Method, computer program product, and computer system for particle-based radiation treatment planning

Abstract

A radiation treatment planning method for particle-based arc therapy, comprising steps of determining an initial set of candidate energy layers (S11, S21); performing calculations on the initial set; determining at least one additional energy layer including a previously unused energy level based on results of the calculations to generate an extended candidate layer set (S13, S23); adding the at least one additional energy layer to the initial set to generate an extended candidate energy layer set (S14, S24); and optimizing the radiation treatment plan based on the extended candidate layer set (S16, S26).

Description

[Technical field]

[0001] The present invention relates to systems and methods for particle-based radiation therapy planning and delivery, and in particular to arc treatment planning. [Background technology]

[0002] In particle-based radiation therapy, a patient is irradiated with charged particles, such as protons. The particles are controlled to deposit most of their energy at a specific depth within the patient to cover the entire target. At the same time, some energy is deposited along the path to the target. Each particle deposits most of its energy toward a stopping point of its path, known as the Bragg peak. The depth of the Bragg peak within the patient can be controlled by adjusting the kinetic energy of the particle. The lateral location of the Bragg peak can be controlled by deflecting the focused beam with electromagnets. This allows for the delivery of a highly localized dose to a well-controlled location within the patient. The dose delivered from a particular combination of the beam's kinetic energy and lateral deflection is called a spot. The number of particles delivered to a spot is commonly referred to as the spot weight. By providing spots at many different locations in three-dimensional space, the target volume can be adequately covered with the desired dose distribution. This procedure is called active scanning ion beam therapy, also known as pencil beam scanning.

[0003] The kinetic energy of the spots is often, but not necessarily, distributed at several discrete energies. A group of spots with the same kinetic energy but different lateral deflections is often called an energy layer. To cover a desired volume within the patient, different energy layers are defined such that the particles of a particular energy layer deposit their energy at a particular depth within the patient. The energy layers are selected to distribute the Bragg peaks throughout the volume to be treated.

[0004] The spot weights for all energy layers are determined through optimization in the treatment planning system, where the spot weights are iteratively varied to achieve a desired goal or constraint related to dose level or other relevant quantity. The case where the spot weights are allowed to vary freely across the beam is called Intensity Modulated Proton Therapy (IMPT).

[0005] Proton arc therapy involves the application of a series of multiple beams, each with several energy layers, for example up to 20, from various angles, for example 10 or more, 50 or more, or 100 or more. The adjustment of each energy layer and each angle takes time. The arc can be delivered as several individual beams or through a continuously moving arc. In the latter case, as known from photon arc therapy, the beam serves as a control point. Hereinafter, the terms beam, beam direction, or beam angle can be equivalently replaced by the term control point when the arc is continuous. The term energy layer in this specification refers to a specific energy level of a specific beam. If the same energy level is used in two beams, this is considered to be two separate energy layers.

[0006] The selection of energy layers, including the selection of the beam angle across the arc, is critical to achieving a high quality plan. A higher number of energy layers often allows for a higher plan quality, but also increases the delivery time to the patient, which increases the cost of the resulting plan in terms of clinical efficiency and economic considerations as well as the risk of patient movement within the fraction and patient discomfort during treatment. There is always a desire to reduce the delivery time while still maintaining a high quality plan that ensures effective treatment of the patient. Thus, treatment planning should seek to select energy layers that allow for a high quality plan with a limited set of energy layers.

[0007] US Patent Application Publication No. 2019 / 0091488 proposes an iterative method for proton arc therapy, including resampling control points, redistributing energy layers, filtering energy layers, and resampling energy layers at each iteration. Each control point corresponds to a beam angle. A small number of control points, each with several energy layers, are defined and the plan is optimized. After optimization, one control point is randomly selected, split, and replaced with a pair of adjacent sub-control points, e.g., 5 degrees on either side of the original control point. The energy layer of the original control point is copied to both sub-control points or split between the sub-control points before optimizing the plan again. Since the sub-control points are at a small angle from the original control point, it is assumed that the same energy level should be used. This is repeated many times, and therefore the sub-control points are also split and replaced with two adjacent control points with all or a subset of the same energy layers. Spots and / or energy layers with low weights are then filtered out by conventional spot or energy layer filtering methods.

