A lung cancer radiotherapy dose optimization method based on internal tumor heterogeneity

Abstract

The application discloses a lung cancer radiotherapy dose optimization method based on tumor internal heterogeneity, which comprises the following steps: delineating a solid tumor target area on a CT image; constructing a three-dimensional neighborhood window with each voxel as the center, and extracting local image features of each voxel; constructing a voxel-level feature vector based on the local entropy value and the CT gray value of each voxel, clustering all voxels in the tumor target area, and dividing the solid tumor into several first habitat sub-regions with different imaging features; constructing and calculating a tumor heterogeneity index ITHscore; and giving each sub-region an individual radiotherapy dose according to the ITHscore value of the different sub-regions. By using the application, automatic identification of lung cancer internal heterogeneity sub-regions can be realized, the radiotherapy dose of a high radioresistance region is improved, the radiotherapy dose of a low radioresistance region is reduced, the local tumor control rate is improved, and the normal tissue radiotherapy toxicity is reduced, so that individualized precision radiotherapy can be implemented according to the specific condition of a patient.

Description

Technical Field

[0001] This invention relates to the field of radiotherapy dose optimization technology, and in particular to a method for optimizing radiotherapy dose for lung cancer based on tumor intratumoral heterogeneity. Background Technology

[0002] Lung cancer is one of the most common malignant tumors worldwide, with approximately 80%-85% being non-small cell lung cancer (NSCLC). Treatment strategies include surgery, chemotherapy, targeted therapy, and radiotherapy. For locally advanced NSCLC that cannot be surgically removed, radiotherapy plays an irreplaceable role as a primary means of local control. Current clinical radiotherapy guidelines primarily follow the principle of "uniform dose irradiation" of the solid tumor target volume, meaning a single, uniform radiation dose is administered within the radiographically delineated solid tumor (GTV). This "one-size-fits-all" dosing model ignores the spatial heterogeneity (Intra-tumor heterogeneity) arising from differences in genetics, phenotype, and function among different cells within the same tumor. Numerous studies have shown the existence of distinct heterogeneous subregions within tumors, including differences in tumor cell proliferation activity, hypoxia levels, and blood perfusion. These heterogeneous subregions exhibit varying sensitivities to radiation. Therefore, the current clinical application of a single, uniform radiation dose to solid tumors can lead to two adverse outcomes: on the one hand, insufficient radiation dose to the high radiation tolerance subregions within the tumor increases the risk of local recurrence and distant metastasis; on the other hand, excessive radiation dose to the low radiation tolerance subregions within the tumor results in excessive doses to organs adjacent to the tumor target area, such as the aorta, trachea, bronchial tree, esophagus, heart, and lungs, which can easily cause damage to these normal tissues or serious radiation complications.

[0003] Currently, although there are some studies based on imaging techniques, such as using PET to conduct localized doses to areas of tumors with high metabolic activity or low oxidation, these studies are mostly focused on empirical adjustments based on overall uniform doses, and lack non-invasive tumor subregion identification algorithms and methods for optimizing precise doses in subregions. Summary of the Invention

[0004] The technical problem to be solved by the embodiments of the present invention is to provide a method for optimizing radiotherapy dose for lung cancer based on intratumoral heterogeneity. This method is a new technology that can non-invasively identify heterogeneous subregions within the tumor and optimize individualized radiation doses for different subregions. This technology is expected to reduce the risk of tumor recurrence and metastasis, reduce the radiation dose to normal tissues, and thus significantly improve the quality of life of lung cancer patients.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide a method for optimizing radiotherapy dose for lung cancer based on tumor intratumor heterogeneity, comprising the following steps: S1: Obtain simulated localization CT images of lung cancer patients, and delineate the target area of ​​solid tumors based on the CT images; S2: Within the delineated tumor target area, a three-dimensional neighborhood window is constructed with each voxel as the center, and the local image features of each voxel are extracted. The local image features are local entropy. S3: Construct voxel-level feature vectors based on the local entropy value and CT gray value of each voxel, and use the K-means clustering algorithm to cluster all voxels in the tumor target area, dividing the solid tumor into several primary habitat subregions with different imaging features. S4: Based on the spatial connectivity characteristics of each primary habitat subregion after division, construct and calculate the tumor heterogeneity index ITHscore; S5: Based on the calculated tumor heterogeneity index (ITHscore) values ​​of different subregions within the tumor, individualized radiotherapy doses are given to each subregion.

