Generative AI is transforming medical imaging research through synthesis, enhancement, and reconstruction of clinical images. While these advances show promise in addressing data scarcity and supporting diagnostics, clinical adoption remains limited due to challenges in assessing the trustworthiness of generated content. This study aims to systematically evaluate the integration of uncertainty quantification (UQ) methods within generative models for medical imaging to enhance result reliability. A systematic review following PRISMA guidelines was conducted, analyzing studies from January 2018 to December 2025 that combined generative models with UQ techniques in medical imaging. The search strategy covered major medical and computer science databases, with studies evaluated against predefined inclusion criteria focusing on implementation methodology and performance metrics. From the analysis of 41 eligible studies, 29 focused on radiology, 8 on microscopy, and 4 on optical coherence tomography. Across this heterogeneous body of evidence, integrating UQ was frequently associated with improved performance or with more informative reliability assessment, including reported gains in reconstruction quality, segmentation accuracy, and anomaly detection. Notably, 56% of studies (n=23) were published in 2025, indicating rapid field growth. UQ integration represents a crucial advancement toward trustworthy generative AI systems in medical imaging. Key priorities identified include standardizing uncertainty metrics, developing computationally efficient frameworks, and embedding uncertainty awareness within generation processes. These findings suggest that UQ methods can enhance the clinical reliability of generative AI applications in medical imaging.
Variability in low monitor unit (MU) deliveries has been previously documented, but the underlying pulse‑level behavior has not been fully described due to the integrating nature of most conventional detectors. A plastic scintillation detector (PSD) enables direct visualization of MU microstructure by isolating individual linac pulses. To use a high temporal resolution PSD to characterize pulse‑level MU substructure on a modern linac and quantify how variation in integral pulse magnitude and pulse count influence the reproducibility of low MU and fractional MU deliveries. A Blue Physics Model 11 PSD was used to measure response from individual linac pulses of a Varian TrueBeam operating at energies of 6 MV, 6 MV FFF, 10 MV, and 10 MV FFF. Initial measurements benchmarked reproducibility in pulse count and dose per pulse by delivering 100 MU at a constant dose rate of 400 MU/min. Subsequent measurements focused on low MU cases, delivering 1-3 MU at dose rates of 5-2400 MU/min. The dose per pulse is variable during a single pulse train with a measured coefficient of variation (COV) of up to 13.3%. However, the average dose per pulse across multiple deliveries is stable (< 0.33% COV). This enables highly reproducible readings (< 0.23% COV) for moderate to high MU deliveries. The variation in dose per pulse coupled with a lower number of required pulses leads to challenges when delivering a low number of MUs with FFF beams. Discrepancies in the planned versus delivered MUs were observed on both detectors and the treatment delivery system. Time-resolved analysis shows that dose per pulse fluctuations and the limited number of pulses comprising low MU FFF deliveries lead to measurable variability in delivered MU. These findings characterize the pulsed MU substructure on a modern linac, in which each MU is composed of a finite number of discrete radiation pulses, and quantify the achievable precision of fractional MU delivery, which is fundamentally limited by the dose per pulse (∼0.1 MU for TrueBeam FFF beams).
The objective of this work was to measure the spectra of standardized diagnostic X-ray beams (DXRBs) using a cadmium zinc telluride (CdZnTe) spectrometer. The mean energy as well as energy resolution were also estimated from the measured spectra. The CdZnTe radiation detector of nominal volume 500 mm3 in combination with an in-house-developed tungsten pinhole collimator was used to measure the spectra of standardized DXRBs of nominal energies 40 kV to 100 kV. The detector was placed at a distance of 1000 mm from the focal spot of the X-ray tube, and the field size of 200 mm × 200 mm was opened. The 241Am, 133Ba, and 57Co reference gamma ray sources were used for the calibration of the CdZnTe spectrometer. The measured spectra were corrected using the stripping procedure and the methodology was validated by comparing the corrected and reference spectra. The mean energy values and energy resolution were determined from the corrected spectra. The effective energies of beams were also determined from the first half-value layer measured experimentally. The mean energies estimated from corrected X-ray spectra agreed within 5% for the beam qualities studied. The energy resolution deteriorated with the energy of DXRBs. In-house-developed pinhole collimator, and CdZnTe spectrometer were successfully used for the measurement of DXRBs spectra. The X-ray beams have been fully characterized for the calibration of diagnostic X-ray dosimeters and quality control instruments.
