[Paper Review] Evaluation of Compton scattering sequence reconstruction algorithms for a portable position sensitive radioactivity detector based on pixelated Cd(Zn)Te crystals

K. Karafasoulis, K. Zachariadou+5 more|arXiv (Cornell University)|Nov 11, 2010

Radiation Detection and Scintillator Technologies3 references3 citations

TL;DR

This paper evaluates Compton sequence reconstruction algorithms—Dual Cluster Sequence (DCS) and Multiple Cluster Sequence (MCS)—for a portable Cd(Zn)Te-based Compton camera (COCAE) designed to localize and identify radioactive sources. Using Monte Carlo simulations, the MCS algorithm achieves up to 70% reconstruction efficiency with optimized constraints on quality factor (QF) and lever arm, outperforming DCS-B, which reaches ~90% efficiency for dual-cluster events.

ABSTRACT

We present extensive simulation studies on the performance of algorithms for the Compton sequence reconstruction used for the development of a portable spectroscopic instrument (COCAE), with the capability to localize and identify radioactive sources, by exploiting the Compton scattering imaging. Various Compton Sequence reconstruction algorithms have been compared using a large number of simulated events. These algorithms are based on Compton kinematics, as well as on statistical test criteria that exploit the redundant information of events having two or more photon interactions in the active detector's volume. The efficiency of the best performing technique is estimated for a wide range of incident gamma-ray photons emitted from point-like gamma sources.

Motivation & Objective

• To develop a portable, cryogenics-free spectroscopic detector for accurate localization and identification of radioactive sources.
• To evaluate and compare the performance of Compton sequence reconstruction algorithms in a pixelated Cd(Zn)Te detector system.
• To optimize reconstruction efficiency by leveraging Compton kinematics and statistical criteria using redundant interaction data.
• To assess the impact of energy resolution, Doppler broadening, and detector geometry on sequence reconstruction accuracy.
• To determine optimal thresholds for quality factor (QF) and lever arm to improve reconstruction efficiency without excessive event loss.

Proposed method

• Simulates Compton scattering events in a 10-layer, 2mm-thick pixelated Cd(Zn)Te detector with 400μm pitch using Monte Carlo methods.
• Applies Compton kinematics and geometric reconstruction to estimate incident photon direction from energy depositions and interaction positions.
• Uses a generalized χ²-based quality factor (QF) to evaluate sequence validity: QF = Σ[(cosφ_geo - cosφ_kin)² / (σ_cosφ_geo² + σ_cosφ_kin²)] for each interaction sequence.
• Implements constraints on QF threshold and minimum lever arm (≥2 cm) between first two interactions to improve sequence selection.
• Compares DCS (for two-cluster events) and MCS (for multiple-cluster events) algorithms using simulated photo-peak events across 100–2000 keV.
• Analyzes efficiency, event reduction, and QF distribution to determine optimal reconstruction thresholds.

Experimental results

Research questions

• RQ1 How does the performance of DCS and MCS algorithms compare in reconstructing Compton sequences across varying gamma-ray energies?
• RQ2 What is the impact of QF threshold and lever arm constraints on the reconstruction efficiency and event statistics?
• RQ3 To what extent does Doppler broadening degrade sequence reconstruction performance at low energies?
• RQ4 How do energy resolution and spatial resolution affect the accuracy of Compton angle estimation and sequence selection?
• RQ5 Can statistical quality factors effectively distinguish true from false interaction sequences in multi-cluster events?

Key findings

• The MCS algorithm achieves a maximum reconstruction efficiency of ~55% for multiple-cluster events at energies above 600 keV.
• By applying a QF threshold of 0.01 and requiring a minimum lever arm of 2 cm, MCS efficiency improves to ~70%.
• At 1000 keV, only ~40% of events have a QF value below 0.01, indicating a significant reduction in usable events with stricter thresholds.
• DCS-B and DCS-C algorithms achieve ~50–90% efficiency for dual-cluster events, with DCS-B performing best at higher energies.
• The overall reconstruction efficiency drops at low energies (<300 keV), not due to algorithmic failure but due to masking by Doppler broadening.
• QF distribution shows a clear separation between true sequences and false sequences, validating the statistical approach for sequence selection.