Online program Overview Session: RTSD-03

Abstract

Fully spectroscopic photon-counting CT enables material identification by recording the energy-dependent transmission spectrum of an object. Timepix3 (TPX3) detectors extend conventional threshold-based photon counting CT by recording position, time-of-arrival and energy information for individual events, allowing for post-acquisition energy binning and extraction of spectral features such as K-edges. However, fully spectroscopic photon counting CT with TPX3 detectors remains limited by detector area, photon statistics, scan time and raw data volume.

Previous work introduced a two-stage TPX3 acquisition scheme that combines frame-based and event-based readout. Here, we investigate whether the frame-based TPX3 intensity image can be replaced by a geometrically matched flat-panel intensity measurement. The CMOS/GdOS flat-panel detector provides full-field intensity projections, while a 12 x 1 TPX3 array with 1 mm CdTe sensors provides fully spectroscopic region-of-interest (ROI) information. The method combines the low-noise intensity information of the flat-panel detector with the material-specific spectroscopic information of TPX3. Projection geometry is matched by adapting cone-beam magnification, followed by empirical intensity mapping in the overlap region. The approach is benchmarked against the original TPX3-MPX-normalized acquisition reference. Initial results indicate that replacing the TPX3 intensity image with flat-panel projections preserves TPX3-derived material characteristics in the investigated ROI and enables material labels to be extrapolated to repeated structures in the full-field reconstruction.

Abstract

Cherenkov Charge Induction (CCI) detectors consist of transparent semiconductor detectors that combine the conventional charge induction readout with the detection of Cherenkov light generated by gammas depositing energy above the Cherenkov threshold. CCI detectors show strong potential for applications requiring time-resolved multi-MeV gamma detection, such as proton range verification in proton therapy.
In this work we report on a pixelated CCI TlBr detector consisting of a 14×14×11 mm3 TlBr crystal with a 4×4 pixelated anode matrix of 2.3×2.3 mm2 pixels with 2.5 mm pitch. The lateral faces of the crystal were polished to maximize light extraction, and a 2×2 silicon photomultiplier (SiPM) array with 4×4 mm2 active area per channel was used for Cherenkov light detection. The detector was instrumented with a multi- channel acquisition system based on a CAEN R5560 open FPGA digitizer, enabling simultaneous acquisition of charge induction and Cherenkov light signals from all detector channels. The acquisition was triggered by the Cherenkov readout and a 22Na point source was used for detector characterization.
First results demonstrate the feasibility of simultaneous Cherenkov light and charge induction acquisition using a multi-channel readout, with the optical signal successfully providing the acquisition trigger and a timing reference to estimate the electron drift time. Future work will focus on optimization of the triggering and synchronization scheme, waveform processing methods, and detector operating conditions to improve event selection and detector characterization.

Abstract

Cadmium Zinc Telluride (CZT) detectors are widely used for X-ray and γ-ray detection due to their high atomic number, wide band gap, and excellent room-temperature performance. Their response is governed by charge transport processes, including carrier mobility, trapping, diffusion, and electric-field nonuniformities, which are critical in pixelated geometries where charge sharing and weighting-field effects dominate signal formation. In this work, a 3D CZT detector model is developed in COMSOL Multiphysics, solving semiconductor transport equations self-consistently without Gaussian charge-cloud approximations. Signal induction is computed using the Ramo–Shockley theorem applied to the full carrier distribution. Validation against synchrotron measurements shows excellent agreement in normalized charge collection profiles (r = 0.98), confirming accurate reproduction of charge transport and sharing effects.

Abstract

Depth-of-interaction (DOI) estimation in CdZnTe detectors is conventionally based on the cathode-to-anode amplitude ratio (C/A), while timing-based methods have been explored as alternative indicators in pixelated geometries. However, DOI characterization in cross-strip CdZnTe detectors using digitized waveform signal parameters, including drift time, cathode amplitude, and anode rise time, and the degree to which these parameters provide independent depth information has not been systematically studied. In this work, we present a waveform-based DOI characterization of a 5 mm thick cross-strip CdZnTe detector (7 anode strips, 8 cathode strips, and 1 steering electrode). A collimated ¹³⁷Cs slit beam was scanned across 26 discrete depth positions 0.2 mm apart. Drift time and cathode amplitude exhibit a strong correlation (R² = 0.99), establishing that both contribute to the practical lower bound on the achievable single-event depth resolution. At each collimator slit position, the measured signal distributions reflect the distribution of interaction depths, due to finite beam width and intracrystal scatter. The C/A ratio computed from the two highest-amplitude cathode strips provides the best trade-off between DOI resolution and measurement consistency. Using a leave-one-step-out cross- validation framework, we obtain a DOI resolution of 1.35 mm RMSE with single-feature calibration and 1.12 mm RMSE using a 9-feature random forest regression model.

Abstract

An ultra-high resolution CZT gamma camera platform is presented, integrating advanced detector readout, machine learning (ML) based signal processing and practical calibration procedures for sub‑pixel 3D interaction localisation. The system employs Kromek’s D‑Matrix readout with 2 mm pixel pitch detectors and utilises an ML model trained on physics‑based simulations and calibrated with experimental flood‑field data. Recent developments focus on improving calibration flexibility, robustness to detector performance fluctuations and stability during prolonged operation through refined detector simulations, multi-model optimisation strategies and incorporation of measured data in the training process.

Performance evaluation demonstrates improved spatial uniformity and significant suppression of inter‑pixel artefacts in uncorrected flood field images, with stable behaviour maintained over extended acquisition periods without recalibration. Previously reported intrinsic spatial resolution (~250 µm) and depth‑of‑interaction accuracy (~0.5–1 mm) are preserved, with improved consistency across the detector volume. Reduced reliance on flat‑field correction limits amplification of Poisson noise in reconstructed images, enhancing effective image quality. The ML predicted energy incorporates corrections that effectively eliminate energy DOI dependence and charge sharing losses, resulting in a strongly reduced low-energy tail of the spectrum.

Abstract

Hybrid semiconductor detectors based on the TimePIX-family read-out chips enable direct detection of ionizing radiation with high spatial and spectroscopic resolution. This work investigates the application of femtosecond laser ablation for precise structural modification of detector components while minimizing thermal damage and preserving detector functionality. The study focused on three-dimensional shaping of the read-out chip in a MiniPIX TimePIX3 detector with a 1 mm thick cadmium telluride sensor and on edge cutting of silicon sensors. An optimized laser ablation process enables significant thinning of the read-out chip while maintaining full detector operation. The modified configuration allows backside illumination of the detector, resulting in reduced X-ray attenuation and improved detection efficiency for low-energy photons in the energy range from 5 to 25 keV. Experimental measurements also demonstrate improved spectroscopic performance and enhanced spatial resolution. In addition, femtosecond laser ablation was applied to precise cutting of silicon sensor edges, enabling efficient integration of individual modules into large-area monolithic detection systems. The presented results demonstrate that femtosecond laser ablation is a promising tool for advanced modification of hybrid semiconductor detectors and can significantly expand their applicability in future radiation detection systems.

Abstract

High resolution detection modules for multi-isotope SPECT imaging are strongly required in nuclear medicine. In this framework, we proposed and developed new cadmium–zinc–telluride (CZT) drift strip detectors for room temperature gamma-ray spectroscopic imaging. As widely demonstrated, CdZnTe (CZT) is one of the key materials for the development of room temperature X-ray and gamma ray detectors and several research groups are still involved in advances in both device and the crystal growth technologies. The CZT detectors, equipped with orthogonal anode/cathode collecting strips, drift strips and dedicated digital pulse processing allow a detection area of 6 x 20 mm², excellent room temperature energy resolution of 0.8% FWHM at 662 keV and sub-millimetre spatial resolution.  The timing performance of the detectors will be also presented. These activities are within an Italian PRIN 2022 PNRR project (No. P2022SE49Y; CUP: B53D23024100001), funded by the Italian Ministry for University and Research (MUR).

Abstract

High-resolution Cd(Zn)Te radiation detectors are essential for quantitative imaging in medical diagnostics, high-energy physics, and industrial non-destructive testing [1]. This study examines the impact of charge sharing (CS) [2] in a 2-mm CdTe/Timepix detector with 55 μm pixels [3], and proposes a spectroscopic correction method.

From Am-241 and Co-57 experimental data, we categorized interaction events by their multiplicity and a newly introduced shape descriptor called circularity. Analysis showed that non-circular events (C < 0.9) resulted in energy peaks shifting toward lower energies due to unrecorded below-threshold pixels. To address these high-multiplicity CS events, we developed a multi-stage spectroscopic correction method that corrects spectra based on circularity and multiplicity values. This approach improved the average FWHM for the Am-241 peak from 16.70% to 10.05% and for Co-57 from 9.64% to 6.73%.

Additionally, AllPix-Squared simulations [4] were used to model charge cloud dynamics and quantify losses occurring during both propagation and digitization. This work enabled a more robust energy calibration and unbiased spectroscopic performance for small-pixel semiconductor detectors.