Optimization apparatus, method implemented by optimization apparatus, and non-transitory computer-readable storage medium for storing program

A method implemented by an optimization apparatus, the optimization apparatus comprises a processor coupled to an Ising machine, the method comprising:

• Receiving, by the processor, an input information indicating a first cost function regarding a radiation irradiation plan in a medical field, the first cost function being a numerical expression including a part being a continuous function and a part being a discontinuous function, the discontinuous function including an organ-dependent irradiation limit in the radiation irradiation plan;
• Transforming, by the processor, the input first cost function indicated by the received input information to a second cost function by using at least “Legendre transformation” and “Wolfe's duality theorem,” in order to obtain the second cost function without changing a graph shape of the discontinuous function included in the first cost function;
• Acquiring, by the processor, first values being values in an Ising format by performing binary expansion on values relating to coefficients of the second cost function obtained by the transforming from the first cost function;
• Outputting, by the processor, the first values to the Ising machine, and causing the Ising machine to perform a simulated annealing (SA) on the second cost function to seek values that represent a ground state regarding the first values in the Ising format relating to the coefficients of the second cost function;
• receiving, from the Ising machine, a result including the sought values that represent the ground state;
• in response to the receiving of the result from the Ising machine, executing, by the processor, reverse binary expansion processing that includes acquiring values that minimize the first cost function by carrying out reverse binary expansion for the sought values that represent the ground state;
• and
• in response to the acquiring of the second values, output, by the processor, the radiation irradiation plan optimized based on the acquired second values, the outputted radiation irradiation plan being a radiation irradiation plan to be used in a radiation therapy to irradiate a tumor of a patient while reducing irradiation of normal tissue of the patient,
• wherein the transforming of the first cost function is configured to:
• perform the Legendre transformation twice on the part being the discontinuous function included in the first cost function;
• and
• perform transformation to the second cost function from the first cost function by replacing the discontinuous function included in the first cost function to a first function, the first function being a function obtained through finding a Wolfe dual problem with respect to the first cost function expression that has undergone the Legendre transformation twice, the Wolfe dual problem being a dual problem of “Wolfe's duality theorem”.

A non-transitory computer-readable storage medium for storing a program which causes a processor to perform processing, the processor coupled to an Ising machine, the processing comprising:

• Receiving, by the processor, an input information indicating a first cost function regarding a radiation irradiation plan in a medical field, the first cost function being a numerical expression including a part being a continuous function and a part being a discontinuous function, the discontinuous function including an organ-dependent irradiation limit in the radiation irradiation plan;
• Transforming, by the processor, the first cost function indicated by the received input information to a second cost function by using at least “Legendre transformation” and “Wolfe's duality theorem,” in order to obtain the second cost function without changing a graph shape of the discontinuous function included in the first cost function;
• Acquiring, by the processor, first values being values in an Ising format by performing binary expansion on values relating to coefficients of the second cost function obtained by the transforming from the first cost function;
• Outputting, by the processor, the first values to the Ising machine, and causing the Ising machine to perform a simulated annealing (SA) on the second cost function to seek values that represent a ground state regarding the first values in the Ising format relating to the coefficients of the second cost function;
• receiving, from the Ising machine, a result including the sought values that represent the ground state;
• in response to the receiving of the result from the Ising machine, executing reverse binary expansion processing that includes acquiring second values being values that minimize the first cost function by carrying out reverse binary expansion for the sought values that represent the ground state;
• and
• in response to the acquiring of the second values, output the radiation irradiation plan optimized based on the acquired second values, the outputted radiation irradiation plan being a radiation irradiation plan to be used in a radiation therapy to irradiate a tumor of a patient while reducing irradiation of normal tissue of the patient,
• wherein the transforming of the first cost function is configured to:
• perform the Legendre transformation twice on the part being the discontinuous function included in the first cost function;
• and
• perform transformation to the second cost function from the first cost function by replacing the discontinuous function included in the first cost function to a first function, the first function being a function obtained through finding a Wolfe dual problem with respect to the first cost function expression that has undergone the Legendre transformation twice, the Wolfe dual problem being a dual problem of “Wolfe's duality theorem”.