Nuclear Medicine Tomographic Imaging Technology Development

Single Photon Emission Computed Tomography (SPECT) is a widely used nuclear imaging technique that visualizes physiological processes by detecting gamma rays emitted from radiotracers, aiding in the diagnosis of diseases such as cancer, heart disease, and neurological disorders. Traditional SPECT systems rely on mechanical collimators, which inherently trade off imaging resolution (clarity of the image) for sensitivity (ability to detect emitted photons), limiting their performance. Our team has pioneered a transformative design called Self-Collimation SPECT (SC-SPECT), which replaces conventional collimators with a novel Multi-layer Interspaced Mosaic Detectors (MATRICES) architecture. In SC-SPECT, front-layer detectors also serve as collimators for rear-layer detectors, while a tungsten plate with a high fraction of apertures provides collimation for the front detectors. This innovative setup eliminates the resolution-sensitivity trade-off, achieving unprecedented performance improvements. Our preliminary work suggests SC-SPECT could revolutionize SPECT imaging, leveraging the established infrastructure of SPECT tracers with long half-lives and robust manufacturing.

However, the complex geometry of the MATRICES architecture and the multi-stage collimation in SC-SPECT introduce unique challenges. Existing models for conventional SPECT, which rely on simple geometric assumptions, are inadequate for characterizing SC-SPECT’s intricate system response. The projection probability density functions (PPDFs), which describe how gamma rays are detected, exhibit significant variations in shape, width, and intensity due to the complex detector arrangement. Additionally, SC-SPECT generates multiplexed data—overlapping signals that enhance sensitivity but risk introducing artifacts if not properly managed. Current evaluation methods, such as Fisher Information Matrix analysis, are computationally infeasible due to the large system response matrix and complicate optimization by involving reconstruction algorithms. 

To address these challenges, we propose developing three novel voxel-wise metrics: the Angular Sampling Completeness Index (ASCI), Multiplex Index (MPXI), and Projection Probability Density Sensitivity (PPDS). These metrics will provide a detailed, voxel-by-voxel assessment of SC-SPECT’s imaging performance, enabling optimized system design. Preliminary results show a strong correlation between ASCI and imaging resolution, indicating their potential to guide design improvements.

By the end of the project, students will be able to:

1. Explain the working principles of nuclear-medicine SPECT, including detector physics, the resolution–sensitivity trade-off, and how Self-Collimation SPECT (SC-SPECT) mitigates it.
2. Implement features in one of the core elements of an imaging pipeline: YAML geometry specs, emission phantom generation, system-response (PPDF) computation, emulated acquisitions, MLEM/OSEM reconstruction, and image-quality evaluation.
3. Compute and interpret voxel-wise design metrics—Angular Sampling Completeness (ASCI), projection spread/intensity, and related sensitivity measures—to compare digital-twin configurations.
4. Use high-performance computing effectively: submit, monitor, and optimize jobs on UB’s CCR cluster (or GPU array); profile runtime/memory; and scale experiments reproducibly.
5. Apply modern software practices (version control, unit tests, parameterized runs, and clear documentation) to produce reusable, benchmarked code and datasets.
6. Analyze and visualize results with clear figures/tables; quantify uncertainty and articulate limits of models and simulations.
7. Communicate findings in a short technical report and a conference-style poster, demonstrating audience-appropriate writing, data presentation, and Q&A skills.
8. Reflect on research ethics and rigor, including data management, citation, and reproducibility, and identify next steps for improving SC-SPECT design.

Please read the articles linked below: 

medical imaging technology, computer science, engineering, medical physics, nuclear medicine, python, linux