[0006] Furthermore, the extraction of local entropy in S2 specifically involves: using each voxel... Centered on a 5×5×5 three-dimensional cubic neighborhood window, the local entropy of each voxel is calculated using the following formula:

[0007] in, Entropy(x i ) voxels x i Local entropy; Ω i Indicated by voxels x i The 3×3×3 neighborhood of the center; p(x j ) Representing the neighborhood Ω i Endogenous x j The probability of grayscale values ​​appearing; ∈ is a very small positive integer used to prevent the logarithm from being meaningless.

[0008] Furthermore, in step S3, before performing K-means clustering, a step of Z-score standardization of the local entropy features is included:

[0009] in, and They represent the first The mean and standard deviation of a local entropy feature within the tumor region.

[0010] Furthermore, in step S3, when using the K-means clustering algorithm to cluster all voxels, the number of cluster categories... K The selection range is from 3 to 9, and the stability and separation of different clustering structures are evaluated using the CH index, which is calculated using the following formula:

[0011] Among them, tr(B) k ): The trace of the inter-cluster discrete matrix Bk, representing the overall degree of dispersion among different cluster centers, used to measure inter-cluster separability; tr(W k ): Intraclass discrete matrix W k The trace represents the degree of dispersion among samples within the same cluster, and is used to measure intra-cluster tightness. N Total number of samples; in this study, this usually represents the total number of elements within the tumor ROI that participated in cluster analysis. K 1 Inter-class degrees of freedom; N K Intraclass degrees of freedom.

[0012] Furthermore, in step S3, after dividing the solid tumor into primary habitat subregions, a nested clustering step is further included: extracting the corresponding voxel set and its local entropy features within each primary habitat subregion to form an independent voxel-local entropy value matrix, and then executing the K-means clustering algorithm again to further subdivide each primary habitat subregion into multiple secondary habitat subregions.

[0013] Furthermore, the tumor heterogeneity index ITHscore is calculated according to the following formula:

[0014] in, S total The total number of all voxels in the tumor volume; V The number of subregions into which the tumor is divided; i Representing the i Subregions; S i,max For the first i The largest number of a certain type of voxel in a habitat subregion; n i For the first i The total number of primes in each habitat; the range of ITHscore values ​​is... [0,1] A higher value indicates greater intratumoral heterogeneity.

[0015] Furthermore, S5 specifically includes: S51: Obtain the standard radiotherapy dose of 60 Gy and its corresponding median tumor heterogeneity index (ITHscore). 总 ; S52: For a specific subregion A within a patient's tumor, the calculated tumor heterogeneity index is the ITHscore. A Then, an individualized radiotherapy dose D is given to subregion A. A Calculate using the following formula: .

[0016] Furthermore, following step S5, a radiotherapy planning step is also included: designing a radiotherapy plan according to the following dose constraints: Whole lung V5 ≤ 60%, V 10 ≤40%, V 20 ≤30%, average dose to the whole lung ≤15Gy; highest dose to the spinal cord ≤45Gy; highest dose to the esophagus ≤64Gy; cardiac V 30 ≤40%, V 40 ≤30%, average cardiac dose ≤26Gy; highest dose to target area <110% of prescription dose.

[0017] Furthermore, when optimizing radiotherapy plans, if there are multiple sub-regions with different prescription doses, a dose gradient descent gap with a width of 5-8 mm is established at the boundary between different sub-regions to prevent the formation of too many high-dose regions due to conflicting conditions during the optimization of different prescription doses.

[0018] Implementing the embodiments of the present invention has the following beneficial effects: (1) It can automatically identify heterogeneous subregions within lung cancer; (2) It can increase the radiotherapy dose in areas with high radiation tolerance; (3) It can reduce the radiotherapy dose in areas with low radiation tolerance; (4) It can improve the local control rate of tumors and reduce the radiotherapy toxicity of normal tissues; (5) It can implement individualized and precise radiotherapy according to the specific condition of the patient. Attached Figure Description

[0019] Figure 1 This is a flowchart of the technology for characterizing local tumor features, clustering, dividing habitat subregions, and calculating heterogeneity index in this invention; Figure 2 This is a schematic diagram of the framework of the present invention; Figure 3 This is a comparison of the dose distribution of two radiotherapy plans. Plan A is Plan 1: the entire solid tumor is treated with a radiotherapy plan of 60 Gy. Plan B is Plan 2: the solid tumor is divided into 3 sub-regions and radiotherapy plans are designed to be given to different sub-regions. Figure 4 yes Figure 3 Comparison of dose parameters for two radiotherapy plans in China. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings.

