A virtual lesion-based radiotherapy intensity modulation optimization method

Abstract

This invention provides a method for intensity-modulated radiotherapy (IMRT) optimization based on virtual lesions, comprising the following steps: S1, inferring the ideal dose distribution: obtaining the ideal dose distribution through dose prediction methods; S2, generating the initial dose distribution: setting the target area field parameters and obtaining the initial dose distribution through dose calculation; S3, generating the dose to be removed: differencing the ideal dose distribution with the initial dose distribution to obtain the dose to be removed distribution; S4, adding virtual lesions: determining the location and size of virtual lesions using a concave-priority strategy; S5, determining the virtual lesion dose parameters: setting the virtual lesion field parameters; S6, outputting the results: obtaining the final IMRT optimization result and outputting the DVH diagram and other dose indicators. This invention uses virtual lesions to achieve targeted dose reduction, thereby stabilizing the optimization of concave target areas, and has advantages such as interpretability, low computational cost, and DVH key nodes closely aligned with the prescription.

Description

Technical Field

[0001] This invention relates to the field of intensity-modulated radiotherapy (IMRT), and more specifically to an IRT optimization method based on virtual lesions. Background Technology

[0002] Radiotherapy is a primary treatment for cancer. During treatment, the radiation energy needs to be completely concentrated within the tumor tissue, while surrounding tissues remain untouched. However, due to the physical properties of radiation scattering, this goal cannot be fully achieved. Therefore, the practical requirement of clinical radiotherapy is to concentrate the dose as much as possible in the target area and strictly control the radiation dose to surrounding normal tissues, especially vital organs. To further optimize radiotherapy outcomes, many optimization methods have been developed in radiotherapy planning, with intensity-modulated radiation (IMRT) being one of the classic and highly effective methods.

[0003] Intensity-modulated radiotherapy (IMRT) divides each radiation field into numerous subfields, optimizing the fluence distribution within the field to achieve non-uniform irradiation. This results in conformal targeting in three-dimensional space, effectively concentrating the dose within the target area and limiting the radiation dose to surrounding normal tissues. The main steps of IMRT are first optimizing the fluence intensity of the radiation field, then optimizing the subfield segmentation to obtain an executable multi-page raster sequence. The first step, optimizing the fluence map intensity, is crucial for the entire radiotherapy plan. The mainstream optimization method involves dividing the radiation field into numerous subfields, defining an objective function in nonlinear optimization, iterating to obtain the optimal fluence intensity value for each subfield, and then summing the doses to obtain the final optimized result. Such methods often require extremely high computing power to process a large number of subfield variables. Although the optimization results can meet clinical standards, the computational and time costs are high, making it impractical for less developed regions. Summary of the Invention

[0004] To address the aforementioned issues, this invention aims to propose a method for intensity-modulated radiotherapy (IMRT) optimization based on virtual lesions. By adding virtual lesions, quantitative reduction of dose distribution is achieved, thereby optimizing the IMRT and obtaining results such as fluence map intensity distribution, dose distribution, and DVH.

[0005] The technical solution adopted by the present invention to achieve the above objectives is as follows:

[0006] A method for intensity-modulated radiotherapy (IMRT) optimization based on virtual lesions includes the following steps:

[0007] S1. Reasoning the ideal dose distribution: The ideal dose distribution is obtained by reasoning through dose prediction methods.

[0008] S2. Generate initial dose distribution: Set the target area's field direction and field weight, and obtain the initial dose distribution through dose calculation.

[0009] S3. Generate the dose to be removed: Difference the ideal dose distribution with the initial dose distribution to obtain the dose distribution to be removed.

[0010] S4. Add virtual lesions: Determine the location of virtual lesions using a depression-first strategy, and determine the size of lesion expansion based on the dose decrease gradient around the location in the dose distribution to be removed.

[0011] S5. Determine the virtual lesion dose parameters: Set the field direction and field weight of the virtual lesion.

[0012] S6. Output Results: Obtain the final intensity-modulated radiation (IMRT) optimization results, and output the injection intensity distribution, dose distribution, DVH diagram, and other dose indicators for each radiation field direction.

