optiGAN: a deep learning-based alternative to optical photon tracking in Python-based GATE (10+).

Area of Science:

• Medical Physics
• Computational Science
• Deep Learning

Background:

• Optical Monte Carlo methods are accurate but computationally expensive for photon transport simulations.
• Accelerating these simulations is crucial for advancing medical imaging and detector design.
• The GATE simulation framework is a key tool in medical physics research.

Purpose of the Study:

• To accelerate optical photon transport simulations in the GATE medical physics framework.
• To integrate a generative adversarial network (GAN), named optiGAN, into the new Python-based GATE 10.
• To ensure high modeling accuracy while reducing computational cost.

Main Methods:

• Integrated optiGAN, a GAN model, into GATE 10.
• Validated GATE 10 optical photon transport modules against GATE v9.3.
• Compared full Monte Carlo simulations in GATE 10 with GATE 10-optiGAN simulations.

Main Results:

• GATE 10 results were consistent with GATE v9.3.
• GATE 10-optiGAN simulations showed over 92% similarity to Monte Carlo results.
• Simulation time was reduced by approximately 50% using GATE 10-optiGAN.

Conclusions:

• Confirmed fidelity of optical photon transport modeling in GATE 10.
• Demonstrated effective deep learning-based acceleration via optiGAN.
• Enabled large-scale, high-fidelity optical simulations with reduced computational cost for medical imaging.