3D in-system calibration method for PET detectors | JoVE Visualize

Area of Science:

Medical Imaging Physics
Nuclear Instrumentation
Machine Learning in Science

Background:

Light-sharing detectors, particularly (semi-)monoliths, offer excellent 2D positioning, energy, and timing resolution for PET systems.
These detectors provide intrinsic depth-of-interaction (DOI) for homogeneous spatial resolution across the field of view (FoV).
Complex positioning calibration procedures currently limit the widespread adoption of these advanced detectors in large-scale clinical PET scanners.

Purpose of the Study:

To introduce a novel 3D in-system calibration method for efficient and convenient recalibration and quality control of assembled PET scanners.
This method is designed for all PET detector types requiring individual calibration, including complex segmented designs.
To evaluate and compare the proposed in-system calibration against a state-of-the-art fan-beam calibration and assess its applicability to various scanner geometries through simulations.

Main Methods:

A proof-of-concept (PoC) scanner with a 120 mm inner diameter and 150 mm axial extent, featuring five finely segmented slab detectors, was utilized.
A 22Na point source was employed, and virtual collimation with near-perpendicular gamma rays was used to train a 2D positioning model via gradient tree boosting (GTB).
Oblique ray data was acquired for angular DOI calibration, calculating DOI geometrically from the ray path to establish 3D training data.

Main Results:

The in-system method demonstrated comparable positioning performance to fan-beam collimator results, with a mean absolute error (MAE) of 0.8 mm and 1.19 mm full-width at half maximum (FWHM).
Depth-of-interaction (DOI) performance reached approximately 90% with an MAE of 1.13 mm and FWHM of 2.47 mm, closely matching fan-beam collimator results.
Analytical calculations indicate that this method's performance is expected to improve in larger scanner geometries.

Conclusions:

The 3D in-system positioning calibration method was successfully validated in a PoC PET scanner, showing performance on par with bench-top fan-beam calibration.
This technique facilitates the calibration and testing of fully assembled PET systems, paving the way for more intricate light-sharing detector architectures in clinical applications.
The collected data can be further leveraged for advanced energy and timing calibration procedures.