[0021] Combination Figure 1 , Figure 2 The method for optimizing radiotherapy dose for lung cancer based on intratumoral heterogeneity, as described in this embodiment of the invention, is implemented through the following steps.

[0022] S1: Obtain simulated localization CT images of lung cancer patients, and delineate the target area of ​​solid tumors and normal organs based on the CT images.

[0023] The patient underwent a simulated localization CT scan, and the target area was delineated on the obtained CT images.

[0024] The lung cancer CT images in this invention are acquired using a radiotherapy-specific CT simulation positioning machine, and are used for subsequent tumor target delineation, tumor internal heterogeneity analysis, and radiotherapy planning and optimization.

[0025] (1) Patient position fixed The patient was placed in a supine position with both arms raised and clasped behind their head. A vacuum bag and wing-shaped plate were used to fix the position. The patient held their breath to minimize the impact of respiratory movements and positional errors on image quality and radiotherapy positioning accuracy.

[0026] (2) CT simulation localization scan range The scan area typically extends from the upper margin of the cricoid cartilage to the lower margin of both adrenal glands, completely covering the lung lesion. Intravenous contrast-enhanced CT scans are used to improve the clarity of tumor boundaries.

[0027] (3) Scanning parameters Tube voltage: 120 kVp, tube current: 200-350 mA; layer thickness: 3 mm, interlayer spacing: 3 mm; reconstruction matrix size: 512 × 512; field of view (FOV): 500 mm; pitch: 0.8-1.5. Reconstruction algorithm: Standard lung window reconstruction algorithm.

[0028] (4) CT image output After the scan was completed, all CT images were exported in DICOM standard format and transferred to the radiotherapy planning system (TPS) for tumor target delineation, voxel-level heterogeneity feature extraction, habitat cluster analysis, and subsequent radiotherapy dose optimization.

[0029] (5) Target delineation Radiation oncologists outline the GTV and normal organs under the lung window.

[0030] S2: Within the delineated tumor target area, a three-dimensional neighborhood window is constructed with each voxel as the center, and the local image features of each voxel are extracted. The local image features are local entropy.

[0031] To systematically characterize the spatial heterogeneity of the internal microstructure and grayscale distribution of lung cancer tumors, a voxel-level local feature modeling strategy is adopted to describe the internal heterogeneity of the tumor at high resolution, based on the completion of tumor target region segmentation.

[0032] Specifically, each voxel within the tumor region Centered on, construct a scale of A three-dimensional cubic neighborhood window is used to extract multidimensional local image features from two levels: gray-level statistical characteristics and texture spatial relationships. These features reflect the complexity of tissue structures, the degree of texture variation, and differences in spatial correlation at different spatial locations. This local modeling method can effectively capture the potential microenvironmental differences within tumors without relying on manually defined subregion boundaries. This invention selects local image features with good interpretability and stability, such as local entropy, to construct a voxel-level heterogeneity representation. Local entropy: Used to measure the degree of disorder in the gray-level distribution within a local neighborhood, reflecting the complexity and irregularity of the organizational structure.

[0033] x i : Represents the current central voxel. Entropy(x i ) voxels x i Local entropy; Ω i : indicates voxels x i The neighborhood range centered on the image. Given that medical imaging data is typically three-dimensional, this study uses... x i The computational region is defined by a 3×3×3 voxel range, which includes 26 adjacent voxels including the central voxel. p(x j ) : indicates neighborhood Ω i Endogenous x jThe probability of a gray value (or intensity value) appearing is usually obtained by calculating the histogram distribution of each gray value in the neighborhood; ∈: is a very small positive integer used to balance the logarithmic function log to prevent the logarithm from being meaningless when p(xj) is 0.

[0034] To ensure the numerical stability of the local entropy calculation, a small constant ∈ is added to the probability term, where ∈ is preferably taken as 10. 10 .

[0035] Using the above formula, the local entropy value of each voxel within the ROI range is calculated one by one.

[0036] S3: Construct voxel-level feature vectors based on the local entropy value and CT gray value of each voxel, and use the K-means clustering algorithm to cluster all voxels in the tumor target area, dividing the solid tumor into several primary habitat subregions with different imaging characteristics.

