Author: Keisuke Fujii, Masahiro Kitagawa, Arezoo Modiri, Yuichiro Nakano, Ken N. Okada, Robabeh Rahimi, Akira SaiToh, Amit Sawant, Satoyuki Tsukano, Baoshe Zhang 👨🔬
Affiliation: University of Maryland, University of Maryland in Baltimore, Department of Computer and Information Sciences, Sojo University, Center for Quantum Information and Quantum Biology, Osaka University, Maryland University Baltimore, Department of Radiation Oncology, University of Maryland School of Medicine 🌍
Purpose: Fully personalized radiotherapy requires computational resources far exceeding those of conventional CPU/GPU systems. This study explores the use of quantum computing (QC) in radiotherapy planning, both in simulation and on actual quantum hardware, for a simplified, proof-of-concept prostate cancer scenario.
Methods: An objective function used in current treatment planning was converted into binary format, and subsequently to an Ising Hamiltonian for annealing model QC. The Quantum Approximate Optimization Algorithm (QAOA) was applied to the Hamiltonian with a resulting quantum circuit usable for a circuit model QC. In this proof-of-concept study, we considered a bilateral beam plan with one target and one organ at risk (OAR). The target was represented by one volume unit and the OAR by two. For each volume unit, a two-dimensional threshold-dependent high/low level dose was assumed, thus making six qubits sufficient for the implementation. If exact doses were to be considered, qudits (higher dimensional qubits) would be required, which would be out of scope for this proof-of-principle study. The job was submitted to IBM Quantum for simulation and real QC run, and to the D-Wave machine for comparison purposes.
Results: Our quantum circuit was successfully tested using 6 qubits on IBM Quantum both in simulation and on real QC hardware. Validation was performed using the D-Wave machine: For our Ising model Hamiltonian, the optimization returned the ground state, which dosimetrically meant 1.45Gy (assumed prescribed dose) to the target with lower than threshold dose to the OAR confirming the IBM Quantum results.
Conclusion: This study demonstrates the feasibility of using QC for treatment plan optimization in radiotherapy and provides a detailed followable/repeatable framework. The goal beyond current scope is to enable personalized treatment without concerns about computational power limitations.