Elisabeth Pfaehler

Objective. Inter-crystal scattering (ICS) degrades spatial resolution and contrast by causing spurious lines of response in small-pitch, staggered three-layer pixelated LSO scintillation detectors used in a high-resolution BrainPET-7T insert. This work aimed to investigate the distribution of ICS events in our BrainPET-7T detector and understand its underlying mechanisms. Approach. Six physical approaches for computing probabilities to identify the first interaction positions of ICS events were studied using Geant4 simulations, focusing specifically on events with two interactions, which represent the most frequent ICS occurrence. Performance was evaluated using ideal interaction positions and energies directly from simulations, ground truth crystal-center positions with ideal energies, as well as crystal-center positions and energies determined by a maximum likelihood positioning (MLP) algorithm. Main results. A method based on Klein-Nishina probabilities, which were computed using geometric information (KN-G) and MLP-derived crystal-center positions and energies, achieved superior accuracy, particularly when the distance between successive interactions in Compton events was shorter than 6 mm and when incident angles were below 15 • . This advantage is partly attributable to crystal-center discretization, which masks a substantial portion of backscattering events. It reached a peak accuracy of 92.2 % at a distance of 19 mm. Significance. These findings demonstrate the feasibility of combining the KN-G method with the MLP algorithm to mitigate the ICS effects, with the potential to improve the image quality of the 7T BrainPET insert.

Artificial intelligence (AI) and machine learning (ML) are reshaping cancer research and care. In pediatric oncology, early evidence—most robust in imaging—suggests value for diagnosis, risk stratification, and assessment of treatment response. Pediatric endocrine tumors are rare and heterogeneous, including intra- and extra-adrenal paraganglioma (PGL), adrenocortical tumors (ACT), differentiated and medullary thyroid carcinoma (DTC/MTC), and gastroenteropancreatic neuroendocrine neoplasms (GEP-NEN). Here, we provide a pediatric-first, entity-structured synthesis of AI/ML applications in endocrine tumors, paired with a methods-for-clinicians primer and a pediatric endocrine tumor guardrails checklist mapped to contemporary reporting/evaluation standards. We also outline a realistic EU-anchored roadmap for translation that leverages existing infrastructures (EXPeRT, ERN PaedCan). We find promising—yet preliminary—signals for early non-remission/recurrence modeling in pediatric DTC and interpretable survival prediction in pediatric ACT. For PGL and GEP-NEN, evidence remains adult-led (biochemical ML screening scores; CT/PET radiomics for metastatic risk or peptide receptor radionuclide therapy response) and serves primarily as methodological scaffolding for pediatrics. Cross-cutting insights include the centrality of calibration and validation hierarchy and the current limits of explainability (radiomics texture semantics; saliency ≠ mechanism). Translation is constrained by small datasets, domain shift across age groups and sites, limited external validation, and evolving regulatory expectations. We close with pragmatic, clinically anchored steps—benchmarks, multi-site pediatric validation, genotype-aware evaluation, and equity monitoring—to accelerate safe, equitable adoption in pediatric endocrine oncology.

Purpose: Pediatric adrenocortical tumors (pACTs) are rare and clinically heterogeneous. Existing risk stratification systems rely on fixed thresholds and linear assumptions, which may limit their prognostic accuracy—particularly for nonmetastatic, locally advanced cases. We aimed to develop an interpretable machine learning (ML) model for individualized survival prediction using only routine clinical features. Methods: We retrospectively analyzed 97 patients with pACT from the German Pediatric Oncology Hematology-Malignant Endocrine Tumors Registry (1997-2024). An Extreme Gradient Boosting Cox proportional hazards model was trained using 4 features—tumor volume, distant metastases, pathologic T stage, and resection status—identified via systematic feature evaluation across 11 737 model combinations. Performance was assessed using a stratified 80/20 train–test split, 500 bootstrap iterations, and Harrell's concordance index (C-index). SHapley Additive exPlanations (SHAP) were used for interpretability. Results: The model achieved strong prognostic performance (test-set C-index: 0.925; bootstrap mean: 0.891, 95% confidence interval: 0.817-0.946). SHAP analysis confirmed the dominant influence of metastatic status, followed by tumor volume, T stage, and resection status. The model uncovered nonlinear and additive effects, including a SHAP- and bootstrap-guided tumor volume cut-off (190 mL, 95% confidence interval 127-910 mL) that only slightly differed from conventional thresholds. Stratification remained robust in subgroups, including nonmetastatic patients with advanced local disease. Conclusion: This interpretable ML model enables individualized survival prediction in pACT using only routine clinical data. It offers a clinically accessible and clinically meaningful complement to existing scoring systems, particularly in patients with ambiguous risk profiles who may benefit from more personalized management.

