft tissue and air-containing cavities, so that grayscale features are accurately correlated with the true dose distribution, ensuring the accuracy of target area boundary positioning.

[0069] In one embodiment, the dose-grayscale mapping matrix is ​​calculated, and the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix are analyzed to obtain the dynamic grayscale threshold of each pixel, including:

[0070] For each pixel in the dose-grayscale mapping matrix, within the neighborhood of that pixel, the slope of the mapping value as grayscale changes is analyzed to generate a local gradient sensitivity that represents the sensitivity of the dose-grayscale mapping relationship at each pixel to grayscale changes. Specifically, for each pixel in the dose-grayscale mapping matrix, for the grayscale value of each pixel in the square neighborhood of that pixel and the corresponding mapping value of that pixel in the dose-grayscale mapping matrix, the grayscale value is used as the independent variable and the mapping value is used as the dependent variable. All different pixel pairs in the neighborhood are traversed. For each pair of pixels, the ratio of the difference in mapping value to the difference in grayscale value is calculated to obtain the slope of that pair of pixels. If the difference in grayscale value is zero, the pixel pair is skipped.

[0071] Then, the median of all slopes is calculated, which is the local gradient sensitivity, representing how sensitive the dose-grayscale mapping relationship at each pixel is to grayscale changes. The larger the local gradient sensitivity value, the more sensitive the mapping value is to grayscale changes, that is, the steeper the local dose-grayscale mapping relationship.

[0072] Based on local gradient sensitivity, the consistency of the bending direction of the mapping curve between a pixel and its surrounding pixels is analyzed, and a mapping curvature anomaly index representing the distortion of the mapping trend of each pixel in the mapping matrix is ​​generated. Specifically, for the square neighborhood of each pixel in the dose-grayscale mapping matrix, the local curvature sign of the center pixel is calculated within the neighborhood: find two pixels in the neighborhood whose grayscale values ​​are higher and lower than the center pixel respectively. If there are multiple pixels with higher and lower grayscale values, select the pixel with the smallest grayscale difference from the center pixel as the reference pixel. If there are no pixels in the neighborhood that simultaneously meet the conditions of being higher and lower, the curvature sign of the center pixel is directly determined to be zero.

[0073] Calculate the slope from the high grayscale reference pixel to the center pixel, and then calculate the slope from the center pixel to the low grayscale reference pixel. These two slopes are the corresponding mapping value difference divided by the grayscale difference. Then divide the difference between the high grayscale value and the low grayscale value by 2 to obtain the grayscale calculation value. Divide the difference between the two slopes by the grayscale calculation value. The sign of the result is the local curvature sign of the center pixel, with positive indicating convex and negative indicating concave.

[0074] According to the above calculation method, the local curvature sign of each pixel in the neighborhood is calculated. Then, the number of local curvature signs of the surrounding 8 pixels in the square neighborhood that are the same as the local curvature sign of the center pixel is counted. After subtracting the number from 1, the result is normalized to the 0-1 interval, which is the mapping curvature anomaly index that represents the mapping trend distortion of each pixel in the mapping matrix. The closer the mapping curvature anomaly index is to 1, the more inconsistent the bending direction of the center pixel is with the surrounding pixels, that is, the more significant the mapping trend distortion is.

[0075] In one embodiment, the dosage-grayscale mapping matrix is ​​calculated, and the mapping trend and characteristics of each pixel in the dosage-grayscale mapping matrix are analyzed to obtain the dynamic grayscale threshold of each pixel. The method further includes:

[0076] The coordinated variation law of local gradient sensitivity and mapping curvature anomaly index in the dose-grayscale mapping matrix is ​​analyzed. The sorting consistency of each pixel in the mapping is analyzed, and dose transition feature value representing the mapping inflection critical point of each pixel is generated. Specifically, for each pixel's square neighborhood, the local gradient sensitivity of each pixel in the neighborhood is sorted in ascending order of value, and the sequence number of each pixel in the sort is recorded, which is the rank, with a value range of 1 to 9, forming an ordered column of length 9.

[0077] Starting from the top-left pixel of the square neighborhood, traverse the first row from left to right, then the second row from left to right, and finally the third row from left to right. This results in a natural position sequence consisting of 9 positions, which is the position sequence from 1 to 9.

[0078] Calculate the Kendall rank correlation coefficient between the ordered sequence and the natural position sequence. The Kendall rank correlation coefficient mainly reflects the consistency between the magnitude of the sensitivity value and its spatial position.

