Contour delineation for radiation therapy planning with real-time contour segment impact rendering

[0022] reference figure 1 The radiation therapy system includes an imaging modality 10, which is adapted to acquire planning images for planning radiation therapy. In some embodiments, the imaging modality 10 is a computer tomography (CT) scanner, such as Brilliance TM Big Bore TM CT scanner (available from Royal Philips Electronics Co., Ltd., Eindhoven, The Netherlands). The Brilliance TM Big Bore TM The CT scanner is an illustrative CT scanner with a large patient aperture of 85 cm, which is large enough to accommodate a patient arranged in a typical radiation therapy position. Other CT scanners, as well as other imaging modalities, such as positron emission tomography (PET), magnetic resonance (MR), single photon emission computed tomography (SPECT), etc., may be used instead. The captured image is stored in the image memory 12. Preferably, planning images are acquired for the subject planned to receive radiation therapy. For example, the subject may be a tumor patient, or a subject of veterinary medicine. The term "planning image" in this context refers to a planning image (or a collection of planning images) used for contouring.

[0023] Before the radiation treatment procedure, the radiation treatment planning module 20 executes the radiation treatment planning. The planning module 20 includes a contour delineation sub-module 22 and a display sub-module 24, which cooperate to perform contour delineation to define and outline the radiation target area and one or more risk areas in the planning image. For example, the radiation target area may be a cancer tumor, and the one or more risk areas may include adjacent vital organs or tissues, and the radiation exposure amount of the adjacent vital organs or tissues should be kept below a maximum value. To this end, radiation therapy planning is about dose planning parameters 26, which indicate doses or dose constraints for the radiation target area and one or more risk areas.

[0024] The intensity modulation optimization sub-module 30 receives at least the contours generated by the contouring components 22 and 24 as input, and generates a radiation treatment plan and a radiation dose map 32 that complies with the calculation of the dose planning parameters 26. The intensity modulation optimization sub-module 30 optionally receives other relevant inputs, such as an attenuation map indicative of radiation absorption (eg, appropriately calculated based on planning images and/or anatomical models). Intensity modulation optimization is typically an iterative "reverse" process, in which: (1) initialize radiation therapy parameters, such as beam intensity, collimator settings, etc., then (2) use these parameters to calculate the calculated dose map, and then ( 3) Use the dose planning parameter 26 to evaluate the compliance of the calculated dose map, and then (4) update the radiation treatment parameters in a way that is expected to make the calculated dose map more closely match the dose planning parameter 26. Then operations (2)-(4) are repeated to iteratively make the calculated dose map best meet the dose planning parameters 26.

[0025] During the contour drawing performed by the contour drawing components 22, 24, the display sub-module 24 draws a contour with a code (eg, color coding and/or line thickness coding) indicating the influence of the contour on the intensity modulation optimization. To this end, the differential analysis sub-module 40 is configured to call the intensity modulation optimization sub-module 30 to estimate the partial derivative of the output of the intensity modulation optimization with respect to the contour. In some embodiments, the output of intensity modulation optimization is the dose distribution, each partial derivative quantifies how the dose accumulation changes when the contour point or contour segment changes, and the partial derivative quantifies the change in the dose accumulation at the contour segment, or The derivative with respect to the dose distribution at the contour position. The contour drawing sub-module 22 is configured to call the differential analysis sub-module 40 after each iteration of contour adjustment to estimate the partial derivative with respect to the contour segment, and is configured to draw the contour segment on the display of the planning image, the contour segment being based on all the contour segments. The estimated partial derivative is coded to indicate the influence of the contour segment on the intensity adjustment optimization.

[0026] In one approach, the radiation therapy planning module 20 is implemented on a suitable computer 44, which has a display device 46 (eg, LCD screen, cathode ray tube display device, etc.) and one or more user input devices, such as illustrated Keyboard 48, or a mouse, trackball, or other pointing user input device, etc. The display sub-module 24 displays a planning image with a contour segment on the display device 46, and the contour segment is encoded based on the estimated partial derivative to indicate the influence of the contour segment on the optimization of intensity adjustment. The user manually inputs the adjustment to the contour via the user input device 48. Then the contour rendering sub-module 22 calls the differential analysis sub-module 40 to estimate the partial derivative with respect to the contour (including the adjusted contour), and draws (updated) contour fragments on the display device 46, which are superimposed on the planning image And is coded to indicate the (updated) influence of the contour segment on the intensity adjustment optimization.

