Fast image reconstruction for Compton camera using stochastic origin ensemble approach.

Area of Science:

• Physics
• Computer Science
• Medical Imaging

Background:

• Compton cameras are valuable tools in various fields, including astronomy, industry, homeland security, and medical diagnostics.
• Standard image reconstruction techniques like filtered backprojection and maximum likelihood-expectation maximization (ML-EM) are often slow and ineffective for complex Compton camera data.

Purpose of the Study:

• To demonstrate a fast image reconstruction method for Compton cameras using a novel stochastic origin ensembles (SOE) approach.
• To address the limitations of existing reconstruction techniques in terms of speed and effectiveness for distributed sources.

Main Methods:

• The SOE algorithm stochastically assigns event origins to conical surfaces, analogous to lines-of-response in PET.
• Event origins are iteratively moved based on a predefined acceptance probability proportional to changes in event density.
• The reconstructed image emerges as the distribution of origins converges to a quasistationary state.

Main Results:

• The postfiltered SOE algorithm achieves comparable image quality to list-mode ML-EM.
• SOE significantly outperforms ML-EM in terms of reconstruction time, especially for simple Compton camera geometries.
• The computational advantage of SOE is expected to increase with more complex system models and geometries.

Conclusions:

• A novel SOE image reconstruction algorithm based on Markov chains has been successfully implemented and tested for Compton cameras.
• The SOE algorithm is suitable for list-mode data, parallelizable, and adaptable to various Compton camera designs.
• While SOE shows promise for faster reconstruction, future work will focus on improving intensity value accuracy and reducing image variance.