[0079] Determine the rank of the center pixel, calculate the absolute value of the difference between the rank of the center pixel and the median of all ranks in the neighborhood to obtain the center rank deviation. Multiply the absolute value of the Kendall rank correlation coefficient by the center rank deviation to obtain the preliminary sorting anomaly value. Then, add 1 to the mapping curvature anomaly index and multiply it by the preliminary sorting anomaly value. Then multiply it by 1 and the absolute value of the difference between the Kendall rank correlation coefficient to obtain the dose transition feature value representing the mapping inflection point of each pixel. The larger the value of the dose transition feature value, the more serious the deviation between the sensitivity ranking of the center pixel and the neighborhood spatial sorting trend is, and the center sensitivity is in an extreme position, that is, the direction of the mapping curve curvature is abnormal.

[0080] In one embodiment, the dosage-grayscale mapping matrix is ​​calculated, and the mapping trend and characteristics of each pixel in the dosage-grayscale mapping matrix are analyzed to obtain the dynamic grayscale threshold of each pixel. The method further includes:

[0081] Based on the dose transition feature value, a mapping consistency analysis is performed on the distribution of dose transition feature values ​​of other pixels in the neighborhood of the pixel. This generates a neighborhood mapping coordination degree that represents the correction of the dose transition feature value deviation. Specifically, for each pixel's square neighborhood, the dose transition feature value of the center pixel is denoted as C. The dose transition feature values ​​of the remaining 8 pixels are arranged clockwise starting from the pixel directly above. The difference between the dose transition feature value of each pixel and C is calculated. If the difference is positive, its sign is 1; if the difference is negative, its sign is -1; if the difference is zero, its sign is 0, resulting in a sign sequence of length 8.

[0082] Calculate the number of times adjacent symbols in the symbol sequence change, divide the number of changes by 8, and normalize the result to the 0-1 interval to obtain the symbol inconsistency rate;

[0083] Simultaneously, the difference between the maximum and minimum values ​​of all nine dose transition characteristic values ​​in the neighborhood is calculated to obtain the range. The sign inconsistency rate is multiplied by the range, and then multiplied by 1 by the difference between the sign inconsistency rate and the range. This yields the neighborhood mapping consistency, which represents the correction of the dose transition characteristic value deviation. The smaller the value of the neighborhood mapping consistency, the more consistent the spatial variation of the dose transition characteristic value in the local area, i.e., the higher the mapping transition consistency.

[0084] The dose transition feature value is adjusted according to the neighborhood mapping coordination degree to generate the dynamic grayscale threshold of each pixel. Specifically, the mean of the grayscale values ​​of all pixels in the square neighborhood of the pixel is calculated, and the dose transition feature value is multiplied by 1 by the difference between the dose transition feature value and the neighborhood mapping coordination degree, and then multiplied by the mean value to obtain the dynamic grayscale threshold of the pixel.

[0085] This technical solution generates an individualized dynamic grayscale threshold by analyzing the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix. Local gradient sensitivity reflects the sensitivity of the mapping value to grayscale changes, mapping curvature anomaly index identifies distortions where the bending direction of the mapping curve is inconsistent with the neighborhood, dose transition feature value captures the degree of deviation between sensitivity ranking and spatial position, and neighborhood mapping coordination corrects the deviation of local transitions. The final dynamic grayscale threshold can adaptively adjust in areas of density abrupt change, such as the junction of bone and soft tissue and air-containing cavities, so that the segmentation boundary accurately fits the real dose distribution. This solution reduces the missegmentation and missed segmentation rates caused by density modulation, ensuring that each pixel can be accurately identified as a target area or normal tissue area.

[0086] In one embodiment, determining the region (target area and normal tissue area) to which a pixel belongs based on a dynamic grayscale threshold includes:

[0087] On the preprocessed dose distribution map, for each pixel, the gray value of the pixel is compared with the dynamic gray value threshold, and the gray value in the local neighborhood of the pixel is modulated to generate a comparison offset. Specifically, on the preprocessed dose distribution map, the difference between the gray value of the center pixel and its dynamic gray value threshold is calculated to obtain the base difference.

[0088] Within the same neighborhood, the gray value of each pixel is compared with its own dynamic gray value threshold. The number of pixels above the threshold and the number of pixels below the threshold are counted. The smaller value of the number of pixels is divided by the larger value to obtain the neighborhood comparison consistency coefficient.

[0089] The comprehensive modulation factor is obtained by multiplying the three parameters of the pixel's local gray-level anisotropy index, gray-level variation factor, and neighborhood mapping compatibility.