Objective. In this study, we propose a fast implementation of a maximum likelihood positioning (MLP) algorithm to estimate the energy and identify the active scintillator pixel in staggered layer scintillation detectors for positron emission tomograph (PET). Approach. High processing speed of the MLP algorithm was achieved by developing an iteration-free implementation paired with additional computational optimizations. Further, the MLP was implemented for a staggered layer design with pixelated scintillators which enables the determination of the gamma’s depth of interaction. The pixelation of the scintillator facilitates an iteration-free formulation of the MLP algorithm. The efficacy of the algorithm optimization was tested on a scintillation detector block designed for an ultra-high field BrainPET 7 T, comprising three scintillator pixel layers. Calibration measurements, in combination with an automated calibration script, were used to obtain the expected counts of scintillation photons required in the MLP algorithm. Main results. Using single-instruction-multiple-data parallelization, multi-threading and optimized cache lines, a maximum processing speed of approximately 22.5 million singles per second was achieved on a platform with four Intel Xeon Platinum 8168 CPUs and 60 threads, encompassing all required processing steps. The automatic calibration failed for 1–15 individual scintillator pixels in approximately 10% of the 120 scintillation detector blocks, necessitating manual correction. After applying the energy correction to the positioned single events, an energy resolution of Δ E = 12 % ± 2 % FWHM was obtained for the entire scintillation block. Significance. The obtained energy resolution is very close to the energy resolutions measured for the individual scintillator pixels, proving that the MLP accurately identifies the scintillating pixel and that the energy correction method effectively compensates for the light collection variations of the SiPM array. The presented method is not limited to PET and can be used as fast positioning and energy determination method with scintillation detectors for PET, SPECT and gamma cameras.

Purpose This study investigated whether radiomic features extracted from [¹⁸F]FDG-PET scans acquired before and two weeks after neoadjuvant treatment, and their variation, provided prognostic parameters in locally advanced cervical cancer (LACC) patients treated with neoadjuvant chemo-radiotherapy (CRT) followed by radical surgery. Methods We retrospectively included LACC patients referred to our Institution from 2010 to 2016. [¹⁸F]FDG-PET/CT was performed before neoadjuvant CRT (baseline) and two weeks after the start of treatment (early). Radiomic features were extracted after semi-automatic delineation of the primary tumour, on baseline and early PET images. Delta radiomics were calculated as the relative differences between baseline and early features. We performed 5-fold cross-validation stratified for recurrence and cancer-specific death, integrating dimensionality reduction of the radiomic features and variable hunting with importance within the folds. After supervised feature selection, radiomic models with the best-performing features for each timepoint, as well as clinical models and combined clinico-radiomic models, were built. Model performances are presented as C-indices, for prediction of recurrence/progression (disease-free survival, DFS) and cancer-specific death (overall survival, OS). Results 95 patients were included. With a median follow-up of 76.0 months (95% CI: 59.5–82.1), 31.6% of patients had recurrence/progression and 20.0% died of disease. None of the models could predict DFS (C-indices ≤ 0.72). Model performances for OS yielded slightly better results, with mean C-indices of 0.75 for both the radiomic and combined model based on early features, 0.79 and 0.78 for the radiomic and combined model derived from delta features, and 0.76 for the clinical models. Conclusion [¹⁸F]FDG-PET early and delta radiomic features could not predict DFS in patients with LACC treated with neoadjuvant CRT followed by radical surgery. Although slightly improved performances for the radiomic and combined models were observed in the prediction of OS compared to the clinical model, the added value of these parameters and their inclusion in the clinical practice seems to be limited.

In the acquisition of Magnetic Resonance (MR) images shorter scan times lead to higher image noise. Therefore, automatic image denoising using deep learning methods is of high interest. In this work, we concentrate on image denoising of MR images containing line-like structures such as roots or vessels. In particular, we investigate if the special characteristics of these datasets (connectivity, sparsity) benefit from the use of special loss functions for network training. We hereby translate the Perceptual Loss to 3D data by comparing feature maps of untrained networks in the loss function. We tested the performance of untrained Perceptual Loss (uPL) on 3D image denoising of MR images displaying brain vessels (MR angiograms - MRA) and images of plant roots in soil. In this study, 536 MR images of plant roots in soil and 450 MRA images are included. The plant root dataset is split to 380, 80, and 76 images for training, validation, and testing. The MRA dataset is split to 300, 50, and 100 images for training, validation, and testing. We investigate the impact of various uPL characteristics such as weight initialization, network depth, kernel size, and pooling operations on the results. We tested the performance of the uPL loss on four Rician noise levels (1%, 5%, 10%, and 20%) using evaluation metrics such as the Structural Similarity Index Metric (SSIM). Our results are compared with the frequently used L1 loss for different network architectures. We observe, that our uPL outperforms conventional loss functions such as the L1 loss or a loss based on the Structural Similarity Index Metric (SSIM). For MRA images the uPL leads to SSIM values of 0.93 while L1 and SSIM loss led to SSIM values of 0.81 and 0.88, respectively. The uPL network’s initialization is not important (e.g. for MR root images SSIM differences of 0.01 occur across initializations, while network depth and pooling operations impact denoising performance slightly more (SSIM of 0.83 for 5 convolutional layers and kernel size 3 vs. 0.86 for 5 convolutional layers and kernel size 5 for the root dataset). We also find that small uPL networks led to better or comparable results than using large networks such as VGG (e.g. SSIM values of 0.93 and 0.90 for a small and a VGG19 uPL network in the MRA dataset). In summary, we demonstrate superior performance of our loss for both datasets, all noise levels, and three network architectures. In conclusion, for images containing line-like structures, uPL is an alternative to other loss functions for 3D image denoising. We observe that small uPL networks have better or equal performance than very large network architectures while requiring lower computational costs and should therefore be preferred.