[0008] Co-pending European Patent Application No. 18165472.4 discloses a treatment planning method for particle-based arc therapy that starts with a set of energy layers, the number of energy layers included in the plan can be reduced as part of the optimization step by including in the optimization problem a penalty designed to limit the number of energy layers. This allows for optimization of the number of energy layers and their positions in the same operation as optimizing the treatment plan. The method only selects energy layers already in the set and does not foresee adding additional energy layers. Summary of the Invention [Problem to be solved by the invention]

[0009] It is an object of the present invention to provide a set of energy layers upon which arc radiation treatment planning can be based that provides high quality treatment plans while minimizing the computational effort required. [Means for solving the problem]

[0010] This object, according to the invention, is to provide a radiation therapy planning method for creating a plan for the delivery of at least one particle-based arc to a patient using a device configured to deliver radiation in the form of charged particles from several different directions, comprising: a. determining an initial set of candidate energy stratums, each of which includes a particular energy level at a particular beam angle; b. performing calculations on the initial set; c. determining at least one additional energy tier including energy levels not previously used in any energy tier based on results of said calculations to generate an expanded candidate tier set; d. adding at least one additional energy tier to the initial set to generate an extended candidate energy tier set; e. optimizing the radiation treatment plan based on the expanded candidate layer set; This is achieved by a method comprising:

[0011] Therefore, according to the present invention, the candidate set of energy layers can be expanded by adding energy layers that contribute significantly to the resulting plan. The energy layers to be added are determined based on the results of plan optimization or dose calculation. For example, energy layers can be added in areas where a lot of energy is deposited, where another energy layer would be useful. Alternatively or additionally, energy layers can be added to create spots in areas where coverage is lacking to improve coverage. Importantly, since energy levels not previously used in any energy layer may be used, one or more of the added energy layers will have energy levels not previously used in any energy layer. Energy layers with previously unused energy levels can be added to existing beam angles or at newly added beam angles.

[0012] In one embodiment, the calculation comprises a dose calculation, and the determination of the at least one additional energy layer is based on the results of the dose calculation identifying, for example, an area with a low Bragg peak density, and the additional energy layer is determined to produce a Bragg peak in said area, which involves limited computational effort and in particular allows the addition of new energy layers without first performing an optimization.

[0013] In a preferred embodiment, the calculation includes optimizing the treatment plan based on an optimization problem, and the determination of at least one additional energy layer is based on the resulting treatment plan. The optimization requires more computing power, but also produces a better basis for determining the new energy layer in the best possible manner.

[0014] In a preferred embodiment, the method of extending the energy layers is combined with a method step for discarding energy layers. One such method includes optimizing the plan using an optimization problem formulation designed to provide a suitable dose distribution in such a way that the plan uses only a subset of beam angles and energy layers from the set of beam angles and the set of energy layers, respectively. This can be done using a greedy or reverse greedy method, both starting from a candidate set of energy layers. In the reverse greedy method, the full candidate set is included from the beginning, and layers are discarded according to their spot weights and / or their gradient-based scores. After a first optimization, the layers to be discarded are determined. For example, the n layers with the lowest scores are discarded, n being an integer, for example 10% of the candidate set. Alternatively, all layers with scores below a threshold can be discarded. The optimization is then repeated, discarding the m layers with the smallest sum of spot weights, m being an integer equal to t or different from n. This can be repeated again until a certain number of energy layers remains. A combination of greedy and reverse greedy approaches may be used by first generating a subset using the greedy method and then refining the subset using the reverse greedy method.

