Image-based tumor response prediction method and system

Abstract

This invention relates to the field of medical image analysis technology, specifically to a method and system for predicting tumor treatment efficacy based on images. The method includes the following steps: extracting inter-frame pixel coordinates from CT image data at multiple time points before and after treatment; calculating the inter-frame pixel offset vectors; performing inter-frame position calibration; and generating time-series image correction data. In this invention, precise inter-frame position calibration is achieved through the extraction of inter-frame pixel coordinates and the calculation of offset vectors from multi-time-point CT image data. By setting an extended threshold and fitting regional boundaries, the outer contour of the tumor region is identified. The set of pixel coordinates at the tumor region boundary is dynamically analyzed, and its positional changes over time are evaluated. This provides crucial data for predicting tumor treatment response. The analysis of density distribution characteristics supplements the assessment of tumor region structural homogeneity, providing a basis for personalized treatment prediction and effectively predicting the changing trends of tumor treatment.

Description

Technical Field

[0001] This invention relates to the field of medical image analysis technology, and in particular to an image-based method and system for predicting the efficacy of tumor treatment. Background Technology

[0002] The field of medical imaging analytics encompasses the acquisition, analysis, and interpretation of medical imaging data to support clinical decision-making and medical research. Within this field, professionals utilize various imaging techniques, such as X-rays, MRI, CT, and ultrasound, to acquire imaging data in pathological states. Through the analysis of this imaging data, medical imaging technology can help physicians more accurately diagnose diseases, assess treatment effectiveness, and monitor disease progression.

[0003] One approach to predicting cancer treatment efficacy utilizes processed and analyzed medical imaging data to forecast the responsiveness to cancer treatment and the patient's condition. Technical aspects of this topic include the use of image processing techniques and analytical methods to assess tumor features in images. These methods typically involve steps such as image segmentation, image enhancement, feature recognition, and image quality assessment. Through these technical processes, quantitative information that decisively influences the prediction of treatment outcomes can be extracted from medical images.

[0004] Traditional prediction methods have significant limitations in tracking the dynamic changes of tumors during treatment. Particularly in terms of the accuracy and real-time performance of tumor treatment efficacy prediction and monitoring, traditional methods lack sufficient automation and accuracy in handling inter-frame pixel position offsets and tumor region boundary identification. They often require manual intervention for correction and boundary determination, increasing workload, introducing subjective errors, and limiting the optimization of diagnosis and treatment plans. Regarding continuous monitoring of tumor morphological changes, traditional methods fail to provide sufficient quantitative tools to accurately measure the rate of tumor boundary expansion and structural complexity. This limits real-time and dynamic assessment of treatment response, potentially delaying the optimal timing for adjusting treatment plans. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and to propose an image-based method and system for predicting the efficacy of tumor treatment.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: an image-based method for predicting tumor treatment efficacy, comprising the following steps:

[0007] S1: Based on CT image data at multiple time points before and after treatment, extract the inter-frame pixel position coordinates of the CT image data, calculate the offset vector of the inter-frame pixel position, perform inter-frame position calibration, and generate time-series image correction data.

[0008] S2: Based on the time-series image correction data, select the first frame image in the image sequence, mark the tumor seed points in the first frame image, and extract the outer contour of the tumor region point by point by calculating the intensity difference and gradient change between the seed points and neighboring pixels to generate tumor region boundary data.

[0009] S3: Based on the tumor region boundary data, calculate the position change of each pixel in the time series, compare the difference of boundary coordinates in adjacent frames, combine the frame sequence change, accumulate and analyze the overall change trend, and generate dynamic change data of tumor boundary.

[0010] S4: Based on the dynamic change data of the tumor boundary, calculate the gradient direction and gradient magnitude change of each pixel, statistically analyze the gradient distribution variance in the target area, and judge the uniformity of the internal structure of the region by the consistency of the gradient change direction of the pixels, analyze the density difference of the target area, and obtain density distribution characteristic data.

[0011] S5: Based on the dynamic change data of the tumor boundary and the density distribution characteristic data, the changes of the tumor region in the time series are analyzed. The consistency between the tumor volume expansion rate, the change amplitude of boundary complexity and the change direction of density gradient is calculated frame by frame. Combined with the change trend in the time dimension, quantitative indicators of regional morphological and structural changes are extracted to predict the changes in tumor treatment effect in the future period and obtain dynamic characteristic indicators for efficacy prediction.

[0012] As a further aspect of the present invention, the steps of extracting inter-frame pixel position coordinates from CT image data at multiple time points before and after treatment, calculating the inter-frame pixel position offset vector, performing inter-frame position calibration, and generating time-series image correction data are as follows:

[0013] S101: Based on CT image data from multiple time points before and after treatment, the pixel position coordinates in each frame are extracted. By calculating the spatial position difference of each pixel between frames, the corresponding displacement vector data is extracted. The displacement vector of each pixel is accumulated to obtain the inter-frame pixel offset vector data.

[0014] S102: Based on the inter-frame pixel offset vector data, the position of the pixels in each frame of the image is adjusted, the actual position of the pixels is corrected by accumulating the offset vector, and each frame of the image is rearranged according to the adjusted position of the pixels to obtain the position adjusted image sequence.

[0015] S103: Based on the position adjustment image sequence, extract the boundary feature points in each frame image, compare the position differences of the feature points in adjacent frames point by point, adjust the inter-frame alignment error by calculating the average displacement of the feature points, and perform global alignment processing on the boundary feature points of each frame in the image sequence to obtain time series image correction data.

[0016] As a further aspect of the present invention, based on the time-series image correction data, the first frame image in the image sequence is selected, tumor seed points in the first frame image are calibrated, and the outer contour of the tumor region is extracted point by point by calculating the intensity difference and gradient change between the seed points and neighboring pixels to generate tumor region boundary data. The specific steps are as follows:

[0017] S201: Based on the time-series image correction data, select the first frame image in the image sequence, mark the center point of the tumor region in the image as the seed point, scan the neighboring pixels around the seed point point by point, and generate neighborhood intensity difference data by calculating the gray intensity difference between the seed point and the neighboring pixels.

[0018] S202: Based on the neighborhood intensity difference data, gradually expand the neighborhood scanning range, calculate the gray-level gradient change value in the expanded area, and use the set gradient change threshold as the expansion stop condition to adjust the expansion range, match the critical value of gray-level gradient change, and obtain the gradient data of the expanded area.

[0019] S203: Based on the gradient data of the extended region, extract the boundary pixels of the extended region, perform fitting analysis on the boundary pixels, and construct the outer contour of the tumor region point by point by determining the belonging relationship of the boundary pixels to obtain the boundary data of the tumor region.