[0090] Multiplying the base difference by the neighborhood comparison consistency coefficient, and then multiplying it by the difference between 1 and the comprehensive modulation factor, yields the contrast offset of each pixel.

[0091] Based on the comparative offset, and combined with the spatial distribution of the comparative offsets of other pixels in the surrounding neighborhood, the region to which the pixel belongs is determined. Specifically, this includes: taking the square neighborhood of each pixel as the analysis range, for the comparative offset of the center pixel, determining the sign based on the magnitude of the comparative offset: a sign of positive if the offset is greater than zero, a sign of negative if the offset is less than zero, and a sign of zero if the offset is equal to zero. Then, counting the signs of the comparative offsets of all 9 pixels in the neighborhood, calculating the number of positive and negative signs respectively. If the number of positive signs > the number of negative signs, the neighborhood sign tendency is positive; if the number of negative signs > the number of positive signs, the neighborhood sign tendency is negative; otherwise, the neighborhood sign tendency is 0.

[0092] The sign of the center pixel is compared with the sign tendency of its neighbors. If the two are the same and the sign of the center pixel is non-zero, the sign that the two are the same is directly used as the judgment sign.

[0093] If the sign of the center pixel is zero, count the number of positive and negative signs in the eight pixels in the neighborhood excluding the center, and use the sign with the larger number as the judgment sign. If the number of positive and negative signs is equal, mark the center pixel as the boundary region.

[0094] If the sign of the center pixel is inconsistent with the neighborhood tendency, calculate the absolute value of the center pixel's contrast offset and the average of the absolute values ​​of all non-zero contrast offsets in the square neighborhood. If the absolute value of the center pixel's contrast offset is greater than the average, then the sign of the center pixel is used as the determination sign; otherwise, the neighborhood tendency is used as the determination sign.

[0095] When the determination symbol is positive, the pixel belongs to the target area; when the determination symbol is negative, the pixel belongs to the normal tissue area; otherwise, the pixel is marked as a boundary area.

[0096] By analyzing the calculated contrast offset and dynamic grayscale threshold, positive signs are identified as target areas, negative signs as normal tissue areas, and inconsistent or zero signs are marked as boundary areas. This improves the accuracy of the segmentation results and ensures that the region attribution of each pixel truly reflects the radiation dose distribution.

[0097] In one embodiment, an image processing-based dose distribution map analysis system is applied to the above-described analysis method, comprising:

[0098] The image analysis unit is used to acquire the dose distribution map and perform pixel analysis on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship of the tissue density corresponding to a single pixel.

[0099] The influence analysis unit is used to correlate and calibrate the gray value of each pixel with the corresponding tissue density modulation coefficient according to the dose distribution map, eliminate the deviation of gray value and dose mapping in different tissue regions, and obtain the dose-gray value mapping matrix representing the relationship between dose and gray value of each pixel.

[0100] The threshold calculation unit is used to calculate the dose-grayscale mapping matrix, analyze the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix, and obtain the dynamic grayscale threshold of each pixel.

[0101] The region determination unit is used to determine the region to which a pixel belongs based on a dynamic grayscale threshold.

[0102] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A dose distribution map analysis method based on image processing, characterized in that, include: Step S1: Obtain the dose distribution map and perform pixel analysis on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship of the tissue density corresponding to a single pixel. Step S2: Based on the dose distribution map, the gray value of each pixel is correlated and calibrated with the corresponding tissue density modulation coefficient to eliminate the deviation between gray value and dose mapping in different tissue regions, and a dose-gray value mapping matrix representing the relationship between dose and gray value of each pixel is obtained. Step S3: Calculate the dose-grayscale mapping matrix, analyze the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix, and obtain the dynamic grayscale threshold of each pixel. This includes: for each pixel in the dose-grayscale mapping matrix, analyze the slope of the mapping value as grayscale changes within the neighborhood of each pixel, and generate a local gradient sensitivity that represents the sensitivity of the dose-grayscale mapping relationship at each pixel to grayscale changes. Based on local gradient sensitivity, the consistency of the bending direction of the mapping curve between a pixel and its surrounding pixels is analyzed, and a mapping curvature anomaly index representing the distortion of the mapping trend of each pixel in the mapping matrix is ​​generated. The coordinated variation law of local gradient sensitivity and mapping curvature anomaly index is analyzed in the dose-grayscale mapping matrix. The sorting consistency of each pixel in the mapping is analyzed, and dose transition feature value representing the mapping inflection critical point of each pixel is generated. Based on the dose transition feature value, and combined with the distribution of dose transition feature values ​​of other pixels in the neighborhood of each pixel, a mapping consistency analysis is performed to generate a neighborhood mapping coordination degree that represents the correction of dose transition feature value deviation. The dose transition feature value is adjusted based on the neighborhood mapping coordination degree to generate a dynamic grayscale threshold for each pixel. Step S4: Determine the region to which the pixel belongs based on the dynamic grayscale threshold.