[0015] In the Greedy method, the initial optimization is based on a subset of the candidate set. A gradient-based point-based score is determined. Based on the score, layers of the candidate set that are not included in the subset can be added to the subset and a new optimization can be performed. For example, a fixed number of layers can be added, or all layers with a score above a certain threshold can be added. This can be repeated until a certain number of energy layers is reached. In both the Greedy and Reverse Greedy methods, the optimization can be repeated, for example, until a predefined number of energy layers or a certain fraction of the candidate set remains, or until all remaining energy layers have a spot weight sum above a threshold. The steps of discarding energy layers in the Reverse Greedy method and including energy layers in the Greedy method are defined to operate on the entire arc or on one or more subsets of the arc, each of which includes one or more beams.

[0016] In a preferred embodiment, the optimization is penalized in a way designed to limit the number of energy layers, ensuring that the best energy layers, i.e., those that contribute most effectively to dose, are retained and that those that result in low dose are discarded.

[0017] A penalty may be defined as one or more constraints or terms of the objective function used in the optimization, or a combination of the two.

[0018] The penalty can be expressed in terms of the number of energy layers. Changing to a different energy takes time, so minimizing the number of energy layers used will reduce the delivery time of the plan. Alternatively, the penalty can be expressed in terms of the number of energy layer changes between adjacent angles. This would allow, for example, to preserve the energy layers when changing the beam angle, thus saving time. The penalty can also be expressed in terms of both the number of energy layers and the number of energy layer changes.

[0019] The penalty may be defined to act on the entire arc, or on one or more subsets of the arc, each of which may contain one or more beams.

[0020] The invention also relates to a computer program product for controlling a radiation therapy planning device, preferably stored on a carrier such as a non-transitory storage means, and comprising computer readable code means which, when running on a processor of the radiation therapy planning device, causes the device to carry out the aforementioned method.

[0021] The invention also relates to a radiation therapy planning apparatus comprising a processor and a program memory holding a computer program product according to the above, adapted to run on the processor to control the radiation therapy planning apparatus. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The invention will now be explained in more detail, by way of example only, with reference to the accompanying drawings, in which: FIG. [Brief description of the drawings]

[0023] [Figure 1] 1 is a flow chart of a first embodiment of the method according to the present invention; [Diagram 2] 4 is a flow chart of a second embodiment of the method according to the invention. [Diagram 3] FIG. 1 illustrates a schematic diagram of a system in which the present invention may be implemented; DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0024] The method according to the invention is based on a candidate set of energy layers that can be expanded according to certain criteria.

[0025] In its simplest form, the method is as shown in FIG. 1. In a first step S11, an initial set of energy layers is obtained. As mentioned above, an energy layer is an energy level associated with a particular beam angle. In this embodiment, the initial set should preferably be small enough to allow an initial dose calculation or optimization to be performed within a reasonable time frame. The set may consist of, for example, one beam angle, or several beam angles each with several energy layers. Then, in step S12, a calculation, for example, optimization of the treatment plan, is performed based on the initial set. In step S13, the result of the calculation is used to determine at least one additional energy layer at a new beam angle or at a beam angle already included in the initial set. Then, in step S14, the initial set of energy layers is expanded by adding one or more energy layers from step S13 to generate an expanded set. An energy layer can be added at an existing beam angle and / or one or more new beam angles can be selected. This can be repeated one or more times by treating the expanded set as the initial set to be expanded again and returning to step S11.

[0026] In one embodiment, the calculations in step S12 include a treatment plan optimization. In that case, information from the resulting treatment plan can be used to determine additional energy layers in step S13. In other embodiments, step S12 instead includes simpler calculations such as dose calculations for each energy layer in the initial set. In the latter case, step S13 determines one or more additional energy layers based on the resulting doses calculated for each energy layer. This provides less information than treatment plan optimization, but it does provide information such as Bragg peak locations that can be used to determine additional energy layers. For example, regions with low density of Bragg peaks can be identified, and additional energy layers can be determined to ensure that new Bragg peaks arise in these regions to ensure adequate Bragg peak density at the target.