[0020] As a further aspect of the present invention, the steps of calculating the position change of each pixel in the time series based on the tumor region boundary data, comparing the differences in boundary coordinates in adjacent frames, and combining the frame sequence changes to cumulatively analyze the overall change trend and generate dynamic change data of the tumor boundary are as follows:

[0021] S301: Based on the tumor region boundary data, extract the pixel coordinates of the tumor boundary in the image frame by frame, calculate the position difference of each pixel in adjacent frames, and obtain the boundary pixel position change data by accumulating the position change of adjacent frames and calculating the cumulative change of the pixel.

[0022] S302: Based on the boundary pixel position change data, the change amount of the tumor boundary in each frame is statistically analyzed, the expansion rate of the boundary coordinate change is analyzed, and the boundary expansion trend is analyzed by accumulating the expansion rate in the frame sequence to obtain the boundary expansion rate data.

[0023] S303: Based on the boundary expansion rate data, compare the complexity differences of the boundary coordinates of adjacent frames frame by frame, calculate the overall change trend by statistically analyzing the differences in coordinate changes in each frame, and generate dynamic change data of the tumor boundary.

[0024] As a further aspect of the present invention, the formula for calculating the overall trend of change is:

[0025]

[0026] Where S is the cumulative trend value of the boundary expansion rate, w i C represents the weighting coefficient of the i-th frame. i and C i-1 E represents the set of boundary coordinates for the i-th frame and the (i-1)-th frame, respectively. i and E i-1 T represents the boundary spread rate of the i-th frame and the (i-1)-th frame, respectively. i and T i-1 α represents the timestamps of the i-th frame and the (i-1)-th frame, respectively. α is the spread rate adjustment parameter, β is the time effect adjustment parameter, and n represents the total number of image frames.

[0027] As a further aspect of the present invention, based on the dynamic change data of the tumor boundary, the steps of calculating the gradient direction and gradient magnitude changes of each pixel, statistically analyzing the gradient distribution variance within the target region, determining the uniformity of the internal structure of the region by the consistency of the gradient change direction of the pixels, and analyzing the density differences of the target region to obtain density distribution characteristic data are as follows:

[0028] S401: Based on the dynamic change data of the tumor boundary, extract the gray value within the tumor area pixel by pixel, calculate the gradient direction of each pixel, and obtain the pixel gradient direction change data by comparing the change in pixel gradient direction point by point.

[0029] S402: Based on the pixel gradient direction change data, the local consistency and distribution difference of the gradient direction are statistically analyzed in each region, the consistency ratio of the local gradient direction is calculated, and the uniformity of the internal structure of the tumor region is analyzed by accumulating the change of the consistency ratio to obtain gradient consistency distribution data.

[0030] S403: Based on the gradient consistency distribution data, extract the range and magnitude of density gradient changes in each region, statistically analyze the distribution characteristics of density differences, and analyze the overall density change trend by combining the uniformity of gradient distribution within the target region to obtain density distribution characteristic data.

[0031] As a further aspect of the present invention, based on the dynamic change data of the tumor boundary and the density distribution characteristic data, the changes of the tumor region in the time series are analyzed. The consistency between the tumor volume expansion rate, the change amplitude of boundary complexity, and the change direction of density gradient is calculated frame by frame. Combined with the change trend in the time dimension, quantitative indicators of regional morphological and structural changes are extracted to predict the changes in tumor treatment effect in the future period. The specific steps for obtaining the dynamic characteristic indicators for efficacy prediction are as follows:

[0032] S501: Based on the dynamic change data of the tumor boundary and the density distribution characteristic data, the pixel position and density value in the tumor area are extracted frame by frame, the rate of change of tumor volume in each frame of image is calculated, the volume change in adjacent frames is compared, and the expansion trend in the time series is statistically analyzed to obtain the volume expansion rate data.

[0033] S502: Based on the volume expansion rate data, extract the coordinate set of the tumor boundary frame by frame, calculate the change in the complexity of the boundary in each frame, and analyze the evolution characteristics of the tumor boundary complexity in the time dimension by statistically analyzing the cumulative trend of the complexity change in the time series, so as to obtain the boundary complexity change data.

[0034] S503: Based on the boundary complexity change data and volume expansion rate data, analyze the time series trend of density gradient change direction pixel by pixel, combine the volume expansion rate and boundary complexity change trend, extract quantitative indicators of regional morphological and structural changes, analyze tumor change indicators, predict changes in tumor treatment effect in future periods, and generate dynamic feature indicators for efficacy prediction.

[0035] As a further aspect of the present invention, the formula for analyzing tumor change indicators is as follows:

[0036]

[0037] Where I is a quantitative indicator of tumor changes, and ΔD j ΔV represents the change in boundary complexity at time j. j w represents the change in the rate of volume expansion at time j. D w is the weighting coefficient for the change in boundary complexity. V T is the weighting coefficient for the change in volume expansion rate. j and T j-1 γ represents the timestamps of time j and the previous time, respectively, γ is the time smoothing parameter, and N is the total number of time points.

[0038] An image-based tumor treatment efficacy prediction system, wherein the image-based tumor treatment efficacy prediction system is used to execute the above-described image-based tumor treatment efficacy prediction method, the system comprising:

[0039] The image sequence correction module extracts the inter-frame pixel position coordinates of the CT image data based on CT image data at multiple time points before and after treatment, calculates the offset vector of the inter-frame pixel position, performs inter-frame position calibration, and generates time-series image correction data.

[0040] Based on the time-series image correction data, the boundary recognition module selects the first frame image in the image sequence, marks the tumor seed points in the first frame image, and extracts the outer contour of the tumor region point by point by calculating the intensity difference and gradient change between the seed points and neighboring pixels, thereby generating tumor region boundary data.

[0041] Based on the tumor region boundary data, the boundary change analysis module calculates the position change of each pixel in the time series, compares the differences in boundary coordinates in adjacent frames, combines the frame sequence change, accumulates and analyzes the overall change trend, and generates dynamic change data of the tumor boundary.

[0042] Based on the dynamic change data of the tumor boundary, the density distribution analysis module calculates the gradient direction and gradient magnitude changes of each pixel, calculates the gradient distribution variance in the target area, and judges the uniformity of the internal structure of the region by the consistency of the gradient change direction of the pixels, analyzes the density difference of the target area, and obtains density distribution characteristic data.