To analyze the inhibitory effect of carbon ion (C-ion) radiation-induced biological effects on the proliferation of lung adenocarcinoma (LUAD) by multi-omics integration. We investigated C-ion radiation effects on LUAD cellular responses using A549 and LLC cell lines. Clonogenic survival assays, DNA damage, and metastasis quantified radiation sensitivity, while a multi-cell co-culture system (A549/Beas-2B/LLC/MLE-12) differentiated direct vs. bystander effects. Integrated transcriptomics and targeted metabolomics identified radiation-responsive genes/metabolites, with pathway analysis conducted through MetaboAnalyst 6.0. Clinical relevance of CPT1, GCH1, and EPAS1 was assessed using UCSC Xena and Kaplan-Meier Plotter survival data. Radiation-induced molecular changes were validated by RT-qPCR and immunoblotting across cell types. An A549 tumor-bearing mouse model was established, and the growth and tumor size of the tumor-bearing mice were observed after irradiation with C-ions. Blood was taken from mice after anesthesia and necropsy to detect changes in the major differential metabolism factor arachidonic acid (AA), and tumor tissues were examined to detect changes in the expression of CPT1, GCH1, and EPAS1 in the tissues. C-ion irradiation exerts a dual anti-proliferative effect on lung adenocarcinoma cells: it directly induces DNA damage (increased γ-H2AX/53BP1 foci), suppresses clonogenic survival, and inhibits tumor cell migration and invasion (p < 0.05); meanwhile, it amplifies the bystander effect from co-cultured normal cells through metabolic reprogramming, thereby enhancing overall cytotoxicity. Transcriptomic-metabolomic integration identified CPT1, GCH1, and EPAS1 as central regulators of radiation-induced metabolic suppression, with AA, 3-hydroxytetradecanoic acid, and 2-methylglutaric acid constituting critical downstream mediators. Database analysis revealed that the differential expression of CPT1, GCH1, and EPAS1 was correlated with the prognostic survival status of patients with lung adenocarcinoma. Carbon ion irradiation downregulated the expression of GCH1 and upregulated CPT1 and EPAS1. Animal experiments demonstrated that carbon ions markedly inhibited tumor proliferation in tumor-bearing mice. The level of arachidonic acid in mouse blood was significantly increased (p < 0.05). Carbon ion radiation suppressed GCH1 expression and promoted the expression of CPT1 and EPAS1 in tumor tissues (p < 0.05). C-ion irradiation suppresses lung adenocarcinoma through a dual mechanism involving direct induction of cellular DNA damage and a metabolically enhanced bystander effect, driven by the CPT1/GCH1/EPAS1 regulatory axis, with arachidonic acid serving as a key downstream mediator.
Establishing the reliability of spinal cord functional magnetic resonance imaging (fMRI) is critical before employing it to assess experimental or clinical interventions. Previous studies have mapped human motor activity primarily to the ipsilateral ventral horn, aligning with myotomal and dermatomal projections. Despite these insights, the test-retest reliability of spinal fMRI remains under-investigated. Here we assessed spinal cord activation during a sensorimotor paradigm involving right-hand grasping and grip force estimation in 30 healthy volunteers. Participants completed two identical scanning visits, each time performing the same task twice, enabling the investigation of test-retest reliability both within a single experimental visit and between visits performed on different days. Aggregating all task runs, motor-evoked activation was observed in ipsilateral ventro-dorsal regions of spinal segmental levels C5-T1, as well as in medial regions of levels C2-C3. Despite highly reliable task performance (grip force) and fMRI signal quality (temporal signal-to-noise ratio), the reliability of motor activation was predominantly poor-to-fair both within and between visits, with notable variability in spatial distribution observed across task runs. Increasing the number of task runs per individual improved the robustness of group-level activation, as indexed by higher activated voxel count, larger cluster spatial extent, and attenuated t-statistic distribution. Although we demonstrated that motor-evoked activation corresponds to the known neuroanatomical organisation of motor circuits, its low test-retest reliability presents a challenge for wider applications of spinal fMRI. Understanding the drivers of low reliability in functional imaging is warranted, but we suggest that looking beyond measurement error is required, including careful consideration of inherent within-individual variability underpinned by neurophysiological and psychological factors.