[0027] Step S15 is a decision step to determine whether another expansion of the set of energy layers should be performed. If yes, return to step S12 and perform the calculations of step S12 with the expansion set. If no, in step S16, perform optimization based on the expansion set to obtain an arc treatment plan. Step S17 is an optional step to filter spots with low spot weight. Spot filtering is performed not only to increase the treatment efficiency, but also because machines have physical limits on how small a spot weight they can deliver. Usually, the spots are also sorted, i.e., the order in which they are irradiated is determined to minimize the time required. Spot filtering and spot sorting are known procedures and can be performed at any appropriate point in the procedure.

[0028] In all embodiments, the selection in step S13 and the following step S23 should be done in such a way as to add energy tiers that may be useful in the plan, i.e. energy tiers that contribute significantly to the total dose. In practice, several different criteria can be applied to achieve this.

[0029] For example, new energy layers can be added near some energy layers, resulting in high spot weights in the resulting treatment plan. Alternatively or additionally, the geometry of the Bragg peak locations can be taken into account, such that new energy layers are added that result in new Bragg peaks in areas with low density of Bragg peaks. A further criterion can be that if there is a large amount of radiation from one beam angle or a group of adjacent beam angles, new energy levels are added near those beam angles, preferably in the form of new beam angles each having several energy layers. To increase the energy level density in the beam, new energy layers can be added near energy layers with high weights in the same beam. The beam angles and energy layers are selected based on what makes the most useful contribution to the total dose, regardless of the beam angles and energy levels of the existing energy layers, in the sense that new energy layers can be added appropriately and accurately by being able to use new energy levels that were not originally used. This means that in the case of an existing beam angle, one or more energy layers can be added wherever it is deemed useful, regardless of the existing energy layers of this or other beams. In the case of a new beam angle, the energy layer is selected as the energy layer that is optimal for this new beam angle, regardless of the energy layers of the existing beam angles.

[0030] After a desired number of iterations, the resulting expansion set is used in an optimization step to generate a treatment plan.

[0031] There may be a predefined candidate set of possible energy tiers including possible beam angles that can be selected to add to the initial set of energy tiers, or new beam angles and associated energy tiers may be freely defined.

[0032] Preferably, the method also includes a method for reducing the number of energy layers to ensure that energy layers that are not important for the plan are discarded. This usually means removing energy layers that contribute little to the total dose, which reduces the computational effort with minimal impact on the plan quality. This can also be done by simple methods such as energy layer filtering, where energy layers with a small sum of the layer spot weights are filtered out. According to a preferred embodiment, this is achieved by an optimization problem with a penalty designed to reduce the number of energy layers.

[0033] FIG. 2 illustrates such an embodiment, in which a series of expansions and contractions of a set of energy layers is performed iteratively.

[0034] In a first step S21, a potential set of energy layers is defined. In the case of pencil beam scanning therapy, a set of spots for each layer is also defined. This may involve obtaining an arc trajectory and specifying several beam angles of the arc and the energy layers associated with each beam angle.

[0035] In the next step S22, an optimization is performed, where the treatment plan is optimized based on the potential set of energy layers defined in step S21. In the optimization problem, one or more penalties are set for the number of energy layers and / or the number of changes of the energy layers and / or the number of beams. In the following, the singular "penalty" is used, but it should be understood that several penalties may be applied together. Further details of this step are provided below. This optimization results in a reduced set of energy layers, which can be used as an initial set of energy layers in the method of FIG. 1 or can be used to create an expanded set as described below.

[0036] In the following steps, the reduced set of energy layers from step S22 is expanded to generate an expanded set. In step S23, the results of the optimization of the treatment plan in step S22 are used to determine new beam angles or at least one additional energy layer at a beam angle already included in the initial set. Then, in step S24, the initial set of energy layers is expanded by adding one or more energy layers from step S23 to generate an expanded set. An energy layer can be added to an existing beam angle and / or one or more new beam angles can be selected.