[0027] The final radiation treatment plan generated by the radiation treatment planning module 20 is stored in the radiation treatment plan memory 50. At the planned date and time for performing the radiation treatment procedure, the radiation treatment device 52 is used to control the therapeutic radiation delivered to the subject through the radiation treatment control system 54 according to the radiation treatment plan stored in the memory 50. For example, in the illustrated embodiment, the radiation therapy delivery device 52 is a tomographic linear accelerator (linear accelerator), and the radiation therapy control system 54 operates a multi-leaf collimator (MLC) or other radiation beam appearance shaping of the linear accelerator 52 The device modulates the beam intensity and appearance when the linear accelerator rotates around the object, thereby delivering to the object a radiation dose distribution that provides the target feature with the desired integrated radiation dose according to the radiation treatment plan that complies with the dose planning parameter 26, while appropriately Limit or restrict the radiation exposure of sensitive important features.

[0028] The radiation therapy planning module 20 may be variously embodied by a single digital processor, two or more digital processors, a computer, an application specific integrated circuit (ASIC), etc. For example, the illustrated computer 44 may include the planning module 20. Likewise, a radiation treatment plan including contour delineation with contour coding to influence optimization of intensity adjustment may be embodied by a storage medium storing instructions that can be executed by a digital processor to execute the plan including contour delineation with contour coding. For example, the storage medium may be a hard disk or other magnetic storage media, optical disks or other optical storage media, random access memory (RAM), read only memory (ROM), flash memory, or other electronic storage media, various combinations thereof, and so on.

[0029] reference figure 2 , Describing suitable for figure 1 An illustrative radiation therapy planning method performed by the system, the radiation therapy planning method including contour delineation with contour coding for influencing intensity modulation optimization. In operation 60, the planning image is displayed and the initial outline is generated and displayed. Operation 60 may adopt manual initial contour drawing, automatic initial contour drawing using automatic segmentation of planning images, or a semi-automatic method. The intensity modulation optimization sub-module 30 then performs an intensity modulation optimization operation 62 to generate an output 64 that includes optimized radiation therapy parameters (eg, beam intensity, multi-leaf collimator settings, etc.), a calculated radiation dose map 32 (see figure 1 ), and optional additional parameters such as related total dose parameters. For example, the total dose parameter may include the cumulative radiation dose delivered to the radiation target area and/or one or more risk areas, and each such cumulative radiation dose is calculated by integrating over the area in the calculated radiation dose map 32 . The output 64 may also optionally include one or more comprehensive parameters, such as a measure indicating the maximum deviation of any calculated area dose (radiation target area or risk area) from the corresponding dose planning parameter in an illustrative example.

[0030] Next is the influence of the calculated contour in operations 66, 68, 70 on the optimization of intensity adjustment. To perform the differential analysis, the contour is divided into contour fragments, and the partial derivative of the output of the intensity modulation optimization with respect to the contour fragment is calculated. Therefore, the contour segment is selected for difference analysis in operation 66. In operation 68, the partial derivative of the output of the intensity modulation optimization with respect to the selected contour segment is estimated. Operations 66, 68 iterate 69 over all contour fragments to estimate the partial derivative with respect to the contour fragment. Operation 70 classifies each contour segment with respect to its influence on the intensity modulation optimization based on the estimated partial derivative.

[0031] In operation 74, the display sub-module 24 displays the planning image, optionally displays the selected total dose parameter, and displays the contour fragments that influence the rendering, that is, the contour fragments that indicate the influence of the contour on the output of the intensity modulation optimization Encoded contour fragment. This influence is appropriately quantified by the contour segment classification output of operation 70. In operation 76, the contour segment is updated. The contour segment update operation 76 may be manual, for example, the user receives the contour segment adjustment input via the user input device 48. Optionally, the contour segment update operation 76 may be automatic, for example, performed based on the re-segmentation of the radiation target area and/or the risk area in the planned image. The processing flow then returns to operation 62 to update the calculated radiation dose map 32 and other output 64 of the intensity modulation optimization to take into account the contour segment adjustment, and to update the impact analysis by repeating operations 66, 68, 70, and update the display according to operation 74. In this way, the user observes the influence of the contour drawing operation substantially in real time, that is, during the contour drawing process.

[0032] reference Figure 3-Figure 6 , Which schematically illustrates the adoption of figure 2 The method of operation 66, 68, 70 performs an impact analysis. image 3 The area R delineated by the outline C is shown. Operation 62 calculates the intensity modulation optimized output 64 for this profile C. For illustrative purposes, the output 64 (or its components used for impact analysis) is referred to herein as output Y without loss of generality. Figure 4 The area R and the contour C are shown, but with the adjusted contour, the first selected contour segment CS1 is adjusted outward by the difference Δ CS1. (Alternatively, the difference contour segment adjustment may be an inward adjustment). Operation 68 generates an adjusted output Y by applying intensity modulation optimization to the adjusted contour 1 And the partial derivative about the first contour segment CS1 is calculated as (Y 1 -Y)/Δ CS1 To properly estimate the partial derivative with respect to the first contour segment CS1.