[0043] The efficacy prediction and analysis module analyzes the changes in the tumor region over time based on the dynamic changes in the tumor boundary and the density distribution characteristics data. It calculates the consistency between the tumor volume expansion rate, the magnitude of changes in boundary complexity, and the direction of changes in density gradient frame by frame. Combining the changing trends over time, it extracts quantitative indicators of regional morphological and structural changes to predict changes in tumor treatment efficacy in future periods, thus obtaining dynamic characteristic indicators for efficacy prediction.

[0044] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0045] In this invention, precise calibration of inter-frame positions is achieved by extracting inter-frame pixel position coordinates and calculating offset vectors from multi-time-point CT image data. This provides a more accurate benchmark for the image sequence, ensuring the consistency and alignment of time-series images and effectively improving the accuracy and efficiency of subsequent image processing. By setting an expansion threshold and fitting regional boundaries, the outer contour of the tumor region is identified, and the set of pixel coordinates of the tumor region boundary is dynamically analyzed to evaluate its positional changes over time. This facilitates the accurate assessment of tumor expansion rate and morphological changes, providing key data for predicting tumor treatment response and making medical decisions more scientific and precise. The analysis of density distribution characteristics supplements the assessment of tumor region structural homogeneity, which is crucial for understanding the biological characteristics of tumors and provides a basis for personalized treatment prediction, effectively predicting the changing trends of tumor treatment. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a schematic diagram of the workflow of the present invention;

[0048] Figure 2 This is a detailed flowchart of S1 of the present invention;

[0049] Figure 3 This is a detailed flowchart of the S2 process of the present invention;

[0050] Figure 4 This is a detailed flowchart of the S3 process of the present invention;

[0051] Figure 5 This is a detailed flowchart of the S4 process of the present invention;

[0052] Figure 6 This is a detailed flowchart of S5 of the present invention;

[0053] Figure 7 This is a system flowchart of the present invention. Detailed Implementation

[0054] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0055] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0056] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, their intended meanings are consistent.

[0057] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0058] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0059] Please see Figure 1 This invention provides a technical solution: an image-based method for predicting the efficacy of tumor treatment, comprising the following steps:

[0060] S1: Based on CT image data at multiple time points before and after treatment, extract the inter-frame pixel position coordinates of the CT image data, calculate the offset vector of the inter-frame pixel position, adjust the position of each frame in the image sequence, align with the image boundary feature points, perform inter-frame position calibration, and generate time series image correction data.

[0061] S2: Based on time-series image correction data, select the first frame image in the image sequence, mark the tumor seed points in the first frame image, calculate the intensity difference and gradient change between the seed points and neighboring pixels, limit the expansion area by setting an expansion threshold, and perform region boundary fitting. Combined with the attribution analysis of region boundary pixels, extract the outer contour of the tumor region point by point to generate tumor region boundary data.

[0062] S3: Based on tumor region boundary data, extract the set of pixel coordinates of tumor boundary in the image frame by frame, calculate the position change of each pixel in the time series, statistically analyze the expansion rate of tumor region boundary and the trend of boundary coordinate change in each frame, statistically compare the differences of boundary coordinates in adjacent frames, combine the frame sequence change, cumulatively analyze the overall change trend, and generate dynamic change data of tumor boundary.

[0063] S4: Based on the dynamic change data of tumor boundary, analyze the gray-level distribution characteristics within the tumor region pixel by pixel, calculate the gradient direction and gradient magnitude change of each pixel, statistically analyze the gradient distribution variance within the target region, and judge the uniformity of the internal structure of the region by the consistency of the gradient change direction of the pixels. Combined with the density gradient change range and magnitude, analyze the density difference of the target region to obtain density distribution characteristic data.

[0064] S5: Based on dynamic changes in tumor boundaries and density distribution characteristics, the changes in tumor regions over time are analyzed. The consistency between the tumor volume expansion rate, the magnitude of changes in boundary complexity, and the direction of changes in density gradient is calculated frame by frame. Combined with the changing trends over time, the dynamic characteristics of the tumor region are analyzed, quantitative indicators of regional morphological and structural changes are extracted, and changes in tumor treatment effects in future periods are predicted to obtain dynamic characteristic indicators for efficacy prediction.

[0065] Time-series image correction data includes corrected pixel spatial coordinates, grayscale intensity matching values ​​between sequence frames, and overall corrected positional consistency index. Tumor region boundary data includes the coordinate set of boundary pixels, intensity difference values ​​inside and outside the boundary, and cumulative area of ​​region expansion. Tumor boundary dynamic change data includes boundary coordinate changes, expansion rate in the time dimension, and complexity change value. Density distribution characteristic data includes pixel gradient direction consistency value, gradient amplitude distribution variance, and density change range. Dynamic feature indicators for efficacy prediction include volume expansion rate, density gradient change rate, and boundary complexity change trend.

[0066] Please see Figure 2 Based on CT image data from multiple time points before and after treatment, the following steps are taken to extract the inter-frame pixel position coordinates from the CT image data, calculate the inter-frame pixel position offset vector, adjust the position of each frame in the image sequence, align with image boundary feature points, perform inter-frame position calibration, and generate time-series image correction data:

[0067] S101: Based on CT image data from multiple time points before and after treatment, the pixel position coordinates in each frame are extracted. By calculating the spatial position difference of each pixel between frames, the corresponding displacement vector data is extracted. The displacement vector of each pixel is accumulated to obtain the inter-frame pixel offset vector data.

[0068] Based on CT image data from multiple time points before and after treatment, the pixel coordinates of each frame of CT image are obtained. The image data is loaded into image processing tools, such as ImageJ or MATLAB, and the pixel positions are extracted through image segmentation and recognition algorithms. The coordinates of each pixel are automatically identified by changes in image grayscale values. After extraction, a data matrix composed of frame number and pixel position is generated. Then, the spatial position difference of each pixel between frames is calculated. The difference is analyzed point by point using the Euclidean distance change formula between pixel coordinates between frames. By traversing the position change of each pixel, the corresponding displacement vector data is generated. All displacement vectors are accumulated and superimposed to obtain the overall offset vector data of pixels between frames. The data can be directly exported as a vector field representation to correct position changes in the image sequence.

[0069] S102: Based on the inter-frame pixel offset vector data, the position of the pixels in each frame of the image is adjusted. The actual position of the pixels is corrected by accumulating the offset vector. Each frame of the image is rearranged according to the adjusted position of the pixels to obtain the position-adjusted image sequence.