[0037] Step S25 is a decision step to determine whether another expansion of the set of energy layers should be performed. If yes, return to step S22 and perform the optimization of step S22 using the expanded set as a new potential set. If no, in step S26, perform the optimization based on the expanded set using the same type of optimization problem as described for step S22 to obtain the arc treatment plan. Step S27 is an optional step of filtering out energy layers with low spot weights in a conventional manner. As with the procedure of FIG. 1, spot filtering and spot sorting can be performed at any appropriate point in the procedure.

[0038] In steps S22 and S26, the optimization is performed using an optimization problem that includes a penalty designed to reduce the number of energy layers. The penalty may be defined as a constraint that sets an absolute limit or as a term in the objective function that reflects a desire to limit the number of energy layers as much as possible. The penalty may be defined to limit the number of energy layers, to limit the number of energy layer changes during treatment, or in any other suitable way that ensures that the optimization algorithm systematically reduces the time spent changing energy layers during treatment delivery. In step S22, the optimization can start with a full initial set of beam angles and energy layers, and thus the optimization process reduces the number of beam angles and energy layers by removing those that contribute the least to the plan. When an energy layer is removed, the spot weights of that layer are distributed to the remaining energy layers to minimize the disturbance to the solution caused by the layer removal. Methods for doing this are known in the art and may include building a three-dimensional Delaunay triangulation of the Bragg peaks of the remaining layers and distributing the weight of each spot to be removed according to its position in the triangulation. Alternatively, the optimization can start without beam angles and energy tiers, and the optimization process proceeds to add those that make the most useful contribution to the plan. In both cases, the possible beam angles and energy tiers are from the set defined in step S21. Some more specific examples of both types of optimization methods are provided below.

[0039] The penalty can be set to act on the whole arc and limit the number of energy layers per arc to a maximum value. Alternatively, the penalty can be set to act on each angle separately and set the maximum number of energy layers per beam. As a third option, sub-arcs can also be defined, each including a subset of the arc, and the penalty can be set to limit the number of energy layers in each sub-arc. The same applies when the penalty is set in terms of the objective function. A sub-arc can include any suitable number of consecutive beams, for example three or five, but a sub-arc can also include only one beam. In particular, if a penalty is specified to limit the number of energy layer changes, it would be appropriate to set the penalty on the whole arc or on sub-arcs including more than one angle. This is because in this case such a penalty can avoid energy layer changes between two beams by using the last energy layer of one beam as the first energy layer of the next beam.

[0040] The energy layer selection penalty is a vector of spot weights x ∈ R m A function y that maps (defined over all energy layers and angles) to R m →R n can be mathematically formulated by introducing a specific measure of how much of each energy layer is used at each angle. n is preferably y if the energy layer and angle combination at index i has a spot with a positive weight. i The component function y(x) can be defined to be positive and zero otherwise. i One possible formulation of is the subvector x associated with a particular energy layer and angle. i Norm of

number

[0041] Such methods start with a complete set of beams and energy layers and remove some during optimization. The cardinal operators are now approximated by continuous functions, which removes the combinatorial aspect of the problem. The resulting approximate problem can be solved by several methods for continuous optimization, such as interior point methods or sequential quadratic programming. One possible reformulation of the problem with a cardinal penalty into a continuous problem is to reduce the cardinal operators to card(y)=Σ i s(y i ), where s is taken to be zero for non-positive numbers and one otherwise, and then the exact function s is replaced by a continuous but approximate step function. Examples of functions that can be used to approximate s are the logistic function or the error function, or simply the positive component function. This method can also be applied in several steps, so that multiple optimizations are performed, and the accuracy of the approximation of the step function is increased between optimizations until the result is acceptable, taking into account the cardinality penalty. If the cardinality penalty is formulated as a constraint, this means that the constraint is satisfied. If the cardinality penalty is formulated as a term in the objective function, the term is minimized.