[0033] Figure 5 The area R and the contour C are shown, but with the contour adjusted as follows: the second selected contour segment CS2 is adjusted outward by the difference Δ CS2. (Also, alternatively, the difference contour segment adjustment may be an inward adjustment). This corresponds to returning the flow to the second iteration 69 of the selection operation 66, which now selects the second contour segment CS2. The second iteration of operation 68 generates an adjusted output Y by applying intensity modulation optimization to the adjusted contour 2 And the partial derivative of the second contour segment CS2 is calculated as (Y 2 -Y)/Δ CS2 To estimate the partial derivative of the second contour segment CS2. (Note that the difference used in the contour segment is the same, that is, Δ CS1 =Δ CS2 , And this is the case for all contour fragments, then optionally omit dividing by Δ CS1 Or Δ CS2 Scaling).

[0034] This process is repeated for each contour segment of the contour C through the operation of the iterative loop 69 to calculate the partial derivative with respect to the contour segments constituting the contour C. Generally, the output Y can be any output optimized for intensity modulation, as long as you are interested in the effect of the output on the contour. In one example, the output Y may be the dose in the radiation target area. In another example, the output Y can be a comprehensive parameter, such as the maximum ratio of the dose in the risk area to the dose constraint of that risk area (over all risk areas). In operation 70, the contour segments are classified based on partial derivatives. For example, in one method, a binary classifier is used that has a contour segment that is classified as "high impact" if the partial derivative with respect to that contour segment is greater than a threshold value, and when the partial derivative with respect to that contour segment is low In the case of a threshold value, it is classified as a "low impact" contour segment. More generally, the classification can be more than two, three, and so on. In some embodiments, the classification operation 70 is replaced by a continuous scoring operation that assigns a score to each contour segment.

[0035] Such as Image 6 As shown schematically, each contour segment is displayed with an appropriate code indicating the influence of the contour segment calculated by the classification (or scoring) operation 70. Three impact categories are used in the illustrative embodiment. The solid contour lines are used to encode high-impact segmentation. The dash is used for segmentation coding with medium impact. Dotted lines are used for low-impact segmentation coding. Image 6 Illustrated for reference Figure 3-Figure 5 To describe the illustrative region R of its impact analysis and such contour coding for two other regions R1, R2. Although in Image 6 Solid lines/dashed lines/dotted lines are used in, but in other embodiments where the display device 46 can generate a color display, the coding may be color coding, for example, the color "red" is used to code high-impact contour fragments , The color "yellow" to encode medium-impact contour fragments, and the color "green" to encode low-impact contour fragments.

[0036] The coding of the contour segment indicating the influence of the contour segment on the intensity modulation optimization advantageously informs the radiologist or other medical staff performing the contouring which contour segment has the most influence on the intensity modulation optimization. reference Image 6 For example, radiologists know that contour segments coded by dashes have a low impact on the intensity modulation optimization, and therefore the contour description of those virtual contour segments can be completed in a relatively rough manner. On the other hand, radiologists know that contour segments encoded by solid lines have a high impact on the intensity modulation optimization, and therefore should be highly careful and accurate to complete the contour delineation of those real contour segments. By telling radiologists which contour fragments have high influence and which have low influence, the contouring process is accelerated (by allowing radiologists to locate low-impact contour fragments more roughly) and the contouring process is also made more accurate (due to radiologists Ensure the accuracy of high-impact contour fragments).

[0037] Continue to refer figure 1 with 2 And for further reference Figure 7 , The partial derivative estimation operation 68 executed by the differential analysis sub-module 40 calling the intensity modulation optimization sub-module 30 can be performed efficiently. This is because the optimization for each partial derivative estimate does not start from the beginning, but can be performed using the output 64 of the intensity modulation optimization operation 62 as a starting value. Such as Figure 7 One is schematically shown. The partial derivative estimation operation 68 for the contour segment selected in operation 66 starts from operation 80, and in operation 80, the selected contour segment is moved outward (or, in another embodiment, inward ) Adjust the difference Δ CS (For example, in Figure 4 Difference adjustment in Δ CS1 ,or Figure 5 Difference adjustment in Δ CS2 ). In operation 82, the intensity modulation optimization iteration is performed with the contour segment difference of operation 80, so as to generate an output 84 having the intensity modulation optimization of the contour segment difference. The iteration 82 is performed using the output 64 of the intensity modulation optimization operation 62 as the starting value. Due to the change of the difference Δ CS It is very small and only affects the selected contour segment, so the expected output 64 is close to the "correct" value that will be produced by multiple iterations. Therefore, it is expected that the output 84 of the single iteration 82 is subtracted from the output 64 according to the subtraction operation 86 (and optionally divided by the difference Δ CS ) Provides a reasonable estimate of the partial derivative of the output of the selected contour segment. For the purpose of contour fragments affecting classification 70, the subtraction operation 86 may alternatively be replaced with a more complex difference metric. Furthermore, because only a single iteration 82 is performed, the partial derivative estimation operation 68 is efficient. Although a single iteration 82 is exemplified, in other embodiments two or more iterations may be performed, either in an open loop manner or iterating until the selected stopping condition is met.

[0038] Back for reference figure 2 After each contour segment update operation 76, an impact analysis 66, 68, 70 is performed. In this way, the user generally observes the influence of the contour drawing operation on the intensity modulation optimization in real time, that is, during the contour drawing process, including the influence of various contour fragments that may occur during the contour drawing process on the intensity modulation optimization Any changes. However, in some radiation therapy applications, it is not expected that the influence of contour fragments will change significantly during the contour drawing process. Therefore, it is actually feasible to pre-calculate the influence of various contour fragments on the intensity modulation optimization. This can be done, for example, based on a previous contouring/radiation therapy planning procedure performed for the current object or for the previous object. The pre-calculated influence can also be based on first-principles analysis, for example using an attenuation model.

[0039] reference Figure 8 ,show figure 2 The processing variant. Figure 8 Processing and figure 2 The difference in processing is that the impact analysis operations 66, 68, 70 are replaced by the retrieval operation 90. The retrieval operation 90 retrieves the pre-calculated contour segment impact classification, and then the display sub-module 24 in the display operation 74 uses the classification to display Contours coded by the influence on intensity modulation optimization. in Figure 8 In the embodiment, the coding of the contour is based on the pre-calculated contour segment influence classification. Appropriately apply this method to the following situation: a class of planning tasks are sufficiently similar that the pre-computed contour segment impact classification is applicable.

[0040] The disclosed contour drawing method with contour influence coding includes calculation and analysis of a treatment plan as part of the contour drawing process. Radiologists or other medical experts who perform contour delineation receive visual "importance" feedback during manual contour delineation, or in the case of automatic or semi-automatic contour delineation, by incorporating important features along the contour into an automatic segmentation algorithm , And benefit. In some embodiments, the contour influence code is input into the automatic segmentation/contour delineation algorithm, which allows relaxation of the accuracy requirements of the automatic algorithm. During manual contour drawing, the technician is told which area of ​​the initial contour drawing needs to be improved. The intensity modulation optimization sub-module 30 is supplemented by the differential analysis sub-module 40 to enable estimation of partial derivatives with respect to contour segments. As reference Figure 7 As mentioned, the differential analysis sub-module 40 optionally estimates the partial derivative in the first approximation through a fast numerical scheme, for example using a single iteration of intensity modulation optimization.

[0041] Other variants are also expected. For example, a radiologist can first quickly make a rough outline of the tumor (or other radiation target area) and all at-risk organs (or, more generally, at-risk areas). Then the treatment plan is calculated and analyzed, specifically, the differential analysis of the contour is performed. The result of this analysis assigns a quantitative "impact" value to each contour segment for changes in plan/dose deposition. Color coding, or line coding, or line thickness coding, etc. can be used to encode different influence values ​​in the display, thereby indicating where the outline drawing should be more precise or less precise. When the technician updates the contour (for example, by operating the mouse or track ball and using pointing and dragging operations), the differential dose calculation is completed again and the contour color coding is updated.

[0042] For some applications, such as prostate treatment planning where the radiation target area is the prostate and the risk area to be delineated typically includes the femoral head, bladder, and rectum, no matter how accurate the delineation is, certain contour fragments are unlikely to have high impact. At the same time other contour fragments need to be very accurate. In this application, reference can be used Figure 8 The described variant method, in which the influence value is pre-calculated in advance (for example, pre-trained with multiple test cases). In the method of automatic or semi-automatic segmentation, the learned or pre-calculated influence value becomes a part of the model (that is, the prior knowledge contained in the model) used for automatic or model-based segmentation. During the model-based segmentation process, the influence value is used as an additional (contingent) feature to balance the accuracy in areas where the contours are difficult to delineate. In such an embodiment, the calculation of the dose during the split can optionally be avoided.

[0043] The invention has been described with reference to the preferred embodiment. Others can think of modifications and variations when they read and understand the foregoing detailed description. The present invention is intended to be construed as including all such modifications and variations as long as they fall within the scope of the appended claims or their equivalents.