Tetris-inspired detector with neural network for radiation mapping.

Area of Science:

• Nuclear physics and instrumentation
• Environmental science and monitoring
• Computer science and machine learning

Background:

• Radiation mapping is crucial for environmental monitoring but faces challenges in performance and cost.
• Existing radiation detectors struggle with complex radiation-matter interactions and data limitations.
• Developing high-resolution, low-cost radiation mapping solutions remains a significant research objective.

Purpose of the Study:

• To present a novel radiation mapping framework utilizing Tetris-inspired detector pixels.
• To demonstrate high-resolution directional prediction and accurate radiation source localization.
• To offer a practical and cost-effective solution for real-world radiation detection applications.

Main Methods:

• Development of a radiation mapping framework with Tetris-inspired detector pixels.
• Application of inter-pixel padding to enhance contrast and neural networks trained with Monte Carlo (MC) simulation data for directional prediction.
• Utilizing a moving detector with Maximum a Posteriori (MAP) for radiation position localization.

Main Results:

• A detector with only four pixels achieved high-resolution directional prediction.
• The Maximum a Posteriori (MAP) method enabled accurate radiation source localization.
• Field testing validated the effectiveness of the MAP method for source localization.

Conclusions:

• The proposed framework enables high-quality radiation mapping with simple detector configurations.
• This approach offers a promising avenue for cost-effective and high-resolution radiation detection.
• The system is anticipated for deployment in real-world environmental monitoring and radiation safety applications.