arXiv subjects

Matthias Guckenberger

Publications and source records attributed to Matthias Guckenberger.

3 recordsLinked to original sources

arXiv2026

GLOW-FDG: Generalized cancer LesiOn Whole-body segmentation model for $^{18}$F-FDG-PET/CT

Whole-body fluorodeoxyglucose positron emission tomography combined with computed tomography is widely used in cancer care, but manual lesion delineation is slow, subjective, and difficult to scale. We present GLOW-FDG, an open-source artificial intelligence model for whole-body cancer lesion segmentation in fluorodeoxyglucose positron emission tomography and computed tomography. The model was trained on 1,563 scans spanning multiple cancer types and evaluated on 185 external scans from independent institutions. Across breast cancer, nonmetastatic and oligometastatic lung cancer, head and neck cancer, and metastatic melanoma, GLOW-FDG consistently outperformed publicly available benchmark models in lesion detection, while reducing false positives and maintaining strong segmentation accuracy. Quantification of total tumor burden and total lesion glycolysis was robust across cohorts, and performance approached the variability observed between expert radiation oncologists. These results support GLOW-FDG as a generalizable tool for automated cancer segmentation and quantitative imaging biomarker extraction in whole-body imaging.

eess.IV↗

arXiv2025

Technical Note: Vendor-Specific Approach for Standardized Uptake Value Calculation

The Standardized Uptake Value (SUV) is a critical metric in positron emission tomography (PET) imaging, used to assess metabolic activity. However, calculating SUV from DICOM files presents challenges due to vendor-specific DICOM attributes and variations in the encoding of radiotracer accumulation times. This technical note introduces a robust, vendor-specific SUV calculation strategy that addresses inconsistencies in current methodologies. We also integrate this strategy into an open-source software solution, Z-Rad, capable of converting raw PET DICOM data into body-weight normalized SUV NIfTI files. Our SUV calculation strategy was developed by reviewing DICOM conformance statements from GE, Philips, and Siemens. Validation was conducted using real-world PET datasets, and the proposed strategy was compared to existing software solutions. Our SUV calculation approach demonstrated improved accuracy, particularly in resolving time-related discrepancies in the studied data. Our analysis also identified inconsistencies in the SUV calculation methods used by popular commercial and open-source software solutions, which do not fully account for vendor-specific DICOM attributes and PET image acquisition times. These limitations resulted in errors in SUV estimation reaching 33\% when comparing our strategy to studied software. The proposed vendor-specific SUV calculation strategy significantly enhances accuracy in PET imaging by addressing key inconsistencies caused by variations in DICOM attributes and image acquisition times across different vendors. This method effectively reduces SUV calculation errors and has been integrated into an open-source software, Z-Rad.

physics.med-ph↗

arXiv2024

Multicenter Privacy-Preserving Model Training for Deep Learning Brain Metastases Autosegmentation

Objectives: This work aims to explore the impact of multicenter data heterogeneity on deep learning brain metastases (BM) autosegmentation performance, and assess the efficacy of an incremental transfer learning technique, namely learning without forgetting (LWF), to improve model generalizability without sharing raw data. Materials and methods: A total of six BM datasets from University Hospital Erlangen (UKER), University Hospital Zurich (USZ), Stanford, UCSF, NYU and BraTS Challenge 2023 on BM segmentation were used for this evaluation. First, the multicenter performance of a convolutional neural network (DeepMedic) for BM autosegmentation was established for exclusive single-center training and for training on pooled data, respectively. Subsequently bilateral collaboration was evaluated, where a UKER pretrained model is shared to another center for further training using transfer learning (TL) either with or without LWF. Results: For single-center training, average F1 scores of BM detection range from 0.625 (NYU) to 0.876 (UKER) on respective single-center test data. Mixed multicenter training notably improves F1 scores at Stanford and NYU, with negligible improvement at other centers. When the UKER pretrained model is applied to USZ, LWF achieves a higher average F1 score (0.839) than naive TL (0.570) and single-center training (0.688) on combined UKER and USZ test data. Naive TL improves sensitivity and contouring accuracy, but compromises precision. Conversely, LWF demonstrates commendable sensitivity, precision and contouring accuracy. When applied to Stanford, similar performance was observed. Conclusion: Data heterogeneity results in varying performance in BM autosegmentation, posing challenges to model generalizability. LWF is a promising approach to peer-to-peer privacy-preserving model training.

eess.IV↗