Magnetic Resonance Imaging (MRI) is a powerful imaging technique widely used for visualizing structures within the human body and in other fields such as plant sciences. However, there is a demand to develop fast 3D-MRI reconstruction algorithms to show the fine structure of objects from under-sampled acquisition data, i.e., k-space data. This emphasizes the need for efficient solutions that can handle limited input while maintaining high-quality imaging. In contrast to previous methods only using 2D, we propose a 3D MRI reconstruction method that leverages a regularized 3D diffusion model combined with optimization method. By incorporating diffusion based priors, our method improves image quality, reduces noise, and enhances the overall fidelity of 3D MRI reconstructions. We conduct comprehensive experiments analysis on clinical and plant science MRI datasets. To evaluate the algorithm effectiveness for under-sampled k-space data, we also demonstrate its reconstruction performance with several undersampling patterns, as well as with in- and out-of-distribution pre-trained data. In experiments, we show that our method improves upon tested competitors.

Objective. Conventionally, if two metabolic processes are of interest for image analysis, two separate, sequential positron emission tomography (PET) scans are performed. However, sequential PET scans cannot simultaneously display the metabolic targets. The concurrent study of two simultaneous PET scans could provide new insights into the causes of diseases. Approach. In this work, we propose a reconstruction algorithm for the simultaneous injection of a β + -emitter emitting only annihilation photons and a β + -γ-emitter emitting annihilation photons and an additional prompt γ-photon. As in previous works, the γ-photon is used to identify events originating from the β + -γ-emitter. However, due to e.g. attenuation and down-scatter, the γ-photon is often not detected and not all events can correctly be associated with the β + -γ-emitter as they are detected as double coincidences. In contrast to previous works, we estimate this number of double coincidences with origin in the β + -γ, emitter including the attenuation of the prompt γ, and incorporate this estimation in the forward-projection of the maximum likelihood expectation maximization algorithm. For evaluation, we simulate different scenarios with varying objects and attenuation maps. The nuclide ¹⁸F serves as β + -emitter, while ⁴⁴Sc functions as β + -γ emitter. The performance of the algorithm is assessed by calculating the residual error of the β + -γ-emitter in the reconstructed β + -emitter image. Additionally, the intensity values in the simulated cylinders of the ground truth (GT) and the reconstructed images are compared. Main results. The remaining activity in the β + -emitter image varied from 0.4% to 3.7%. The absolute percentage difference between GT and reconstructed intensity for the pure β + emitter images was found to be between 3.0% and 7.4% for all cases. The absolute percentage difference between the GT and the reconstructed intensity for the β + -γ emitter images ranged from 8.7% to 10.4% for all simulated cases. Significance. These results demonstrate that our approach can reconstruct two separate images with a good quantitation accuracy.

Purpose: To propose a new quality scoring tool, METhodological RadiomICs Score (METRICS), to assess and improve research quality of radiomics studies. Methods: We conducted an online modified Delphi study with a group of international experts. It was performed in three consecutive stages: Stage#1, item preparation; Stage#2, panel discussion among EuSoMII Auditing Group members to identify the items to be voted; and Stage#3, four rounds of the modified Delphi exercise by panelists to determine the items eligible for the METRICS and their weights. The consensus threshold was 75%. Based on the median ranks derived from expert panel opinion and their rank-sum based conversion to importance scores, the category and item weights were calculated. Result: In total, 59 panelists from 19 countries participated in selection and ranking of the items and categories. Final METRICS tool included 30 items within 9 categories. According to their weights, the categories were in descending order of importance: study design, imaging data, image processing and feature extraction, metrics and comparison, testing, feature processing, preparation for modeling, segmentation, and open science. A web application and a repository were developed to streamline the calculation of the METRICS score and to collect feedback from the radiomics community. Conclusion: In this work, we developed a scoring tool for assessing the methodological quality of the radiomics research, with a large international panel and a modified Delphi protocol. With its conditional format to cover methodological variations, it provides a well-constructed framework for the key methodological concepts to assess the quality of radiomic research papers. Critical relevance statement: A quality assessment tool, METhodological RadiomICs Score (METRICS), is made available by a large group of international domain experts, with transparent methodology, aiming at evaluating and improving research quality in radiomics and machine learning. Key points: • A methodological scoring tool, METRICS, was developed for assessing the quality of radiomics research, with a large international expert panel and a modified Delphi protocol. • The proposed scoring tool presents expert opinion-based importance weights of categories and items with a transparent methodology for the first time. • METRICS accounts for varying use cases, from handcrafted radiomics to entirely deep learning-based pipelines. • A web application has been developed to help with the calculation of the METRICS score (https://metricsscore.github.io/metrics/METRICS.html) and a repository created to collect feedback from the radiomics community (https://github.com/metricsscore/metrics). Graphical