To assess and compare the radiation-induced effects on the spinal cord of rats following fractionated irradiation with doses of 1.36 Gy and 2.0 Gy per fraction. The risk of radiation-induced myelopathy and spinal cord tolerance represent the major dose-limiting factors during re-irradiation of head-and-neck cancers since spinal cord is exposed to maximum tolerance dose of 45 Gy in 1.8-2 Gy per fraction during first radiation treatment. However, modern radiotherapy techniques such as Volumetric-modulated Arc Therapy (VMAT) with simultaneous integrated boost (SIB) usually irradiate spinal cord with much less dose per fraction than 2 Gy per fraction of prescription dose. In this experimental study, the cervical spinal cord of male Wistar rats was irradiated to a total biologically effective dose of 120 Gy2 using different dose-per-fraction, control group was irradiated with 2 Gy per fraction for a total dose of 60 Gy/30F in 6 weeks while experimental group was given 1.36 Gy per fraction for a total dose of 70.72 Gy/52F in 10.5 weeks. The motor function including motor co-ordination and balance, locomotor activity was observed at 1, 3, and 5 months after the completion of radiation and was compared with baseline values. Motor co-ordination and balance were preserved until the 3rd month in both control and experimental groups, with no significant difference from baseline (0 ± 0). Gradual functional deterioration became evident after 3 months and a significant decline was observed at 5 months in both groups, with a greater reduction in the control group (72 ± 2.4) compared to the experimental group (105 ± 2.8) (F = 367.9, P < 0.001). The decline was more pronounced in rats treated with 2 Gy per fraction than in those receiving 1.36 Gy per fraction. Locomotor activity declined progressively after irradiation in both groups. Baseline activity was 388.8 (control group) and 369.7 (experimental group). At 1 month, the control group showed a significantly greater reduction (180.2 ± 8.2) than the experimental group (272.5 ± 12.1), with a significant group and time interaction (F = 110.4, P < 0.001). This difference persisted at 3 months (control group: 130.2 ± 3.6; experimental group: 185.4 ± 4.8; F = 208.7, P < 0.001) and was most pronounced at 5 months (control group: 99.5 ± 1.7; experimental group: 147.4 ± 4.0; F = 610.9, P < 0.001). This study demonstrates that fractionated irradiation of the spinal cord at 1.36 Gy per fraction, as commonly delivered during VMAT SIB, is less damaging in rats compared to the conventional 2 Gy per fraction used in three-dimensional conformal radiotherapy. Motor co-ordination and locomotor activity deterioration were more pronounced in the control group than in the experimental group. These findings suggest that smaller dose-per-fraction schedules may be well within the spinal cord tolerance, thereby potentially enabling safer re-irradiation scenarios.
Immunoassay methods such as enzyme-linked immunosorbent assay using antigen-antibody reactions have been applied to detect biomarker proteins in the case of various diseases (e.g., cancer). However, these methods comprise several-hour-long pre-treatment processes (incubation and washing) before detection, and their sensitivity is limited by the affinity of protein and antibody sets. Here, we demonstrate the rapid and highly sensitive analysis of a cancer biomarker at several hundreds of attograms using a microflow-type light-induced acceleration system (MF-LAC-SYS) to enhance the reaction between the multiple antibodies and glycoprotein (CEACAM-5) from human plasma containing multiple impurities with the help of theoretical analysis. Adjusting the surface charge of the antibody-modified beads and buffer solution components, we could detect pg mL-1 levels of CEACAM-5 at high sensitivity under a laser beam irradiation of several hundred mW for several minutes at a defocused condition equivalent to the microchannel width. Specifically, we discovered the existence of nanoscale CEACAM-5 aggregates related to the sensitivity of the MF-LAC-SYS, using dynamic light scattering and electron microscopy. The results will potentially pave the way for a platform using unconventional immunoassays for liquid biopsy and blood proteomics.
Gain-of-function (GoF) missense variants in the two-pore domain (K2P) K+ channel TASK-1 (KCNK3) result in developmental delay with sleep apnea (DDSA), a neurodevelopmental channelopathy, while loss-of-function (LoF) variants cause pulmonary arterial hypertension. However, for the related TASK-3 channel (KCNK9), both LoF and GoF variants underlie a distinct neurodevelopmental disorder, KCNK9 imprinting syndrome (KIS). The relationship between genotype and phenotype in these disorders is further complicated because TASK-1 and TASK-3 can co-assemble into heteromeric channels with distinct functional properties. Here, we report additional patients with missense variants in KCNK3 and KCNK9 and investigate the effect of four novel genetic variants on the functional properties of homomeric and heteromeric TASK channels. Interestingly, two of these new pathogenic GoF variants (R131H and L122V) are found in both TASK-1 and TASK-3 and have equivalent functional effects on heteromeric TASK-1/TASK-3 channel activity, yet result in different clinical phenotypes. We have also determined a cryo-EM structure for the pathogenic L122V mutant TASK-3 channel, which suggests that subtle changes in gating and permeation within the inner cavity are responsible for its activatory effect. Overall, these results highlight the dominant role that homomeric TASK channels likely play in defining their associated channelopathies as well as the complexity of interpreting K+ channel dysfunction in pathophysiology.
