Purpose Total body irradiation (TBI) is an integral component of conditioning regimens for hematopoietic stem cell transplantation. Despite similar biological objectives, planning and delivery of TBI vary substantially across institutions. The aim of this study was to quantitatively assess inter-institutional and technique-dependent variability in contemporary TBI planning by comparing dose distributions generated under standardized conditions across multiple centers in Germany and Switzerland. Methods Twenty-four radiotherapy centers received anonymized total-body computed tomography (CT) datasets of a standardized dummy patient in head-first supine and prone orientations, including predefined points of interest (POI) and organs at risk (OARs). Each center was asked to generate TBI plans for two prescription levels (8 Gy in 4 fractions and 12 Gy in 6 fractions) using its local planning standards, with the entire body defined as the planning target volume (PTV). Submissions included treatment-planning-system (TPS)–based 3D dose distributions and in-house table-based calculations. TPS plans were centrally imported into a TPS for extraction of target coverage and OAR dose metrics, while table-based submissions were evaluated using aggregated POI doses. Results In total, 45 TBI treatment plans from 24 centers were analyzed, including 24 plans for 8 Gy and 21 plans for 12 Gy. Both TPS-based and table-based planning approaches were represented across a broad range of clinical techniques, including volumetic modulated arc therapy (VMAT), Tomotherapy, sweeping beam, static field, bilateral, and translation couch setups. Marked inter-institutional variability was observed, particularly for lung, lens, and kidney doses. At the 12 Gy prescription level, mean lung doses differed by more than 3 Gy between centers, corresponding to relative variations of up to approximately 30%. Conformal intensity-modulated techniques (VMAT and Tomotherapy) showed the highest consistency across institutions, whereas non-conformal approaches exhibited substantially larger dose variability. Increased lung sparing at 12 Gy was consistently associated with reduced target coverage compared to 8 Gy. Conclusion This study provides the first standardized multicenter planning-based comparison of contemporary TBI approaches in Germany and Switzerland. By quantifying inter-institutional variability in dose distributions under uniform conditions, the results outline the current heterogeneity of clinical practice and the range of achievable dosimetric outcomes. Rather than defining a reference standard, these findings establish a structured benchmark of existing planning strategies and highlight the need for harmonization, thereby providing a foundation for future consensus development and end-to-end validation efforts.
PET image quality degradation due to positron range effects in high-energy positron-emitting radionuclides such as Gallium-68 (68Ga) can be mitigated by applying positron range correction (PRC). Deep learning (DL) models have been shown to improve image quality by reducing image noise. This study aimed to evaluate a DL model for simultaneous denoising and 68Ga-specific PRC in clinical [68Ga]Ga-DOTA-TOC PET scans. Three different U-Net-based DL models were trained to map reduced-count PET images reconstructed without PRC (25%, 33%, or 50% of acquired counts) to full-count PET images reconstructed with tissue-independent PRC (FC-PRC). The generated reduced-count-enhanced images (RCE-25%, RCE-33%, RCE-50%) were evaluated using the mean standardized uptake value (SUVmean) in liver and lung, noise levels in liver, and maximum standardized uptake value (SUVmax) and contrast-to-noise ratio (CNR) of soft tissue lesions. Three experienced nuclear medicine physicians assessed the overall image quality (OIQ), overall diagnostic confidence (ODC), and lesion detectability (sensitivity, specificity, accuracy). Compared to FC-PRC, RCE-33% and RCE-50% exhibited similar noise, whereas RCE-25% showed substantial residual noise. CNR was reduced for RCE-33% (-5.6%, p=0.001) and RCE-25% (-8.0%, p<0.001), but was comparable for RCE-50%. All reduced-count-enhanced images demonstrated lower SUVmax (RCE-25%: -3.8%, p=0.02; RCE-33%: -7.0%, p<0.001; RCE-50%: -5.2%, p<0.001). RCE-33% and RCE-50% achieved non-inferior OIQ and ODC compared to FC-PRC, while RCE-25% demonstrated only non-inferior ODC. All reduced-count-enhanced images showed high sensitivity, specificity, and accuracy (>90%). The proposed DL-based approach achieved comparable noise and image quality down to 50% of counts. Although it did not outperform the reconstruction-based PRC, its benefit lies in simultaneous denoising and PRC in a single step.