[0042] The objective function and constraints in the optimization problem should be optimizable for a variety of quantities, most importantly dose and LET related quantities such as LET and dose weighting LET. (When considered in the optimization, LET (linear energy transfer) can be more efficiently focused inside the target with a proton arc than with ordinary IMPT.)

[0043] At any point in the optimization procedure, usually at the end, spot filtering is applied to filter out spots that have been assigned low weights, as indicated by steps S17 and S27. Preferably, the weights of these filtered out spots are reassigned to other spots that are retained in the plan. Several ways of doing this are known in the art and will not be described in detail here.

[0044] If the initial set of energy layers has too many spots to calculate and store the doses of all spots, it is possible to pre-process the energy layers to remove some of the energy levels before starting the procedure according to the invention. This can be done in any suitable way. For example, for two adjacent beam angles with the same or similar energy levels, some energy levels can be removed from each beam angle in such a way that what is removed from one is kept in the other, or vice versa. The pre-processing method can remove energy layers uniformly until the candidate set is small enough to perform numerical calculations. It can also be based on certain rules, such as prioritizing distal layers over proximal layers. Pre-processing can also be based on the Bragg peak position, if approximate inexpensive techniques are available to calculate the Bragg peak position from the spot positions and energies. For such methods, the pre-processing can remove energy layers in a way that tries to maintain a uniform spread of the Bragg peak throughout the target volume.

[0045] FIG. 3 is an overview of a system for radiation therapy imaging and / or planning adapted to provide radiation to a patient from different directions. As will be appreciated, such a system may be designed in any suitable manner, the design shown in FIG. 3 being merely an example. A patient 11 is placed on a treatment couch 13. The system comprises a treatment unit 10 having a radiation source 15 mounted on a gantry 17 for emitting radiation towards the patient placed on the couch 13. Typically, the couch 13 and the gantry 17 are movable in several dimensions relative to each other in order to provide radiation to the patient as flexibly and precisely as possible. In particular, the gantry can rotate around the couch either between certain angles or in a full 360° rotation. Alternatively, the gantry can be fixed and the couch supporting the patient can rotate instead. Another alternative is to have the patient sit in a chair. The arc treatment in this case can be performed by changing the direction of the beam or by rotating the chair, which are well known to those skilled in the art. The system also comprises a computer 21 that can be used for radiation therapy planning and / or for controlling the radiation treatment. As will be appreciated, the computer 21 may be a separate unit that is not connected to the imaging unit.

[0046] The computer 21 comprises a processor 23, a data memory 24 and a program memory 26. There are also one or more user input means 28, 29, preferably in the form of a keyboard, a mouse, a joystick, voice recognition means or any other available user input means. The user input means may also be arranged to receive data from an external memory unit.

[0047] The data memory 24 comprises clinical data and / or other information used to derive a treatment plan, including a set of clinical targets used in the planning. The data memory 24 also comprises one or more dose maps for one or more patients used in the treatment planning according to embodiments of the present invention. The program memory 26 holds computer programs, known per se, configured for the optimization of treatment plans. The program memory 26 also holds computer programs configured to cause a computer to control the treatment of a patient according to the present invention.

[0048] It is understood that the data memory 24 and the program memory 26 are shown and described only diagrammatically. There may be several data memory units each holding one or more different types of data, or one data memory holding all data in a suitably structured manner, the same applies to the program memory. The one or more memories may also be stored on another computer. For example, a computer may be configured to perform only one of the methods and there may be another computer for performing the optimization.

[0049] To further reduce damage to surrounding tissue, the described optimization method can also be combined with an approach where the plan consists of similar arcs delivered in different treatment fractions delivered on different days, with the different arcs containing different beam angles and therefore the target being covered in the same way, but with different fractions taking different paths from the patient to the tumor.