[0037] S31: Feature Normalization Processing To eliminate differences in the dimensions and numerical ranges of different features and to avoid a single feature having a dominant influence on the clustering results, this study uses Z-score standardization to normalize the features of each dimension:

[0038] in and They represent the first The mean and standard deviation of each feature within the tumor region. This step ensures that each feature contributes equally to the Euclidean distance calculation during clustering. It is the local entropy value of a certain voxel.

[0039] S32: Clustering Strategy and Parameter Optimization Local entropy of each voxel Entropy(x i ) and the grayscale value of its corresponding mode. I(x i ) Based on this, construct voxel-level feature vectors:

[0040] For each sample point, calculate its distance to all cluster centers, and use K-means to cluster all voxels. The clustering objective is to minimize the within-cluster squared error (SSE).

[0041] in, Ck This is the k-th type of habitat area; μ k The cluster center is set as the value of the cluster. After K-means iteration reaches stability, the tumor is divided into k habitat subregions. In the specific implementation: a K-means initialization strategy is used to improve the quality of centroid initialization and accelerate convergence; voxel classification and centroid iterative update are completed based on Euclidean distance; clustering is terminated when the centroid change is less than the threshold or the maximum number of iterations is reached.

[0042] To avoid subjectivity in the selection of the number of clusters, in A systematic search was conducted within the scope, and the Calinski-Harabasz (CH) index was introduced to evaluate the stability and separation of different clustering structures:

[0043] CH: CH index, used to evaluate the quality of clustering results. The higher the CH value, the higher the separation between clusters and the better the compactness within clusters, that is, the better the clustering effect; tr(B k ): Inter-class discrete matrix B k The trace represents the overall dispersion among different cluster centers and is used to measure inter-cluster separability. tr(W k ): Intraclass discrete matrix W k The trace represents the degree of dispersion among samples within the same cluster, and is used to measure intra-cluster tightness. K : Number of cluster categories, i.e., the number of cluster centers in K-means clustering; N Total number of samples; in this study, this usually represents the total number of elements within the tumor ROI that participated in cluster analysis. K 1 Inter-class degrees of freedom; N K Intraclass degrees of freedom.

[0044] S33: Nested Tumor Subregion Reclustering In this invention, K-means clustering is first used to divide solid lung cancer tumors into multiple primary habitat subregions A, B, C, etc., with different imaging characteristics. To further explore the finer spatial heterogeneity within each subregion and calculate the ITHscore of primary habitat subregions A, B, and C, this invention further employs a nested K-means clustering method to further subdivide the primary habitat regions. Specifically, after completing the primary clustering, the voxel sets and their imaging features within subregions A, B, and C are extracted to form independent voxel-local entropy matrices. Subsequently, using the voxel features within a subregion (e.g., region A) as new input data, the K-means clustering algorithm is executed again, and the number of clusters is reset. K The primary habitat subregions A, B, and C are further subdivided into A1, A2, A3, B1, B2, B3, C1, C2, C3, etc.

[0045] S4: Based on the spatial connectivity characteristics of each primary habitat subregion after division, construct and calculate the tumor heterogeneity index ITHscore.

[0046] To quantify tumor heterogeneity holistically at the spatial organizational structure level, a unified, stable, and low-dimensional tumor heterogeneity index is constructed.

[0047] Based on the above spatial connectivity characteristics, the tumor heterogeneity index (ITHscore) is constructed as follows:

[0048] in: S total The total number of all voxels in the tumor volume (i.e., the total volume). V : Represents the number of subregions into which the tumor is divided; i : Represents the first i Subregions; S i,max : No. i The largest number of a certain type of voxel in a habitat subregion; n i : No. i The total number of primes in each habitat.

[0049] The range of ITHscore values ​​is: The higher the value, the more dispersed the spatial distribution and the more complex the connectivity structure of the subregions A, B, and C within the tumor, indicating a higher level of intratumoral heterogeneity.

[0050] The above algorithm can be used to obtain the ITH score of different subregions A, B, and C of solid tumors.

[0051] S5: Based on the calculated ITHscore values ​​of different subregions within the tumor, administer individualized radiotherapy doses to each subregion.

[0052] S51: Determine the tumor ITHscore value corresponding to the standard radiotherapy dose. Three hundred lung cancer CT images were selected from the TCIA public database and two domestic cancer centers. After dividing the tumor into different subregions A, B, and C according to the algorithm described above, the ITHscore for each case was calculated. The median ITHscore was then calculated. 总 This corresponds to the standard radiotherapy dose (60Gy, as the international standard for radiotherapy dose of 60Gy is currently the same for locally advanced inoperable NSCLC).