[0070] Based on inter-frame pixel offset vector data, the position of pixels in each frame of the image is adjusted. The offset vector data is normalized to ensure that the offset value range is consistent between different images. By analyzing the magnitude and direction of the offset vector of each pixel, the position of each pixel in the corresponding frame of the image is adjusted point by point. This includes accumulating the offset vector frame by frame starting from the first frame to form an adjusted position mapping matrix. At the same time, the corrected coordinates are reapplied to each frame of the image. The adjusted image sequence needs to be visually compared and verified to ensure that the geometric position change of each frame of the image conforms to the actual observation. This generates an image sequence with adjusted position.

[0071] S103: Based on the position adjustment image sequence, extract the boundary feature points in each frame of the image, compare the position differences of the feature points in adjacent frames point by point, adjust the inter-frame alignment error by calculating the average displacement of the feature points, and perform global alignment processing on the boundary feature points of each frame in the image sequence to obtain time series image correction data.

[0072] Based on the position adjustment of the image sequence, it is necessary to extract the boundary feature points in each frame of the image. This includes applying edge detection algorithms such as Canny or Sobel, locating edge pixels by calculating the gradient changes of each frame, analyzing the coordinates of the extracted edge feature points, comparing the position changes of feature points in adjacent frames one by one, calculating the average displacement of each feature point through the algorithm and adjusting the inter-frame alignment error, adjusting the position of the global boundary feature points in each frame of the image sequence by accumulating vector fields to ensure global alignment of feature points, and performing a global consistency check on the adjusted data. Finally, the corrected time-series image data is output using image stitching software or programming tools to form time-series image correction data.

[0073] Please see Figure 3 Based on time-series image correction data, the first frame of the image sequence is selected, and tumor seed points in the first frame are labeled. By calculating the intensity difference and gradient change between the seed points and neighboring pixels, the expansion region is limited by setting an expansion threshold, and the region boundary is fitted. Combined with the attribution analysis of the region boundary pixels, the outer contour of the tumor region is extracted point by point to generate tumor region boundary data. The specific steps are as follows:

[0074] S201: Based on time-series image correction data, select the first frame image in the image sequence, mark the center point of the tumor region in the image as the seed point, scan the neighboring pixels around the seed point point by point, and generate neighborhood intensity difference data by calculating the gray intensity difference between the seed point and the neighboring pixels.

[0075] Based on time-series image correction data, the first frame of the image sequence is selected, and the center point of the tumor region in the image is marked as the seed point. The first frame image is loaded using image processing tools, and the center point of the tumor is selected as the seed point using mouse click or automatic segmentation algorithm. Then, the neighboring pixel data around the seed point is extracted, and the gray value of the surrounding pixels is checked point by point using a neighborhood scanning algorithm. Neighborhood intensity difference data is generated by calculating the gray intensity difference between the seed point and each surrounding pixel. The process can be completed based on the gray-level difference calculation formula. The corresponding calculation is performed for each pixel to generate neighborhood intensity difference data.

[0076] S202: Based on the intensity difference data of the neighborhood, the neighborhood scanning range is gradually expanded, the gray-level gradient change value in the expanded area is calculated, and the expansion range is adjusted with the set gradient change threshold as the expansion stop condition to match the critical value of gray-level gradient change and obtain the gradient data of the expanded area.

[0077] Based on neighborhood intensity difference data, the neighborhood scanning range is gradually expanded. The neighborhood is expanded by setting an initial neighborhood radius and an incremental step size, and the gray-level gradient change value in the expanded area is calculated. The pixel gradient change in the expanded area is calculated point by point using the gray-level gradient change formula. A gradient change threshold is set as the expansion stopping condition. When the gray-level gradient change value in the expanded neighborhood is lower than the set threshold, the expansion stops. By dynamically adjusting the expansion range to match the critical value of gray-level gradient change, the gradient change data of the expanded area is obtained. The data indicates the possible location of the tumor region boundary and provides the data basis for contour fitting.

[0078] S203: Based on the gradient data of the extended region, extract the boundary pixels of the extended region, perform fitting analysis on the boundary pixels, and construct the outer contour of the tumor region point by point by determining the belonging relationship of the boundary pixels to obtain the boundary data of the tumor region.

[0079] Based on the gradient data of the extended region, the boundary pixels of the extended region are extracted. Edge detection algorithms such as Canny or Sobel are applied to locate the boundary pixels. The extracted boundary pixels are fitted and analyzed point by point. The fitting curve is constructed based on the coordinate data of the boundary pixels. The boundary fitting result is optimized by determining the belonging relationship of the boundary pixels. The outer contour of the tumor region is constructed point by point. At the same time, the fitted boundary is smoothed to reduce abrupt changes or noise interference. The tumor region boundary data is obtained, providing complete boundary information for subsequent data analysis and visualization.

[0080] Please see Figure 4Based on tumor region boundary data, the specific steps for generating dynamic tumor boundary change data are as follows: First, the set of pixel coordinates of the tumor boundary in the image is extracted frame by frame. Then, the positional change of each pixel in the time series is calculated. The expansion rate of the tumor region boundary and the trend of boundary coordinate changes in each frame are statistically analyzed. The differences in boundary coordinates between adjacent frames are statistically compared. Finally, the overall trend is cumulatively analyzed by combining the frame sequence changes.

[0081] S301: Based on tumor region boundary data, extract the pixel coordinates of tumor boundaries in the image frame by frame, calculate the position difference of each pixel in adjacent frames, and obtain boundary pixel position change data by accumulating the position changes in adjacent frames and calculating the cumulative changes of pixels.

[0082] Based on tumor region boundary data, the pixel coordinates of tumor boundaries in the images are extracted frame by frame. By reading the boundary data of each frame, the boundary pixel coordinates are analyzed and two-dimensional spatial coordinates are extracted. The position difference of each pixel in adjacent frames is calculated frame by frame. The displacement of each pixel is obtained by using the difference calculation method, including extracting the X and Y coordinates of the same pixel in adjacent frames, calculating the magnitude and direction of the displacement vector, and accumulating the difference values ​​of each pixel to form a statistical result. The boundary pixel position change data is obtained, and the data is finally saved in the form of a list or matrix.

[0083] S302: Based on the boundary pixel position change data, the change in the tumor boundary in each frame is statistically analyzed, the expansion rate of the boundary coordinate change is analyzed, and the boundary expansion trend is analyzed by accumulating the expansion rate in the frame sequence to obtain the boundary expansion rate data.

[0084] Based on boundary pixel position change data, the change in tumor boundary in each frame is statistically analyzed. By classifying the pixel position change data within each frame sequence, the boundary change of adjacent frames is integrated into the statistical process of inter-frame expansion rate. The expansion rate of boundary coordinate change is specifically calculated by performing time-step calculation on the cumulative position change value. By accumulating the specific values ​​of the expansion rate in the frame sequence, the overall change trend is evaluated to obtain boundary expansion rate data. The data represents the boundary expansion dynamics of the tumor region and is used for subsequent dynamic characteristic analysis.