BACKGROUND Carbon-ion radiotherapy provides highly conformal dose distributions and enhanced biological effectiveness; however, energy-switching delays in the multi-energy beam scanning system remain a challenge for fast dose delivery to minimize the interplay effect for moving tumors, and ultra-fast dose delivery to trigger the FLASH effect in clinical treatment. PURPOSE This study presents a novel treatment workflow based on patient-specific 3D range modulators (3DRMs) that enables fast, single-energy carbon-ion treatments while maintaining conformal dose distributions and demonstrating feasibility for clinical application. METHODS We developed and validated a patient-specific 3DRM workflow in the TRiP treatment planning system that converts a conventionally optimized multi-layer virtual range shifter plan into a single-energy delivery. Multiple Coulomb scattering introduced by the modulator material was quantified via the Highland formula and incorporated into beam modeling, spot optimization, and dose calculation. The workflow was demonstrated for a non-small cell lung cancer (NSCLC) case using three patient-specific 3DRMs with modulator-to-target distances (MTD) of 50, 70, and 100 cm and a prescription dose of 3 Gy(RBE) delivered in a single fraction. The impact of MTD to plan quality was assessed using dose-volume histogram (DVH) metrics for the planning target volume (PTV) and relevant organs at risk (OARs). The 3DRMs were manufactured using high-precision 3D printing. The workflow was validated by cross-comparison of phantom-based dose distributions obtained from TRiP calculations, FLUKA Monte Carlo simulations, and experimental measurements. RESULTS By integrating the 3DRM into the beamline, all 3DRM-based plans achieved full target coverage using a single energy layer, the irradiation times were reduced to 25-33 s per field, corresponding to 4 synchrotron cycles - significantly faster than conventional beam scanning, which requires approximately 2-5 min. Increasing MTDs resulted in predictable lateral penumbra broadening and increased low- and medium-dose exposure to surrounding lung tissue; however, all plans maintained OAR doses within clinical constraints and good target coverage, with PTV V95 values exceeding 99% and inhomogeneity coefficients (IC) of approximately 0.05. Good agreement (γglobal(2%/2 mm, 10% dose threshold) = 91%-99%) was observed between treatment planning calculations, FLUKA simulations, and experimental measurements in phantom-based validations. CONCLUSION This study demonstrates that patient-specific 3DRMs enable precise and efficient fast carbon-ion treatments. The proposed workflow maintains high dosimetric quality while reducing irradiation time, providing a scalable and clinically feasible pathway toward motion-robust and FLASH-compatible carbon-ion radiotherapy.
Precise fitting of threshold (T-) levels in cochlear implant (CI) sound processors is crucial for optimizing speech recognition and audibility for CI users. This study systematically investigated the effects of T-level fitting on speech recognition in quiet in adult CI users. Thirty-five adult, experienced CI users with a Nucleus system (Cochlear, Sydney, Australia) were included in a prospective exploratory study. T-levels for each electrode were individually fitted to each participant's electrical hearing threshold using a standardised adaptive procedure. Speech recognition at sound pressure levels of 50, 65, and 80 dB SPL and the Oldenburg Categorical Loudness Scaling (CLS) were measured, and the baseline and post-intervention values were compared. The preoperative outcome prediction for speech recognition was compared with the results obtained before and after the intervention. Precise T-level fitting to the electrical hearing threshold resulted in overall changes ranging from a reduction of 4.8CL to an increase of 9.9CL per electrode (mean: 0.5CL, SD: 15.7 CL). Mean monosyllabic word recognition improved significantly at all measured sound pressure levels. At 65 dB SPL, it increased from 64% to 74% (p < 0.001). With precise T-level fitting, more CI users achieved their preoperative outcome prediction for speech recognition. Precise, individualized T-level fitting to the electrical hearing threshold enhances speech recognition and increases the percentage of CI users who achieve their predicted clinical outcomes. Individualised, electrode-specific threshold adjustment is important for achieving an optimal CI outcome. Further research is needed to refine the threshold detection methods and fitting procedures for broad clinical application.