[0013] Furthermore, in step S1, the ideal dose distribution D ideal It was derived from a dose distribution prediction model based on deep learning.

[0014] Furthermore, in step S2, the field direction and field weight are set by the physicist based on experience, and a conformal initial dose distribution D is obtained through a pencil beam dose calculation algorithm. conformal .

[0015] Furthermore, in step S3, the target dose D to be removed is... sub It is obtained by the difference between the conformal initial dose distribution and the ideal dose distribution, and the formula is expressed as:

[0016] D sub =D conformal -D ideal

[0017] Furthermore, in step S4, based on the dose distribution to be removed, the set of starting positions of virtual lesions is determined according to the depression priority strategy. Each virtual lesion is then grown in three dimensions according to the dose descent gradient at its location, and its boundary convexity is constrained to finally obtain a set of virtual lesions.

[0018] Specifically, it can be divided into the following steps:

[0019] S4.1: Locate the depression region around the target area using a depression search algorithm;

[0020] S4.2: Sort the depressions by volume and add virtual lesions starting with the largest depression.

[0021] S4.3: Next, find the local highest dose point as the initial growth point of VTV and start expanding outward. Determine the growth boundary according to the isodose gradient of the growth direction.

[0022] If all the depressed areas have been processed, proceed to step S4.4; if not all the depressed areas have been processed, proceed to step S4.3 to continue adding virtual lesions in the depressed areas.

[0023] S4.4: Record the maximum and average dose values ​​of the dose to be removed within all virtual lesions, providing a reference for the field parameters of adding virtual lesions, and completing the addition of virtual lesions.

[0024] Furthermore, in step S5, the target dose of each virtual lesion is determined based on the maximum and average dose of the dose to be removed within each virtual lesion. A suitable irradiation direction is selected from all irradiation directions in the target area and the target dose is evenly distributed, thereby obtaining the irradiation direction and irradiation weight of each virtual lesion.

[0025] Furthermore, in step S6, after obtaining the dose distribution of all virtual lesions through dose calculation, the final intensity-modulated radiation optimization result can be obtained by performing a difference operation between the conformal initial dose distribution and the dose distribution of all virtual lesions, including the dose distribution map, the injection intensity distribution of each field, the DVH map, and other common dose parameters. Attached Figure Description

[0026] The present invention will now be further described with reference to the accompanying drawings, wherein:

[0027] Figure 1 This is a schematic diagram of the process for intensity-modulated radiotherapy optimization based on virtual lesions.

[0028] Figure 2 A schematic diagram illustrating the specific process of adding virtual lesions. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] like Figure 1 and Figure 2 As shown, this invention discloses a method for intensity-modulated radiotherapy (IMRT) optimization based on virtual lesions. By adding virtual lesions, quantitative dose distribution reduction is achieved, thus optimizing the IMRT and obtaining results such as flux map intensity distribution, dose distribution, and dose-dependent volume-hose (DVH). The method includes the following steps:

[0031] S1. Inferring the Ideal Dose Distribution: The ideal dose distribution D is obtained by inferring using a deep learning-based dose distribution prediction model. ideal .

[0032] S2. Generating the initial dose distribution: Setting the target area's field direction and field weight. The field direction and field weight are set by the physicist based on experience. A conformal initial dose distribution D is obtained through a pencil beam dose calculation algorithm. conformal .

[0033] S3. Generate the dose to be removed: the target dose to be removed, D. sub It is obtained by the difference between the conformal initial dose distribution and the ideal dose distribution, and the formula is expressed as:

[0034] D sub =D conformal -D ideal

[0035] S4. Adding Virtual Lesions: Based on the dose distribution to be removed, the starting location set of virtual lesions is determined according to the depression priority strategy. Each virtual lesion is then subjected to three-dimensional outward growth based on the dose descent gradient at its location, and boundary convexity constraints are applied to it, ultimately obtaining a set of virtual lesions. Specifically, this can be divided into the following steps:

[0036] S4.1: Locate the depression region around the target area using a depression search algorithm;

[0037] S4.2: Sort the depressions by volume and add virtual lesions starting with the largest depression.

