EP-4292647-B1 - PARALLEL PROCESSING FOR MULTI-PASS OPTIMIZATION OF RADIOTHERAPY PLANS

EP4292647B1EP 4292647 B1EP4292647 B1EP 4292647B1EP-4292647-B1

Inventors

DA SILVA, JOAKIM SEBASTIAN
HENNIX, MARCUS LARS ERIC
NORDSTRÖM, Carl Axel Håkan

Dates

Publication Date: 20260506
Application Date: 20230613

Claims (14)

A computer-implemented method for radiotherapy treatment planning, comprising: obtaining a set of first optimization problems for providing radiotherapy treatment to a human subject, the first optimization problems defined by a first plurality of parameters; performing dose optimization for delivery of the radiotherapy treatment to at least one treatment area of the human subject, the dose optimization comprising: converting the first optimization problems into a first problem matrix; performing a first pass of the dose optimization by solving, using a linear programming solver operating based on an alternating direction method of multipliers technique, the first optimization problems represented in the first problem matrix in parallel on parallel processing hardware, the first pass to produce a first set of multiple solutions, corresponding to a first plurality of multiple sets of weights, to the first optimization problems; combining the first set of multiple solutions to the first optimization problems to produce a set of second optimization problems for providing the radiotherapy treatment, the second optimization problems defined by a second plurality of parameters; converting the second optimization problems into a second problem matrix; and performing a second pass of the dose optimization by solving, using the linear programming solver operating based on the alternating direction method of multipliers technique, the second optimization problems represented in the second problem matrix in parallel on the parallel processing hardware, the second pass to produce a second set of multiple solutions, corresponding to a second plurality of multiple sets of weights, to the second optimization problems; and generating treatment plan data based on at least one solution of the second set of multiple solutions to the second optimization problems, wherein the treatment plan data is usable to control delivery of radiotherapy from a radiotherapy machine; wherein the at least one treatment area includes a low-dose region and a target region, wherein a dose to be delivered in the low-dose region is a fraction of a dose to be delivered in the target region, and wherein combining the first set of multiple solutions to produce the second optimization problems comprises: performing a union of the first set of multiple solutions to the first optimization problems for the low-dose region.
The method of claim 1, wherein the first and the second plurality of parameters define constraints for at least one target area and at least one low-dose area in the at least one treatment area.
The method of any of claims 1 to 2, wherein the first and second plurality of multiple sets of weights, corresponding to the first and second set of multiple solutions, relate to points defined for at least one low-dose volume.
The method of any of claims 1 to 3, wherein the first plurality of parameters and the second plurality of parameters relate to radiation delivery parameters of a radiotherapy treatment machine.
The method of any of claims 1 to 4, wherein solving the first optimization problems or the second optimization problems in parallel on the parallel processing hardware comprises, for a respective set of problems: identifying parameterized linear programming equations from the respective set of problems; and converting the parameterized linear programming equations for execution by the parallel processing hardware; and wherein solving the respective set of problems in parallel comprises solving a plurality of the converted parameterized linear programming equations in parallel on the parallel processing hardware, to produce a plurality of solutions to the respective set of problems.
The method of claim 5, wherein converting the parameterized linear programming equations comprises applying the alternating direction method of multipliers technique, and wherein the alternating direction method of multipliers technique comprises transforming the converted parameterized linear programming equations to matrix and projection operations.
The method of any of claims 1 to 6, wherein the parallel processing hardware comprises a set of one or more graphics processing units (GPUs).
The method of any of claim 1 to 7, wherein each low-dose point selected from a common low-dose region of the first set of multiple solutions of the first optimization problems is represented in the second problem matrix, and wherein performing the second pass of the dose optimization includes assigning a non-zero upper bound to a solution vector in a subset of points corresponding a respective low-dose region for each set of weights.
The method of any of claims 1 to 7, wherein each low-dose point selected from a common low-dose region of the first set of multiple solutions of the first optimization problem is represented in the second problem matrix, and wherein performing the second pass of the dose optimization includes applying a new low-dose weight to all low-dose points in the union of the first set of multiple solutions of the first optimization problems.
The method of any of claims 1 to 7, wherein combining the first set of multiple solutions to produce the second optimization problems comprises: performing a sampling of the union of the first set of multiple solutions to identify weights of the second plurality of parameters for the low-dose region.
The method of any of claims 1 to 10, further comprising: selecting a solution to the second optimization problems based on an evaluation of the second set of multiple solutions, wherein the treatment plan data is generated based on the selected solution to the second optimization problems; and optionally, wherein the selected solution to the second optimization problems provides an approximate solution, with the method further comprising receiving an additional optimization to the selected solution, wherein the treatment plan data is generated based on the additional optimization to the selected solution.
The method of any of claims 1 to 11, wherein the treatment plan data for the radiotherapy treatment comprises a set of treatment delivery parameters corresponding to capabilities of a radiotherapy treatment machine; and optionally, wherein the radiotherapy treatment is to be provided with a Gamma knife, and the set of treatment delivery parameters comprises a set of isocenters used for delivery of the radiotherapy treatment; and optionally, wherein the set of treatment delivery parameters further comprises timing for delivery of the radiotherapy treatment and a collimator sequence for the delivery of the radiotherapy treatment; and optionally, wherein the radiotherapy treatment is provided with a Volumetric-modulated arc therapy (VMAT) or Intensity modulated radiation therapy (IMRT) using a Linac radiotherapy machine, and wherein the set of treatment delivery parameters comprises: a set of arc control points for one or more arcs, fluence fields, gantry speed, and dose rate along the one or more arcs.
A computer-readable storage medium comprising computer-readable instructions for radiotherapy treatment planning, wherein the instructions, when executed, cause a computing machine to perform any of the methods of claims 1 to 12.
A computing system configured for radiotherapy treatment planning, the system comprising: one or more parallel processing hardware devices; one or more memory devices to store data for providing radiotherapy treatment to a human subject; and one or more processors configured to perform any of the methods of claims 1 to 12.