Introduction.Pulse oximeters are indispensable devices for noninvasive monitoring of arterial blood oxygen saturation (SpO2); however, verification of their accuracy still relies on invasive CO-oximetry performed under controlled desaturation protocols, which are impractical for routine clinical use. Although functional testing using pulse oximeter simulators is widely employed as a surrogate approach, such simulators generally lack direct metrological traceability to CO-oximetry.Objectives.This study evaluated the accuracy of five commercial pulse oximeters under routine clinical conditions at three hospitals, using site-specific CO-oximeters as reference analyzers. The objectives were to characterize real-world pulse oximeter accuracy without induced desaturation, to examine the potential for metrological traceability of arterial blood oxygen saturation measurements obtained from pulse oximeters with respect to the gold standard, and to assess the suitability of selected pulse oximeters as candidate secondary reference devices for traceable and noninvasive verification.Approach.A total of 488 paired SpO2-SaO2measurements were collected from 148 hospitalized patients during routine care. Accuracy metrics, including bias, mean absolute error, root-mean-square accuracy (Arms), Bland-Altman limits of agreement, and concordance correlation coefficient, were calculated for each site and for a pooled dataset.Results.Across all reference systems, the five pulse oximeters exhibitedArmsvalues between 2.1% and 2.7%, well within the acceptance limits specified by ISO 80601-2-61 (4%) and the U.S. Food and Drug Administration (3%). Measured accuracy remained broadly consistent across pulse oximeter models and reference CO-oximeters when standardized handling procedures were applied. Only minor inter-site differences were observed, demonstrating that consistent multicenter evaluation is feasible under routine clinical conditions within the SaO2range of approximately 80%-100%. These findings provide preliminary support for the future evaluation of selected pulse oximeters as candidate secondary reference devices and for the development of traceable, noninvasive verification approaches using calibrated simulators.
The aim of the study was to investigate the dose calculation accuracy of volumetric modulated arc therapy (VMAT) control point (CP) discretisation approaches in Pinnacle and RayStation treatment planning systems (TPSs). The 6MV beam model for Elekta linear accelerator with Agility head generated in Pinnacle and RayStation was used in this study. The impact of CP discretization on dose calculation was assessed using 10 prostate and ten breast VMAT plans generated in each TPS. In Pinnacle, VMAT plans were generated using both 4° and 2° CP gantry spacing, while RayStation plans used 2° spacing. To assess the dosimetric differences due to CP discretization, VMAT plans from Pinnacle were recalculated in RayStation, and vice versa. The impact of CP discretization on target volume and critical structures was assessed by comparing dose volume histogram metrics. The agreement between ArcCheck-measured and TPS-calculated doses was assessed using gamma analysis with a 2%2 mm tolerance criterion. The prostate and breast plans with 2° spacing generated in Pinnacle and recalculated in RayStation showed a mean (Std Dev) target dose agreement of 1.0 (0.4)% and 0.3 (0.8)%, respectively. The similar comparison for RayStation-generated plans recalculated in Pinnacle was 0.8 (0.2)% and 2.4 (0.2)%, respectively. The ArcCheck measurements of Pinnacle-generated prostate and breast plans showed a mean gamma pass rate of 96.5 (1.2)% and 91.5 (5.2)%, respectively. Similarly, RayStation-generated prostate and breast plans had a mean gamma pass rate of 96.7 (2.3)% and 95.4 (3.7)%, respectively. The CP discretization approach implemented in RayStation results in improved dose calculation accuracy of VMAT plans, particularly for plans with larger CP spacing.