[0085] S303: Based on the boundary expansion rate data, the complexity difference of the boundary coordinates of adjacent frames is compared frame by frame. By statistically analyzing the differences in coordinate changes in each frame, the overall trend of change is calculated, and dynamic change data of the tumor boundary is generated.

[0086] Based on the boundary expansion rate data, the complexity differences of the boundary coordinates of adjacent frames are compared frame by frame. The boundary coordinate data of each frame is extracted, and its geometric complexity is analyzed, including the length change, curvature change, and uniformity of pixel distribution of the boundary lines. The coordinate change data of each frame is statistically analyzed, and the overall change trend is calculated to describe the overall change trend. Through the analysis of the complexity data, dynamic change data of the tumor boundary is generated. The data is the basis for a complete record of boundary changes and can intuitively display the dynamic characteristics of tumor expansion.

[0087] The formula for calculating the overall trend of change is:

[0088]

[0089] Where S is the cumulative trend value of the boundary expansion rate, w i C represents the weighting coefficient of the i-th frame. i and C i-1 E represents the set of boundary coordinates for the i-th frame and the (i-1)-th frame, respectively. i and E i-1 T represents the boundary spread rate of the i-th frame and the (i-1)-th frame, respectively. i and T i-1 α represents the timestamps of the i-th frame and the (i-1)-th frame, respectively. α is the spread rate adjustment parameter, β is the time effect adjustment parameter, and n represents the total number of image frames.

[0090] formula:

[0091]

[0092] Parameter meanings and acquisition methods:

[0093] w i The weighting coefficients for the i-th frame reflect the importance of different frames in the overall analysis. For example, more recent frames may be assigned higher weights because their data is more relevant. These weights can be calculated automatically through expert input or based on previous analysis results.

[0094] C i and C i-1 : The set of boundary coordinates of the i-th frame and the (i-1)-th frame, extracted from image processing algorithms (such as Canny edge detection).

[0095] E i and E i-1 Boundary spread rate: This represents the rate at which the boundary changes between two consecutive frames. It can be calculated by measuring changes in the boundary region (e.g., area or perimeter).

[0096] T i and T i-1: Timestamp, indicating the specific time when an image frame was recorded. It is usually obtained directly from the image file's metadata or the image acquisition system.

[0097] α and β: Adjustment parameters, fixed values ​​derived from experiments or actual observations, used to adjust the weights in the calculation of the spread rate and time effects. Typically, these parameters are optimally estimated in preliminary experiments using optimization algorithms.

[0098] Calculation Example

[0099] The following two frames of image data are provided:

[0100] Frame 1 data:

[0101] Weight w1 = 1.0, boundary coordinates C1 = {(10,20),(15,25)}, boundary expansion rate E1 = 0.0 (initial frame), timestamp T1 = 100 (seconds).

[0102] Frame 2 data:

[0103] Weight w2 = 1.5, boundary coordinates C2 = {(12,22),(17,27)}, boundary expansion rate E2 = 0.4, timestamp T2 = 101 (seconds).

[0104] The Euclidean distance is used to calculate the sum of the distances between all corresponding points between two frames:

[0105]

[0106] Calculate the time difference T2-T1: 101-100 = 1 second;

[0107] Substitute the value into the formula:

[0108]

[0109] The calculation result S represents the change in tumor boundary from the first frame to the second frame. Considering the combined effects of weight, expansion rate and time difference, this quantitative analysis is very useful for understanding the growth rate and dynamic changes of tumors.

[0110] Please see Figure 5 Based on dynamic changes in tumor boundaries, the gray-level distribution characteristics within the tumor region are analyzed pixel-by-pixel. The gradient direction and magnitude changes of each pixel are calculated, the gradient distribution variance within the target region is statistically analyzed, and the uniformity of the internal structure of the region is determined by the consistency of the gradient change direction of each pixel. Combined with the range and magnitude of density gradient changes, the density differences in the target region are analyzed. The specific steps to obtain density distribution characteristic data are as follows:

[0111] S401: Based on the dynamic change data of the tumor boundary, the gray value within the tumor area is extracted pixel by pixel, the gradient direction of each pixel is calculated, and the pixel gradient direction change data is obtained by comparing the change in pixel gradient direction point by point.

[0112] Based on dynamic changes in tumor boundaries, grayscale values ​​within the tumor region are extracted pixel by pixel. Image processing tools are used to load the grayscale matrix data of the tumor region. Each pixel is scanned point by point to extract its grayscale value. The gradient direction of each pixel is then calculated using gradient direction calculation methods, such as the formula for calculating the gradient of pixel grayscale value changes. The changes in pixel gradient directions are compared point by point. By calculating the angular differences in the gradient directions of pixels within their spatial neighborhood, the pixel gradient direction change data is statistically obtained. This data can be saved as a gradient direction change matrix for subsequent analysis.

[0113] S402: Based on pixel gradient direction change data, the local consistency and distribution difference of gradient direction are statistically analyzed in each region, the consistency ratio of local gradient direction is calculated, and the uniformity of the internal structure of the tumor region is analyzed by accumulating the change of the consistency ratio to obtain gradient consistency distribution data.

[0114] Based on pixel gradient direction change data, the local consistency and distribution differences of gradient direction are statistically analyzed in each region. This includes grouping the pixel gradient direction data in each region, calculating the local consistency ratio by counting the number of pixels with the same gradient direction, evaluating the distribution characteristics of gradient direction in the target region using angular distribution analysis, gradually accumulating the change in consistency ratio, performing spatiotemporal analysis on gradient consistency in the tumor region, and quantitatively evaluating the dynamic change trend of gradient consistency ratio to form gradient consistency distribution data.

[0115] S403: Based on gradient uniformity distribution data, extract the range and magnitude of density gradient changes in each region, statistically analyze the distribution characteristics of density differences, and analyze the overall density change trend by combining the uniformity of gradient distribution within the target region to obtain density distribution characteristic data.

[0116] Based on gradient consistency distribution data, the range and magnitude of density gradient changes are extracted region by region. By analyzing the gradient consistency distribution data, the range of density difference changes is extracted. The statistical method of pixel gradient density within the region is used to gradually statistically analyze the characteristics of density difference distribution. Combined with the uniformity of gradient distribution within the target region, the gradient difference of density change is analyzed point by point. The trend of density gradient change in different regions is statistically analyzed to complete the generation of overall density distribution characteristic data. The data is finally used as the basis for dynamic characteristic analysis of density distribution and for subsequent processing.

