Diffusion Correction in Fricke Hydrogel Dosimeters: A Deep Learning Approach with 2D and 3D Physics-Informed Neural...

Area of Science:

• Radiation dosimetry
• Hydrogel science
• Computational physics

Background:

• Fricke gel dosimeters are tissue-equivalent hydrogels used for radiation dose mapping.
• Ferric ion diffusion in gels blurs dose distributions, limiting measurement accuracy.
• Accurate spatial dose measurement is critical in radiation therapy and research.

Purpose of the Study:

• To develop an AI-driven computational method to mitigate ferric ion diffusion artifacts in Fricke gel dosimeters.
• To reconstruct accurate 2D and 3D radiation dose distributions from diffused ferric ion data.
• To enhance the precision and applicability of Fricke gel dosimetry.

Main Methods:

• Development of Physics Informed Neural Networks (PINNs) to model and reverse ion diffusion.
• Utilizing deep learning to solve diffusion-related partial differential equations in 2D and 3D.
• Inputting diffused ion distributions, boundary conditions, and diffusion coefficients to reconstruct original dose maps.

Main Results:

• The PINN model accurately reconstructed 2D and 3D dose distributions, showing significant agreement with simulated data.
• Mean square errors ranged from 1×10⁻⁶ to 1×10⁻⁴.
• Gamma analysis achieved 90-100% passing rates (3%/2 mm), demonstrating high prediction fidelity.

Conclusions:

• The AI-based approach effectively overcomes ferric ion diffusion limitations in Fricke gel dosimetry.
• This method enhances the accuracy of spatial dose measurements, enabling wider applications.
• The technique promises improved volumetric dose analysis and radiation therapy precision.