Encrypted login | home

Program Information

Higher Resolution Radioluminescence Microscopy Image Reconstruction Via Ionization Trajectory Analysis

no image available
S Almasi

S Almasi*, G Pratx , Q Wang , Stanford University, Palo Alto, CA


TH-AB-708-6 (Thursday, August 3, 2017) 7:30 AM - 9:30 AM Room: 708

Purpose: Radioluminescence microscopy is a method for visualizing fluorescent and radionuclide signals in living cells. The goal of this work is to develop a novel reconstruction method that enhances spatial resolution and sensitivity while reducing user input.

Methods: The reconstruction pipeline starts with the automated segmentation of the ionization tracks in 2D scintillation frames using a method that considers the statistical properties of noise inherent to this type of imaging. The segmented tracks are then skeletonized to localize the ionization source points. Given that each track has two endpoints, the source endpoint is selected by clustering features measuring local intensity entropy, which indicates the proximity of cell clusters. By extrapolating the track beyond its endpoints, we compensate for the travel of the beta particle outside the scintillator plate. Aggregating the final reconstructed source points yields the reconstructed image, which can be filtered to achieve a desired level of noise and spatial resolution.

Results: Implementation of the new method on real and synthetic datasets is shown to provide a radioluminescence reconstruction of cell clusters with 55% higher SNR and 18% better spatial resolution. The performance of sequential steps in this process is assessed using a range of metrics including the dice coefficient for segmentation, detection sensitivity for source point detection, and SNR. The results confirm the efficiency of our method, and show that it outperforms the performance of the previous algorithm (ORBIT 1.11) without requiring any user interaction for reproducible results. Also, consistent performance is achieved for various datasets, demonstrating the robustness of the reconstruction.

Conclusion: We present a reconstruction method for radioluminescence microscopy that yields higher quality images through analysis of the scintillation frames. Images obtained this way have higher resolution and SNR without requiring hardware modification or manual interaction, which greatly improves the efficiency of radioluminescence microscopy.

Contact Email: