Online program Overview Session: NSS-15

Abstract

To ensure the safe decommissioning of the reactor buildings at the Fukushima Daiichi Nuclear Power Station, it is essential to reduce the external exposure of workers. In order to achieve this objective, it is necessary to accurately estimate the distribution of  contamination density — the primary source of external exposure — and to evaluate dose rates in each working space within R/Bs with high precision. 3D Air Dose Rate Evaluation System -Indoor (3D-AI) estimates distributions of contamination density and dose rate in the working space using dose-rate data and point clouds via LASSO regression. In this work, we introduce an L2-normalization method that balances each source's contribution to measurement points to improve sensitivity to remote contaminations. To test the ability of this method, we analyzed data measured near the startup transformer of the Fukushima Daiichi Nuclear Power Station Unit 3 and compared them with maps estimated by the Compton camera. Results show that both conventional and new systems estimated contamination distributions and dose rates with high accuracy near the measurement points. In contrast, in areas distant from measurement points, the new method better matches the Compton images and improves surface dose-rate estimates, reducing the ~1/50 underestimation to within a few factors.

Abstract

This paper describes a template-based gamma spectral decomposition algorithm developed for quantitative analysis of Naturally Occurring Radioactive Material (NORM). The method uses a Monte Carlo-simulated template to decompose measured spectra into contributions from the 40K, and the 238U and 232Th decay chains, and to estimate the corresponding absorbed dose rates. The algorithm is integrated into the CAEN Gamon-Drone platform and supports both CeBr₃ and NaI(Tl) scintillation detectors. The proposed approach addresses two operational scenarios. In airborne surveys, dynamic integration and spectral reweighting are applied to continuously acquired spectra to reduce statistical fluctuations. This processing improves detection sensitivity and allows for high-resolution radiological mapping. In ground-based measurements, the same analytical framework provides rapid in situ characterization of soils, rocks, and construction materials through direct estimation of radionuclide concentrations. Validation tests showed strong consistency between reconstructed activities and reference values for multiple detector configurations and environmental conditions. The results demonstrate that Monte Carlo template decomposition can provide accurate and robust NORM quantification using compact scintillation-based gamma spectroscopy systems.

Abstract

Multi-Voltage Threshold (MVT) technology, serving as an advanced method for scintillation pulse digitization, achieves pulse waveform reconstruction and energy calculation in PET systems through a bi-exponential model. However, with the gradual application of crystal miniaturization and 1:1:1 readout technologies, the fitting of low-energy pulses faces severe convergence challenges. This leads to a large number of pulses being discarded, which severely restricts system sensitivity. To address this issue, this paper proposes an effective initial value estimation method. Experiments were conducted using four-threshold MVT sampling and bi-exponential fitting, utilizing the Root Mean Square Error (RMSE) of the fitted waveforms, the proportion of severe deviations, the absolute mean of the pulse area relative error, and the standard deviation as evaluation metrics. Results on low-energy band pulses (40-150 keV) indicate that after utilizing the proposed initial value estimation method, the waveform RMSE was effectively restricted, and the three relative area error metrics were improved by 61.5%, 105.8, and 341.0, respectively, thereby significantly reducing fitting convergence failures.

Abstract

  Muon scattering imaging (MSI) is sensitive to high-Z materials; however, the inherently low cosmic muon flux usually requires long acquisition times to achieve acceptable image quality. To address this limitation, we propose a statistically enhanced dual-modality muon imaging framework based on trigger redesign in a muon scattering tomography system consisting of 8 detector layers. By relaxing the conventional strict coincidence requirement and redesigning the event selection logic, previously discarded non-penetrating muon events can be effectively incorporated into muon absorption imaging (MAI), while penetrating events are retained for MSI reconstruction. This event reutilization significantly increases the effective statistics without additional hardware, thereby improving imaging efficiency and reducing the required measurement time.

  To further exploit the complementary information from MAI and MSI, multiple image fusion methods were investigated, including linear weighted fusion, Fourier decomposition filtering fusion, and Principal Component Analysis-based fusion. Experimental results show that the maximum a posteriori (MAP)-based fusion method achieves the best reconstruction performance. Imaging experiments were conducted using an eight-layer 50 cm × 50 cm plastic scintillator detector system. The target consisted of fifteen 5 cm × 5 cm × 5 cm lead blocks arranged in a U-shaped structure. A total of 4.3 × 10⁴ muon events were collected. Compared with conventional MSI reconstruction, the proposed framework produces clearer structural boundaries, higher contrast, and significantly suppressed background noise. Quantitative evaluation using the structural similarity index (SSIM) shows an improvement from 0.212 to 0.389, corresponding to an increase of approximately 83%.

  The proposed trigger optimization and image fusion strategy enable faster, higher-quality, and more practical muon scattering imaging for security and high-Z material identification.

Abstract

This work investigates Inverse Compton Scattering (ICS) sources and their potential for synchrotron-based medical applications, particularly Microbeam Radiation Therapy (MRT). As compact alternatives to large synchrotron facilities, ICS sources are rapidly approaching the flux and energy requirements necessary for MRT, but require detailed characterisation before clinical implementation can be considered. Geant4 Monte Carlo simulations inspired by the ThomX ICS source were performed to evaluate the influence of key source parameters on the emitted x-ray spectrum. Parameters studied included electron beam energy, energy spread, emittance, and laser wavelength. The simulations show that variations in these parameters produce measurable shifts in spectral peak energy, directly relevant to MRT beam optimisation, as well as asymmetric spectral broadening across the spectrum. These results are the first ICS simulations of these parameters at MRT relevant energies and aid the evaluation of ICS sources as compact alternatives to synchrotron radiation sources.

Abstract

Radiomics analysis in nuclear medicine is commonly performed on predefined features extracted from individual regions, which limits its ability to describe global imaging patterns across the whole body. In this work, we propose a latent-space digital twin framework to characterize multi-organ radiomics behavior in theranostic SPECT/CT imaging. Post-therapy SPECT/CT data from 57 patients treated with ¹⁷⁷Lu-based radiopharmaceuticals were analyzed. Multi-organ and lesion segmentations were obtained, and 107 standardized radiomics features were extracted for each region. These features were aggregated into patient-level representations and projected into a low-dimensional latent space using principal component analysis. Unsupervised clustering and distance-based analysis were then applied to identify underlying radiomics phenotypes and quantify inter-patient variability. The learned latent space revealed structured organization of radiomics patterns, with the first two principal components capturing more than half of the total variance (PC1: 34%, PC2: 19%). Distinct patient groups emerged with clear separation, characterized by lower intra-group distances (≈6) and substantially higher inter-group distances (≈12). Analysis of latent components indicated that structural regions predominantly contributed to the first component, while soft-tissue and vascular regions influenced the second component. In addition, the latent representation enabled identification of outlier patients, reflecting substantial heterogeneity in multi-organ radiomics profiles. These findings demonstrate that latent-space modeling provides a compact and interpretable representation of complex radiomics data. The proposed digital twin framework enables system-level phenotyping and may support improved patient stratification in theranostic SPECT/CT imaging.