[0053] S52: Determining the radiotherapy dose for different subregions of the tumor in lung cancer patients For a specific tumor patient, based on the ITHscore values ​​calculated for tumor subregions A, B, and C, and using 60 Gy and its corresponding ITHscore value as a baseline, different radiotherapy doses are administered to subregions A, B, and C in a linear proportion, followed by radiotherapy planning. For example, if the ITHscore value calculated for subregion A is... A Then the radiation therapy dose D given to subregion A A for:

[0054] The radiation dose for other subregions within the tumor is calculated in the same manner.

[0055] S6: Radiotherapy planning.

[0056] S61: Calculate the radiotherapy dose for different subregions of the tumor Assuming the median ITHscore value corresponding to a standard radiation therapy dose of 60 Gy (ITHscore) 总 If a patient's solid tumor is divided into three subregions (A, B, and C) with ITH scores of 0.540, 0.480, and 0.439 respectively, then the prescribed radiotherapy doses for these three subregions, calculated using this method, should be 64.8 Gy, 57.6 Gy, and 52.7 Gy, respectively. Two independent radiotherapy plans are designed: Plan 1: A uniform prescribed dose of 60 Gy is used for the entire solid tumor; Plan 2: The solid tumor is divided into three subregions (A, B, and C), with prescribed doses of 64.8 Gy, 57.6 Gy, and 52.7 Gy, respectively. The target area and organ-at-risk doses of the two plans are then compared after designing the radiotherapy plans according to the following dose constraints.

[0057] S62: Dose Constraints in Radiotherapy Planning Whole lung V5 ≤ 60%, V 10 ≤40%, V 20 ≤30%, average dose to the whole lung ≤15Gy; highest dose to the spinal cord ≤45Gy; highest dose to the esophagus ≤64Gy; cardiac V 30 ≤40%, V 40 ≤30%, average cardiac dose ≤26Gy; highest dose to target area <110% of prescription dose.

[0058] S63: Optimization of Radiotherapy Planning When designing radiotherapy plans, four dose-limiting rings are generated outside the target area to improve conformity in low-dose regions. For Plan 2, due to the existence of three dose gradients, a dose gradient descent gap with a width of 5-8 mm must be established at the boundary of different subregions during plan optimization to prevent excessive high-dose regions from forming due to conflicting conditions during optimization of different prescription doses. Both radiotherapy plans are optimized using the same gantry angle and the same dose optimization conditions. After dose calculation, normalization is performed by using 100% of the prescription dose to surround 95% of the target area.

[0059] According to experience in radiotherapy planning, the dose drop gradient generated per 1 mm is approximately 1.5 Gy. Therefore, the gap that needs to be reserved between the two doses of 64.8 Gy and 57.6 Gy is: (64.8-57.6) / 1.5=4.80≈5 mm, and the gap that needs to be reserved between the two doses of 64.8 Gy and 52.7 Gy is: (64.8-52.7) / 1.5=8.06≈8 mm.

[0060] The results of the radiotherapy planning design of this invention are as follows: Figure 3 The diagram shows cross-sectional dose distribution for two different plans. Figure 3 The yellow (thick) dose line shown in A is the 60 Gy prescription dose line; Figure 3 In Figure B, the yellow, blue, and sky-blue dose lines (in bold) represent prescription dose lines of 60 Gy, 57.6 Gy, and 52.7 Gy, respectively. Figure 4 This presentation shows a comparison of dose parameters for two different radiotherapy plans. PTV64.8, due to its higher tumor heterogeneity, received a higher tumor radiation dose than the original dose, while PTV57.6 and PTV52.7, due to their lower tumor heterogeneity, received a lower tumor radiation dose than the original dose. The dose parameter results show that the radiotherapy plan obtained using this technique reduced the maximum dose to the aorta, bronchial tree, esophagus, trachea, and spinal cord by 0.95-3.28 Gy; the maximum cardiac dose was reduced by 7.78 Gy, and the cardiac V... 30 V 40Both the average and total doses showed varying degrees of decrease; V5 and V5 in the whole lungs... 10 V 20 The average dose also decreased to varying degrees. These reductions in dose parameters can effectively reduce radiation-related toxicities.