Public leaderboards such as the Therapeutics Data Commons (TDC) ADMET benchmark are widely treated as a ranking of state-of-the-art models. However, a high leaderboard position is only meaningful if the corresponding model can actually be reproduced and deployed by an independent researcher. In this work, we audit whether the top-ranked TDC ADMET models meet that bar. We assessed the top-ranked models of all 22 TDC ADMET leaderboards from the perspective of an end user with access only to the publicly released artifacts of each model─its publication, code repository, and installation instructions. For every end point, the top three models were screened with a unified protocol including an execution environment reproducibility check, a data-leakage assessment, verification of the hyperparameter-optimization procedure, and a reevaluation against the current leaderboard. Only three models (CaliciBoost, MapLight, and MapLight + GNN) passed all stages and reproduced their reported performance. The remaining models failed because of unavailable code, nonreproducible environments, runtime incompatibilities, or methodological flaws. We traced direct or indirect data leakage in the MiniMol, GradientBoost, and XGBoost models, and used deliberately overfitted variants of our own Mol2Vec-based models to show that tuning on the public test set─whether accidental or intentional─can substantially inflate both metrics and leaderboard rank. These results indicate that current TDC leaderboard positions cannot be read as a direct measure of model quality and practical applicability and emphasize the urgent need for better public ADMET benchmarks based on the hidden test sets, strict data set versioning and model submission with standardized inference environments.
Concentrations of extracellular vesicles (EVs) and other particles are measured in plasma for biomarker exploration. A commonly used method, flow cytometry, requires plasma dilution to ensure single-particle detection. Since plasma EVs are outnumbered by variable concentrations of lipoproteins, dilution differs between the plasma samples. Dilution can result in misidentification of fluorescent background signals as labeled EVs. This phenomenon, called the dilution paradox, leads to overestimation of plasma EV concentrations, and likely impacts conclusions from earlier performed biomarker studies. This study reevaluated earlier conclusions from our clinical biomarker studies Antiplatelet Therapy Effect on Extracellular Vesicles (AFFECT EV) and Circulating Nanotraces to Identify the Cause of Stroke (CINTICS), by taking the dilution paradox into account. We developed a model that quantifies the fluorescent background and estimates whether a flow cytometry measurement is dominated by fluorescent background, that is, if the measurement is unreliable. This model was applied to the original datasets of the AFFECT EV and CINTICS studies to identify and exclude unreliable measurements. We investigated whether exclusion of unreliable data affects the original conclusions. Our model estimated that 47% (1156/2457) of the evaluated measurements are unreliable, and conclusions from both biomarker studies required adjustment. Our model improves reliability and reproducibility of EV concentration measurements using flow cytometry. We recommend to reanalyze earlier EV flow cytometry studies using our model and to use a fixed dilution factor in future EV flow cytometry studies to enable reliable EV concentration measurements.
To evaluate the repeatability and consistency of rapid quantitative 23Na MRI of the human calf, an acquisition-weighted stack-of-stars (AW-SOSt) sequence was implemented on a 7T MRI system and used for all measurements. Three variants, differing only in nominal in-plane spatial resolution (2.5, 5.0, and 7.5 mm) and corresponding acquisition times (8:04, 4:02, and 2:41 min), while maintaining a constant slice thickness of 15 mm, were employed for imaging the calf skeletal muscle. Quantitative consistency across resolutions and repeatability was evaluated in simulations and consecutive measurements of 10 healthy volunteers. The apparent tissue sodium concentration (aTSC) was determined using a postprocessing pipeline consisting of a B0, B1, relaxation, and partial volume correction (PVC). Deviations of the determined aTSC from simulated ground truth were below 6.0%. The simulated coefficient of variation (CV) improved with decreasing spatial resolution from 1.1% to 0.9%. The in vivo findings matched the simulation results. The CV improved with increasing voxel size, decreasing from 2.7% at 2.5-mm resolution to 2.0% at 7.5-mm resolution. Measurements remained highly consistent across different in-plane resolutions, with a small increase in aTSC (≈0.2 mM) observed as a difference between lower resolutions (7.5 mm) and the highest resolution (2.5 mm). All applied protocols showed good repeatability and high consistency between the protocols. aTSC quantification using low-resolution 23Na MRI showed improved repeatability, whereas in vivo measurements showed no effect on quantitative accuracy. This study demonstrates that low-resolution 23Na MRI with short acquisition times combined with PVC could be a practical alternative to commonly used "high-resolution" techniques with long acquisition times for quantifying aTSC in calf muscle tissue.