[0117] Please see Figure 6Based on dynamic changes in tumor boundaries and density distribution characteristics, this study analyzes the changes in tumor regions over time. It calculates the consistency between the tumor volume expansion rate, the magnitude of changes in boundary complexity, and the direction of density gradient changes frame by frame. Combining these temporal trends, the study analyzes the dynamic characteristics of the tumor region, extracts quantitative indicators of regional morphological and structural changes, and predicts future changes in tumor treatment efficacy. The specific steps for obtaining dynamic characteristic indicators for efficacy prediction are as follows:

[0118] S501: Based on the dynamic change data of tumor boundaries and density distribution characteristics, the pixel position and density value within the tumor area are extracted frame by frame, the rate of change of tumor volume in each frame is calculated, the volume change in adjacent frames is compared, and the expansion trend in the time series is statistically analyzed to obtain the volume expansion rate data.

[0119] Based on dynamic changes in tumor boundaries and density distribution characteristics, pixel positions and density values ​​within the tumor region are extracted frame by frame. By processing the image sequence frame by frame, the tumor region in the image is extracted, and the grayscale value of each pixel is read. Combined with its position coordinates, a data matrix containing spatial position and grayscale value is constructed. Then, the rate of change of tumor volume in each frame of the image is calculated. The volume data of each frame is differentially processed using a three-dimensional reconstruction method to obtain the rate of change of volume. The volume change of adjacent frames is compared, and the expansion trend in the time series is analyzed by the cumulative volume change method. The volume expansion rate data is statistically analyzed and plotted to obtain the time series curve.

[0120] S502: Based on the volume expansion rate data, the coordinate set of the tumor boundary is extracted frame by frame, the change in boundary complexity of each frame is calculated, and the evolution characteristics of tumor boundary complexity in the time dimension are analyzed by statistically analyzing the cumulative trend of complexity change in the time series, so as to obtain boundary complexity change data.

[0121] Based on volume expansion rate data, the coordinate set of tumor boundaries is extracted frame by frame. Complexity analysis is performed on the boundary of each frame. By statistically analyzing indicators such as the curvature change of the tumor boundary, the uniformity of point distribution, and the contour length, the change in complexity of the boundary of each frame is calculated. The calculation results are processed by time accumulation, and a cumulative trend chart of complexity change in the time series is drawn. The evolution characteristics of tumor boundary complexity in the time dimension are analyzed, and the characteristics are quantified into a numerical set to obtain boundary complexity change data.

[0122] S503: Based on boundary complexity change data and volume expansion rate data, analyze the time series trend of density gradient change direction pixel by pixel, combine volume expansion rate and boundary complexity change trend to extract quantitative indicators of regional morphological and structural changes, analyze tumor change indicators, predict changes in tumor treatment effect in future periods, and generate dynamic feature indicators for efficacy prediction.

[0123] Based on boundary complexity variation data and volume expansion rate data, the time series trend of density gradient change direction is analyzed pixel by pixel. The density gradient direction of each pixel is analyzed frame by frame. By comparing the density gradient change angle and magnitude of adjacent frames in the time series, the dynamic changes of each pixel are accumulated to the global region. Combined with the volume expansion rate and boundary complexity variation trend, tumor change indicators are analyzed. The change trend of tumor is analyzed by the statistical law of density distribution in the region. Quantitative indicators are used to predict the tumor treatment effect in the future period, generating dynamic feature indicators for efficacy prediction.

[0124] The formula for analyzing tumor change indicators is:

[0125]

[0126] Where I is a quantitative indicator of tumor changes, and ΔD j ΔV represents the change in boundary complexity at time j. j w represents the change in the rate of volume expansion at time j. D w is the weighting coefficient for the change in boundary complexity. V T is the weighting coefficient for the change in volume expansion rate. j and T j-1 γ represents the timestamps of time j and the previous time, respectively, γ is the time smoothing parameter, and N is the total number of time points.

[0127] formula:

[0128]

[0129] Parameter meanings and acquisition methods:

[0130] ΔD j : The change in boundary complexity at time j, representing the change in boundary structure complexity from the previous time to the current time. Boundary complexity is extracted using image analysis algorithms (such as fractal dimension calculation), and then the difference between adjacent time points is taken.

[0131] ΔV j The change in the tumor volume expansion rate at time j represents the change in the tumor volume growth rate from the previous time to the current time. The difference in the volume change rate is calculated by measuring the tumor volume in consecutive frames.

[0132] w D and w V : Weighting coefficients, used to adjust the impact of boundary complexity and volume expansion rate on the overall performance. Their value range is usually determined through experimental optimization; here, we set w. D =0.6, w V =0.4.

[0133] T j and T j-1 The timestamps at times j and j-1 are obtained directly from the image data acquisition time and are usually in days.

[0134] γ: Time smoothing parameter, used to prevent the denominator from being too small due to excessively short time intervals. In this example, γ is set to 0.05.

[0135] Calculation Example

[0136] The data is set at two time points as follows:

[0137] Time point 1: Timestamp: T1 = 0 days, Boundary complexity: D1 = 1.2 (set value), Volume: V1 = 1000 cubic millimeters; Time point 2: Timestamp: T2 = 7 days, Boundary complexity: D2 = 1.5 (set value), Volume: V2 = 1200 cubic millimeters.

[0138] Calculate the change in boundary complexity ΔD2:

[0139] ΔD2=D2-D1=1.5-1.2=0.3;

[0140] Calculate the change in volume expansion rate ΔV2:

[0141] Volume expansion rate:

[0142]

[0143] Change in volume expansion rate (relative to initial value):

[0144] ΔV2=E2-0=28.57-0=28.57 cubic millimeters / day;

[0145] Calculation time interval:

[0146] |T2-T1|=|7-0|=7 days;

[0147] Substitute into the formula to calculate the comprehensive morphological index I:

[0148] The formula is as follows:

[0149]

[0150] The obtained index I≈4.37 represents a comprehensive quantitative result combining changes in boundary complexity and volume expansion rate, used to assess the dynamic trend of tumor changes. This value can serve as a dynamic characteristic indicator for predicting treatment efficacy, and can be compared with the index values ​​at subsequent time points to form a trend analysis.