[0061] The above description is merely a preferred embodiment of the present invention and should not be construed as limiting the scope of the invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for optimizing radiotherapy dose for lung cancer based on intratumoral heterogeneity, characterized in that, Includes the following steps: S1: Obtain simulated localization CT images of lung cancer patients, and delineate the target area of ​​solid tumors based on the CT images; S2: Within the delineated tumor target area, a three-dimensional neighborhood window is constructed with each voxel as the center, and the local image features of each voxel are extracted. The local image features are local entropy. S3: Construct voxel-level feature vectors based on the local entropy value and CT gray value of each voxel, and use the K-means clustering algorithm to cluster all voxels in the tumor target area, dividing the solid tumor into several primary habitat subregions with different imaging features. S4: Based on the spatial connectivity characteristics of each primary habitat subregion after division, construct and calculate the tumor heterogeneity index ITHscore; S5: Based on the calculated tumor heterogeneity index (ITHscore) values ​​of different subregions within the tumor, individualized radiotherapy doses are given to each subregion.

2. The method according to claim 1, characterized in that, The extraction of local entropy in S2 specifically involves: using each voxel... Centered on a 5×5×5 three-dimensional cubic neighborhood window, the local entropy of each voxel is calculated using the following formula: in, Entropy(x i ) voxels x i Local entropy; Ω i Indicated by voxels x i The 3×3×3 neighborhood of the center; p (x j ) Representing the neighborhood Ω i Endogenous x j The probability of grayscale values ​​appearing; ∈ is a very small positive integer used to prevent the logarithm from being meaningless.

3. The method according to claim 1, characterized in that, In step S3, before performing K-means clustering, a step of Z-score standardization of the local entropy features is also included: in, and They represent the first The mean and standard deviation of a local entropy feature within the tumor region.

4. The method according to claim 1, characterized in that, In S3, when the K-means clustering algorithm is used to cluster all voxels, the number of cluster categories is... K The selection range is from 3 to 9, and the stability and separation of different clustering structures are evaluated using the CH index, which is calculated using the following formula: Among them, tr(B) k ): The trace of the inter-cluster discrete matrix Bk, representing the overall degree of dispersion among different cluster centers, used to measure inter-cluster separability; tr(W k ): Intraclass discrete matrix W k The trace represents the degree of dispersion among samples within the same cluster, and is used to measure intra-cluster tightness. N Total number of samples; in this study, this usually represents the total number of elements within the tumor ROI that participated in cluster analysis. K 1 Inter-class degrees of freedom; N K Intraclass degrees of freedom.

5. The method according to claim 1, characterized in that, In step S3, after dividing solid tumors into primary habitat subregions, a nested clustering step is further included: extracting the voxel set and its local entropy features corresponding to each primary habitat subregion to form an independent voxel-local entropy value matrix, and then executing the K-means clustering algorithm again to further subdivide each primary habitat subregion into multiple secondary habitat subregions.

6. The method according to claim 1, characterized in that, The tumor heterogeneity index ITHscore is calculated according to the following formula: in, S total The total number of all voxels in the tumor volume; V The number of subregions into which the tumor is divided; i Representing the i Subregions; S i,max For the first i The largest number of a certain type of voxel in a habitat subregion; n i For the first i The total number of primes in each habitat; the range of ITHscore values ​​is... [0,1] A higher value indicates greater intratumoral heterogeneity.

7. The method according to claim 1, characterized in that, S5 specifically includes: S51: Obtain the standard radiotherapy dose of 60 Gy and its corresponding median tumor heterogeneity index (ITHscore). 总 ; S52: For a specific subregion A within a patient's tumor, the calculated tumor heterogeneity index is the ITHscore. A Then, an individualized radiotherapy dose D is given to subregion A. A Calculate using the following formula: 。 8. The method according to claim 1, characterized in that, Following S5, the radiotherapy planning step is also included: designing a radiotherapy plan according to the following dose constraints: Whole lung V5 ≤ 60%, V 10 ≤40%, V 20 ≤30%, average dose to the whole lung ≤15Gy; highest dose to the spinal cord ≤45Gy; highest dose to the esophagus ≤64Gy; cardiac V 30 ≤40%, V 40 ≤30%, average cardiac dose ≤26Gy; highest dose to target area <110% of prescription dose.

9. The method according to claim 8, characterized in that, When optimizing a radiotherapy plan, if there are multiple subregions with different prescription doses, a dose gradient descent gap with a width of 5-8 mm should be established at the boundary between different subregions to prevent the formation of too many high-dose regions due to conflicting conditions during the optimization of different prescription doses.

Images