[0050] Radiotherapy using charged particles such as protons allows for very accurate plan optimization. This in turn means that small errors, for example in the setup, can have a large impact on the delivered dose. Therefore, the optimization method used according to the present invention should preferably be a robust optimization method. This becomes even more important when the number of energy layers is reduced. The method should be robust against various uncertainties to ensure that the quality of the plan is preserved even when there are uncertainties in the delivery (for example, range and setup uncertainties). As an alternative to a robust optimization method, a method aiming to produce a uniform dose at each angle (for example, Single Field Uniform Dose (SFUD) in conventional active scanning plan optimization) or at each set of angles (each sub-arc) would be appropriate. If a low number of energy layers per angle is desired, the most feasible solution is to aim for a uniform dose at each sub-arc.

[0051] The described methods can also be employed for all types of active scanning, including spot scanning, raster scanning, and line scanning. The implementation of line scanning in conjunction with proton arc therapy would be beneficial from the perspective of delivery time. These methods are suitable for planning human treatments, but also for small animals such as mice for preclinical trials. The described methods work equally well for light ions other than protons, such as helium, carbon, and oxygen, as long as the different biological effects of the different ion types are taken into account in the biologically relevant (RBE) model included in the optimization process.

Claims

1. 1. A method of operating a radiation therapy planning device to generate a plan for delivering at least one particle-based arc to a patient using a device configured to deliver radiation in the form of charged particles from a number of different directions, comprising: a. The radiation therapy planning device determines an initial set of candidate energy stratums (S11, S21), each of which includes a specific energy level at a specific beam angle; b. The radiation therapy planning device performs dose calculation using the initial set; c. The radiation therapy planning device determines at least one additional energy stratum including an energy level not previously used in any energy stratum to generate an extended candidate stratum set based on the result of step b (S13, S23); d. The radiation therapy planning device adds the at least one additional energy layer to the initial set to generate an extended candidate energy layer set (S14, S24); e. The radiation therapy planning device optimizes the radiation therapy plan based on the expanded candidate layer set (S16, S26), the optimization being based on an optimization problem formulation designed to produce an appropriate dose distribution; The method of operation includes:

2. 2. The method of claim 1, wherein the determination of the at least one additional energy layer is based on results of a dose calculation that identifies an area with a low Bragg peak density, and the additional energy layer is determined to produce a Bragg peak in the area.

3. The method of claim 1 , wherein the determination of the at least one additional energy layer is based on a resulting treatment plan.

4. 2. The method of claim 1, wherein the optimization (S22, S26) is performed in such a manner that the plan uses only a subset of beam angles and / or a subset of energy layers from a set of beam angles and a set of energy layers, respectively.

5. 5. The method of claim 4, wherein a potential set of energy strata is provided, and the optimization comprises a first optimization procedure based on the entire potential set and a subsequent step (S26) of discarding a portion of the energy strata from the potential set based on spot weights and / or gradients to create a reduced potential set, and performing a second optimization procedure based on the reduced potential set.

6. 5. The method of claim 4, wherein a potential set of energy layers is provided, and the optimization comprises a first optimization procedure based on a first subset of the potential set and a subsequent step of adding to the first subset a portion of the energy layers from the potential set that are not included in the first subset based on a calculation of gradients or spot weights with respect to an optimization goal to create an extended subset, and performing a second optimization procedure based on the extended subset.

7. The method of claim 4 , wherein the optimization is designed to limit a number of energy layers.

8. The method of claim 7 , wherein the energy stratum limit is achieved by a penalty implemented as a constraint or term in an objective function used for optimization.

9. The method of claim 8 , wherein the energy layer limitation is achieved by a penalty expressed in the number of energy layers or the number of changes in the energy layer.

10. A computer program product for controlling said radiation therapy planning apparatus, said computer program product being stored on a carrier comprising storage means and comprising computer readable code means which, when running on a processor of said radiation therapy planning apparatus, causes said apparatus to perform the method of operation of any one of claims 1 to 9.

11. 11. The radiation therapy planning device (10), comprising a processor (23) and a program memory (26) holding a computer program product according to claim 10 adapted to operate on the processor to control the radiation therapy planning device.

Images