[0151] Please see Figure 7An image-based tumor efficacy prediction system is used to execute the aforementioned image-based tumor efficacy prediction method. The system includes:

[0152] The image sequence correction module extracts the inter-frame pixel position coordinates of the CT image data based on CT image data at multiple time points before and after treatment, calculates the offset vector of the inter-frame pixel position, performs inter-frame position calibration, and generates time-series image correction data.

[0153] The boundary recognition module is based on time-series image correction data. It selects the first frame image in the image sequence, marks the tumor seed points in the first frame image, and extracts the outer contour of the tumor region point by point by calculating the intensity difference and gradient change between the seed points and neighboring pixels to generate tumor region boundary data.

[0154] The boundary change analysis module calculates the positional change of each pixel in the time series based on tumor region boundary data, compares the differences in boundary coordinates in adjacent frames, combines the frame sequence change data, accumulates and analyzes the overall change trend, and generates dynamic change data of tumor boundary.

[0155] The density distribution analysis module calculates the gradient direction and gradient magnitude changes of each pixel based on the dynamic change data of the tumor boundary, statistically analyzes the gradient distribution variance within the target area, and judges the uniformity of the internal structure of the region by the consistency of the gradient change direction of the pixels, analyzes the density difference of the target area, and obtains density distribution characteristic data.

[0156] The efficacy prediction and analysis module analyzes the changes in the tumor region over time based on dynamic changes in tumor boundaries and density distribution characteristics. It calculates the consistency between the tumor volume expansion rate, the magnitude of changes in boundary complexity, and the direction of changes in density gradient frame by frame. Combining the changing trends over time, it extracts quantitative indicators of regional morphological and structural changes to predict changes in tumor treatment efficacy in future periods, thus obtaining dynamic characteristic indicators for efficacy prediction.

[0157] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0158] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0159] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0160] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0161] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0162] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0163] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0164] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0165] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0166] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An image-based method for predicting the efficacy of tumor treatment, characterized in that, Includes the following steps: S1: Based on CT image data at multiple time points before and after treatment, extract the inter-frame pixel position coordinates of the CT image data, calculate the offset vector of the inter-frame pixel position, perform inter-frame position calibration, and generate time-series image correction data. S2: Based on the time-series image correction data, select the first frame image in the image sequence, mark the tumor seed points in the first frame image, and extract the outer contour of the tumor region point by point by calculating the intensity difference and gradient change between the seed points and neighboring pixels to generate tumor region boundary data. S3: Based on the tumor region boundary data, calculate the position change of each pixel in the time series, compare the difference of boundary coordinates in adjacent frames, combine the frame sequence change, accumulate and analyze the overall change trend, and generate dynamic change data of tumor boundary. S4: Based on the dynamic change data of the tumor boundary, calculate the gradient direction and gradient magnitude change of each pixel, statistically analyze the gradient distribution variance in the target area, and judge the uniformity of the internal structure of the region by the consistency of the gradient change direction of the pixels, analyze the density difference of the target area, and obtain density distribution characteristic data. S5: Based on the dynamic change data of the tumor boundary and the density distribution characteristic data, the changes of the tumor region in the time series are analyzed. The consistency between the tumor volume expansion rate, the change amplitude of boundary complexity and the change direction of density gradient is calculated frame by frame. Combined with the change trend in the time dimension, quantitative indicators of regional morphological and structural changes are extracted to predict the changes in tumor treatment effect in the future period and obtain dynamic characteristic indicators for efficacy prediction.

2. The image-based tumor treatment efficacy prediction method according to claim 1, characterized in that, The time-series image correction data includes corrected pixel spatial coordinates, grayscale intensity matching values ​​between sequence frames, and overall corrected positional consistency index. The tumor region boundary data includes the coordinate set of boundary pixels, intensity difference values ​​inside and outside the boundary, and cumulative area of ​​region expansion. The tumor boundary dynamic change data includes boundary coordinate changes, expansion rate in the time dimension, and complexity change value. The density distribution characteristic data includes pixel gradient direction consistency value, gradient amplitude distribution variance, and density change amplitude range. The efficacy prediction dynamic feature index includes volume expansion rate, density gradient change rate, and boundary complexity change trend.

3. The image-based tumor treatment efficacy prediction method according to claim 1, characterized in that, The specific steps for generating time-series image correction data based on CT image data from multiple time points before and after treatment are as follows: Inter-frame pixel coordinates are extracted from the CT image data; the offset vector of the inter-frame pixel positions is calculated; inter-frame position calibration is performed; and time-series image correction data is generated. S101: Based on CT image data from multiple time points before and after treatment, the pixel position coordinates in each frame are extracted. By calculating the spatial position difference of each pixel between frames, the corresponding displacement vector data is extracted. The displacement vector of each pixel is accumulated to obtain the inter-frame pixel offset vector data. S102: Based on the inter-frame pixel offset vector data, the position of the pixels in each frame of the image is adjusted, the actual position of the pixels is corrected by accumulating the offset vector, and each frame of the image is rearranged according to the adjusted position of the pixels to obtain the position adjusted image sequence. S103: Based on the position adjustment image sequence, extract the boundary feature points in each frame image, compare the position differences of the feature points in adjacent frames point by point, adjust the inter-frame alignment error by calculating the average displacement of the feature points, and perform global alignment processing on the boundary feature points of each frame in the image sequence to obtain time series image correction data.

4. The image-based tumor treatment efficacy prediction method according to claim 1, characterized in that, Based on the time-series image correction data, the following steps are taken to select the first frame image in the image sequence, mark the tumor seed points in the first frame image, and extract the outer contour of the tumor region point by point by calculating the intensity difference and gradient change between the seed points and neighboring pixels to generate tumor region boundary data: S201: Based on the time-series image correction data, select the first frame image in the image sequence, mark the center point of the tumor region in the image as the seed point, scan the neighboring pixels around the seed point point by point, and generate neighborhood intensity difference data by calculating the gray intensity difference between the seed point and the neighboring pixels. S202: Based on the neighborhood intensity difference data, gradually expand the neighborhood scanning range, calculate the gray-level gradient change value in the expanded area, and use the set gradient change threshold as the expansion stop condition to adjust the expansion range, match the critical value of gray-level gradient change, and obtain the gradient data of the expanded area. S203: Based on the gradient data of the extended region, extract the boundary pixels of the extended region, perform fitting analysis on the boundary pixels, and construct the outer contour of the tumor region point by point by determining the belonging relationship of the boundary pixels to obtain the boundary data of the tumor region.