2. The dose distribution map analysis method based on image processing according to claim 1, characterized in that, Pixel analysis was performed on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship corresponding to the tissue density of a single pixel. This coefficient includes: On the preprocessed dose distribution map, for each pixel, the gradient change in different directions is analyzed to obtain the local gray-level anisotropy index, which represents the difference in texture direction of the dose distribution map caused by the non-uniformity of tissue density.

3. The dose distribution map analysis method based on image processing according to claim 2, characterized in that, Pixel analysis is performed on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship corresponding to the tissue density of a single pixel. This also includes: Based on the local gray-level anisotropy index, combined with the gray-level value of each pixel, the dispersion of gray-level values ​​is analyzed at multiple decreasing scales to generate a gray-level variation factor that represents the degree of influence of tissue density on gray-level distribution variation at different spatial resolutions. By combining and mapping the local gray-level anisotropy index and the gray-level variation factor, a tissue density modulation coefficient is obtained, which represents the degree of modulation of the dose-gray-level mapping relationship of the tissue density corresponding to a single pixel.

4. The dose distribution map analysis method based on image processing according to claim 3, characterized in that, Based on the dose distribution map, the grayscale value of each pixel is correlated and calibrated with the corresponding tissue density modulation coefficient to eliminate the deviation in grayscale-dose mapping between different tissue regions, resulting in a dose-grayscale mapping matrix representing the correspondence between dose and grayscale for each pixel, including: On the dose distribution map, for each pixel, its gray value and tissue density modulation coefficient are analyzed to determine the intensity of the interaction between the two, and a gray-density response representing the gray value of a single pixel after being modulated by tissue density is generated. Based on the gray-density responsivity, the changes in the gray-density responsivity of the neighborhood around each pixel are analyzed to generate a heterogeneity correction value that represents the reverse compensation of local mapping deviation.

5. The dose distribution map analysis method based on image processing according to claim 4, characterized in that, Based on the dose distribution map, the grayscale value of each pixel is correlated and calibrated with the corresponding tissue density modulation coefficient to eliminate the deviation between grayscale and dose mapping in different tissue regions, resulting in a dose-grayscale mapping matrix representing the correspondence between dose and grayscale for each pixel. This matrix also includes: For gray-density responsivity and heterogeneity correction values, at each pixel, the gray-density responsivity of each pixel is superimposed based on the gray value to obtain a dose-gray-level mapping matrix that represents the relationship between dose and gray value at each pixel.

6. The dose distribution map analysis method based on image processing according to claim 5, characterized in that, Based on the dynamic grayscale threshold, the region to which a pixel belongs is determined, including: On the preprocessed dose distribution map, for each pixel, the gray value of each pixel is compared with the dynamic gray value threshold, and the gray value is modulated according to the dispersion of the gray value in the local neighborhood of each pixel to generate a comparison offset. Based on the comparison offset, and combined with the spatial distribution of the comparison offsets of other pixels in the neighborhood of each pixel, the region to which each pixel belongs is determined.

7. A dose distribution map analysis system based on image processing, applied in the analysis method as described in any one of claims 1-6, characterized in that, include: The image analysis unit is used to acquire the dose distribution map and perform pixel analysis on the preprocessed dose distribution map to obtain the tissue density modulation coefficient, which represents the degree of modulation of the dose-grayscale mapping relationship of the tissue density corresponding to a single pixel. The influence analysis unit is used to correlate and calibrate the gray value of each pixel with the corresponding tissue density modulation coefficient according to the dose distribution map, eliminate the deviation of gray value and dose mapping in different tissue regions, and obtain the dose-gray value mapping matrix representing the relationship between dose and gray value of each pixel. The threshold calculation unit is used to calculate the dose-grayscale mapping matrix, analyze the mapping trend and characteristics of each pixel in the dose-grayscale mapping matrix, and obtain the dynamic grayscale threshold of each pixel. The region determination unit is used to determine the region to which a pixel belongs based on a dynamic grayscale threshold.

Images