[0038] S4.3: Next, find the local highest dose point as the initial growth point of VTV and start expanding outward. Determine the growth boundary according to the isodose gradient of the growth direction.

[0039] If all the depressed areas have been processed, proceed to step S4.4; if not all the depressed areas have been processed, proceed to step S4.3 to continue adding virtual lesions in the depressed areas.

[0040] S4.4: Record the maximum and average dose values ​​of the dose to be removed within all virtual lesions, providing a reference for the field parameters of adding virtual lesions, and completing the addition of virtual lesions.

[0041] S5. Determine the virtual lesion dose parameters: Set the field direction and field weight of the virtual lesion. Determine the target dose of each virtual lesion based on the maximum and average dose of the dose to be removed in each virtual lesion. Select an appropriate irradiation direction from all field directions in the target area and distribute the target dose equally to obtain the field direction and field weight of each virtual lesion.

[0042] S6. Output Results: After obtaining the dose distribution of all virtual lesions through dose calculation, the final intensity-modulated radiation (IMR) optimization results can be obtained by performing a difference operation between the initial conformal dose distribution and the dose distribution of all virtual lesions. These results include the dose distribution map, the injection intensity distribution of each field, the DVH map, and other common dose parameters.

[0043] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for intensity-modulated radiotherapy (IMRT) optimization based on virtual lesions, characterized in that, Includes the following steps: S1. Reasoning for the ideal dose distribution: The ideal dose distribution is obtained by reasoning through dose prediction methods; S2. Generate initial dose distribution: Set the target area's field direction and field weight, and obtain the initial dose distribution through dose calculation; S3. Generate the dose to be removed: Difference the ideal dose distribution with the initial dose distribution to obtain the dose distribution to be removed; S4. Add virtual lesions: Determine the location of virtual lesions using a depression-first strategy, and determine the size of lesion expansion based on the dose decrease gradient around the location in the dose distribution to be removed. S5. Determine the virtual lesion dose parameters: Set the field direction and field weight of the virtual lesion; S6. Output Results: Obtain the final intensity-modulated radiation (IMRT) optimization results, and output the injection intensity distribution, dose distribution, DVH diagram, and other dose indicators for each radiation field direction.

2. The method for intensity-modulated radiotherapy based on virtual lesions according to claim 1, characterized in that: The dose prediction method in step S1 refers to a dose distribution prediction model based on deep learning. It is a prediction model trained using a large amount of clinical planning data from a deep learning network, which can infer the ideal dose distribution of the target area.

3. The method for intensity-modulated radiotherapy based on virtual lesions according to claim 1, characterized in that: Adding a virtual lesion in step S4 can be divided into the following steps: S4.1: Locate the depression region around the target area using a depression search algorithm; S4.2: Sort the depressions by volume and add virtual lesions starting with the largest depression. S4.3: Next, find the local highest dose point as the initial growth point of VTV and start expanding outward. Determine the growth boundary according to the isodose gradient of the growth direction. If all the depressed areas have been processed, proceed to step S4.4; if not all the depressed areas have been processed, proceed to step S4.3 to continue adding virtual lesions in the depressed areas. S4.4: Record the maximum and average dose values ​​of the dose to be removed within all virtual lesions, providing a reference for the field parameters of adding virtual lesions, and completing the addition of virtual lesions.

4. The method for adding virtual lesions according to claim 3, characterized in that: In step S4.3, the termination condition related to the three-dimensional outward expansion growth of the virtual lesion is that, along any growth direction, when the decrease ratio of the dose to be removed in that direction reaches a preset ratio threshold, or when the isodose line descent gradient in that direction is lower than a preset gradient threshold, the growth in that direction is terminated, and convexity constraints are applied to the synthesized growth boundary to finally determine the shape and size of the lesion.

5. The method for adding virtual lesions according to claim 3, characterized in that: In step S4.4, the target dose of each virtual lesion is determined by a fixed weighted sum of the maximum and average values ​​of the dose to be removed within it, and the target dose is evenly distributed among the selected irradiation directions to obtain the field weight of each irradiation direction.

Images