5. The image-based tumor treatment efficacy prediction method according to claim 1, characterized in that, Based on the tumor region boundary data, the specific steps for calculating the positional change of each pixel in the time series, comparing the differences in boundary coordinates between adjacent frames, and combining the frame sequence changes to cumulatively analyze the overall trend and generate dynamic change data of the tumor boundary are as follows: S301: Based on the tumor region boundary data, extract the pixel coordinates of the tumor boundary in the image frame by frame, calculate the position difference of each pixel in adjacent frames, and obtain the boundary pixel position change data by accumulating the position change of adjacent frames and calculating the cumulative change of the pixel. S302: Based on the boundary pixel position change data, the change amount of the tumor boundary in each frame is statistically analyzed, the expansion rate of the boundary coordinate change is analyzed, and the boundary expansion trend is analyzed by accumulating the expansion rate in the frame sequence to obtain the boundary expansion rate data. S303: Based on the boundary expansion rate data, compare the complexity differences of the boundary coordinates of adjacent frames frame by frame, calculate the overall change trend by statistically analyzing the differences in coordinate changes in each frame, and generate dynamic change data of the tumor boundary.

6. The image-based tumor treatment efficacy prediction method according to claim 5, characterized in that, The formula for calculating the overall trend of change is: Where S is the cumulative trend value of the boundary expansion rate, w i C represents the weighting coefficient of the i-th frame. i and C i-1 E represents the set of boundary coordinates for the i-th frame and the (i-1)-th frame, respectively. i and E i-1 T represents the boundary spread rate of the i-th frame and the (i-1)-th frame, respectively. i and T i-1 α represents the timestamps of the i-th frame and the (i-1)-th frame, respectively. α is the spread rate adjustment parameter, β is the time effect adjustment parameter, and n represents the total number of image frames.

7. The image-based tumor treatment efficacy prediction method according to claim 1, characterized in that, Based on the dynamic change data of the tumor boundary, the steps of calculating the gradient direction and gradient magnitude change of each pixel, statistically analyzing the gradient distribution variance within the target region, determining the uniformity of the internal structure of the region by the consistency of the gradient change direction of each pixel, and analyzing the density difference of the target region to obtain density distribution characteristic data are as follows: S401: Based on the dynamic change data of the tumor boundary, extract the gray value within the tumor area pixel by pixel, calculate the gradient direction of each pixel, and obtain the pixel gradient direction change data by comparing the change in pixel gradient direction point by point. S402: Based on the pixel gradient direction change data, the local consistency and distribution difference of the gradient direction are statistically analyzed in each region, the consistency ratio of the local gradient direction is calculated, and the uniformity of the internal structure of the tumor region is analyzed by accumulating the change of the consistency ratio to obtain gradient consistency distribution data. S403: Based on the gradient consistency distribution data, extract the range and magnitude of density gradient changes in each region, statistically analyze the distribution characteristics of density differences, and analyze the overall density change trend by combining the uniformity of gradient distribution within the target region to obtain density distribution characteristic data.

8. The image-based tumor treatment efficacy prediction method according to claim 1, characterized in that, Based on the dynamic change data of tumor boundaries and density distribution characteristics, the changes of the tumor region over time are analyzed. The consistency between the tumor volume expansion rate, the magnitude of boundary complexity change, and the direction of density gradient change is calculated frame by frame. Combined with the change trend over time, quantitative indicators of regional morphological and structural changes are extracted to predict changes in tumor treatment efficacy in future periods. The specific steps for obtaining dynamic characteristic indicators for efficacy prediction are as follows: S501: Based on the dynamic change data of the tumor boundary and the density distribution characteristic data, the pixel position and density value in the tumor area are extracted frame by frame, the rate of change of tumor volume in each frame of image is calculated, the volume change in adjacent frames is compared, and the expansion trend in the time series is statistically analyzed to obtain the volume expansion rate data. S502: Based on the volume expansion rate data, extract the coordinate set of the tumor boundary frame by frame, calculate the change in the complexity of the boundary in each frame, and analyze the evolution characteristics of the tumor boundary complexity in the time dimension by statistically analyzing the cumulative trend of the complexity change in the time series, so as to obtain the boundary complexity change data. S503: Based on the boundary complexity change data and volume expansion rate data, analyze the time series trend of density gradient change direction pixel by pixel, combine the volume expansion rate and boundary complexity change trend, extract quantitative indicators of regional morphological and structural changes, analyze tumor change indicators, predict changes in tumor treatment effect in future periods, and generate dynamic feature indicators for efficacy prediction.

9. The image-based tumor treatment efficacy prediction method according to claim 8, characterized in that, The formula for analyzing tumor change indicators is as follows: Where I is a quantitative indicator of tumor changes, and ΔD j ΔV represents the change in boundary complexity at time j. j w represents the change in the rate of volume expansion at time j. D w is the weighting coefficient for the change in boundary complexity. V T is the weighting coefficient for the change in volume expansion rate. h and T j-1 γ represents the timestamps of time j and the previous time, respectively, γ is the time smoothing parameter, and N is the total number of time points.

10. An image-based tumor treatment efficacy prediction system, characterized in that, The image-based tumor treatment efficacy prediction method according to any one of claims 1-9, wherein the system comprises: The image sequence correction module extracts the inter-frame pixel position coordinates of the CT image data based on CT image data at multiple time points before and after treatment, calculates the offset vector of the inter-frame pixel position, performs inter-frame position calibration, and generates time-series image correction data. Based on the time-series image correction data, the boundary recognition module selects the first frame image in the image sequence, marks the tumor seed points in the first frame image, and extracts the outer contour of the tumor region point by point by calculating the intensity difference and gradient change between the seed points and neighboring pixels, thereby generating tumor region boundary data. Based on the tumor region boundary data, the boundary change analysis module calculates the position change of each pixel in the time series, compares the differences in boundary coordinates in adjacent frames, combines the frame sequence change, accumulates and analyzes the overall change trend, and generates dynamic change data of the tumor boundary. Based on the dynamic change data of the tumor boundary, the density distribution analysis module calculates the gradient direction and gradient magnitude changes of each pixel, calculates the gradient distribution variance in the target area, and judges the uniformity of the internal structure of the region by the consistency of the gradient change direction of the pixels, analyzes the density difference of the target area, and obtains density distribution characteristic data. The efficacy prediction and analysis module analyzes the changes in the tumor region over time based on the dynamic changes in the tumor boundary and the density distribution characteristics data. It calculates the consistency between the tumor volume expansion rate, the magnitude of changes in boundary complexity, and the direction of changes in density gradient frame by frame. Combining the changing trends over time, it extracts quantitative indicators of regional morphological and structural changes to predict changes in tumor treatment efficacy in future periods, thus obtaining dynamic characteristic indicators for efficacy prediction.

Images