Description

TECHNICAL FIELD Embodiments of the present disclosure pertain generally to processing and optimization techniques used in connection with a radiation therapy planning and treatment system. In particular, the present disclosure pertains to methods for the use of specific computing hardware configurations to optimize dosage in a radiation therapy session, which involve multiple passes or iterations to perform an optimization. BACKGROUND Radiation therapy (or "radiotherapy") can be used to treat cancers or other ailments in mammalian (e.g., human and animal) tissue. One such radiotherapy technique is provided using a Gamma Knife, by which a patient is irradiated by a large number of low-intensity gamma rays that converge with high intensity and high precision at a target (e.g., a tumor). Another such radiotherapy technique is provided using a linear accelerator (Linac), whereby a tumor is irradiated by high-energy particles (e.g., electrons, protons, ions, high-energy photons, and the like). The placement and dose of the radiation beam must be accurately controlled to ensure the tumor receives the prescribed radiation, and the placement of the beam should be such as to minimize damage to the surrounding healthy tissue, often called the organ(s) at risk (OARs). In radiotherapy, treatment plans are usually generated by solving an optimization problem that balances various conflicting objectives, such as high dose to target, normal tissue sparing, and treatment complexity. It is therefore a multicriteria optimization (MCO) problem. Commonly, the different criteria are combined using a weighted sum, where each weight determines the relative importance of that criterion. Finding acceptable weights to develop an optimal radiotherapy plan with prior techniques may involve a manual and tedious process of trial-and-error, especially because evaluating a single choice of parameters (e.g., a single set of weight values) requires solving a full optimization problem for the radiotherapy treatment. Solving a full optimization problem for a single treatment plan (i.e. corresponding to a single set of weights) with conventional planning techniques may take from a few seconds up to an hour, depending on the application. When performing a dose optimization process for multiple anatomical regions to be treated, two (or more) passes (i.e., two iterations) of processing operations may be needed. The first pass may be used to determine an optimal dose distribution corresponding to a set of weights, which is calculated based on a number of dose points selected from different volumes, such as a target volume (e.g., the designated target region to be treated with radiation, such as a tumor), a ring volume (e.g., a margin region around the target volume where a rapid fall-off of the radiation dose is desired to spare heathy tissue close to the target), a low-dose volume (e.g., a region surrounding the margin again, used as a proxy to limit the volume of healthy tissue further away from the target receiving at least a given fraction (e.g., half) of the prescription dose), and an organs at risk (OAR) volume (e.g., regions which are to be spared from radiation as much as possible). The approximate, geometric description of the low-dose volume used in the first pass is not sufficiently accurate to encompass the relevant region around where the optimized dose falls to below the given fraction of the prescription dose, for all targets and weight combinations. Thus, a second pass, where new low-dose points are selected from the low-dose volume obtained after the first pass, is necessary to achieve good control over the low-dose volume. This second pass requires additional computation time and processing resources. An exemplary two-pass optimization process is described in publication US 2021/158929 A1. OVERVIEW The invention is defined by the claims. Various embodiments, methods, systems, and computer-readable mediums are provided for the generation of radiotherapy plans, which provide an optimized process of calculation of radiotherapy dose distributions, using parallel processing computing hardware. This optimized process may be used for calculating dosage with the parallel processing computing hardware during multiple iterations or passes, where a first pass is used to compute an dose distribution corresponding to a set of weights, such that the first pass is calculated based on a first set of dose points, and where a second pass is used to calculate a more accurate plan based on new dose points obtained from the solution of produced by the first pass (such as new low-dose points selected from a "low-dose" region of the first optimization receiving a given fraction of the prescription dose). The following enables the use of radiotherapy treatment planning optimization using a specialized configuration of a linear programming (LP) solver on parallel processing hardware, to optimize, in parallel, multiple treatment plans where the optimizati