Monte Carlo (MC) simulation is widely used in medical physics for radiation transport modeling, beam commissioning, and dose calculation. However, the optimization of primary electron source parameters for clinical linear accelerators using MC simulations can require many repeated simulations, making the process time-consuming. This study aimed to develop a data-driven surrogate framework to guide the optimization of primary electron beam parameters in Geant4 for an Elekta Synergy linear accelerator operated in 10 MeV electron mode. The objective was to predict the gamma pass rate for the 2%/2 mm criterion and to identify the most influential beam parameters. A Geant4-based Monte Carlo model was used to simulate the 10 MeV electron beam. Simulated dose distributions were compared with experimental measurements in water under reference conditions. A feedforward neural network was trained to predict the gamma pass rate from four primary source parameters: mean energy, energy spread, spatial spread, and angular spread. A Gradient Boosting Regressor was also used to evaluate the relative importance of these parameters. The feedforward neural network showed high predictive performance, with an R² value of 0.9924 for the training dataset and 0.9857 for the testing dataset. The model enabled rapid screening of beam parameter configurations. The Gradient Boosting Regressor indicated that the mean energy was the dominant parameter influencing agreement between simulated and measured dose distributions. The proposed deep learning-guided framework can reduce trial-and-error in Monte Carlo-based electron beam model tuning while maintaining clinically relevant accuracy. This approach may support more efficient optimization of primary electron source parameters for clinical linac modeling.
The current study is to assess and compare the performance of an electronic portal imaging device (EPID)-based dosimetry system electronic portal imaging beam (EPIbeam) and a three-dimensional (3D) diode array system (ArcCHECK) using various gamma criteria (3%/3 mm, 2%/2 mm, 3%/2 mm and 1%/1 mm) in verifying intensity-modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT) plans at different anatomical sites. A total of 100 clinical IMRT and VMAT treatment plans covering head and neck (HN), thoracic and pelvic sites were considered for patient specific quality assurance (PSQA). All the plans were calculated with the Monaco TPS (v6.1.4.0) and were treated with an Elekta Infinity linear accelerator with 6 MV photon beams. EPID-based portal dosimetry (EPIbeam, DosiSoft) and ArcCHECK (Sun Nuclear Corporation) were utilized to carry out PSQA. All plans were analyzed using global gamma with 10% dose threshold and 95% pass-rate. A 3%/3 mm, EPID showed almost perfect pass rates (99.48%-100%) in HN, Thoracic, and Pelvic regions, and ArcCHECK was also close behind (97%-100%). EPID and ArcCHECK performed with a high level of accuracy (97.82-100 and 88.5%-98.8%, respectively) under stricter 2%/2 mm criteria. Using 3%/2 mm gamma criteria, EPID demonstrates higher pass rates (98.6%-100%), and ArcCHECK had also increased (93.2%-99.8%). In the most stringent 1%/1 mm, the performance of EPID declined moderately (77.28%-98.73%), and ArcCHECK had the lowest pass rates (61.6%-91.4%). In general, EPID exhibited better and more reproducible dose delivery accuracy in all conditions. The differences between the quality assurance (QA) systems were statistically significant (P < 0.05) (except thorax cases). The PSQA systems EPID-based and ArcCHECK-based are valuable dosimetric verification of IMRT and VMAT. Their sensitivity and accuracy are emphasized by gamma analysis on several criteria. The EPID system offers Logistical advantage in clinical workflow, whereas ArcCHECK provides 3D dose verification. Both modalities are viable in clinical QA protocols.
The primary objective of this study was to identify the most suitable curve-fitting method for picket fence analysis by comparing commonly used mathematical fitting approaches with film-based reference measurements. A picket fence test was planned on an Elekta Synergy linear accelerator equipped with an MLCi2 multileaf collimator (40 leaf pairs) using 6-mm slit widths and five pickets with a nominal field geometry of 0.6 cm × 24 cm, covering 24 leaf pairs per picket. For analysis, only the central 20 opposing leaf pairs in each picket were evaluated, yielding a total of 100 analyzed peaks. The identical picket fence plan was delivered to both the electronic portal imaging device (EPID) and Gafchromic film. EPID images were exported as TIFF files and analyzed using PTW MEPHYSTO software to obtain peak coordinates, followed by manual determination of peak center position, maximum amplitude, and full width at half maximum (FWHM). Mean FWHM values were 6.90 mm (Gaussian), 5.69 mm (Lorentzian), 6.29 mm (pseudo-Voigt), 6.92 mm (manual), and 6.56 mm (film). Gaussian fitting and manual measurements demonstrated systematic overestimation of FWHM relative to film, while Lorentzian fitting showed the largest deviation and highest root mean square error (RMSE). Based on percentage error, RMSE, and Bonferroni-corrected nonparametric statistical analysis, pseudo-Voigt curve fitting provides the most accurate and consistent estimation of slit width in EPID-based picket fence analysis.
Measures of speech intelligibility in noise show limited correspondence with difficulties people with hearing loss report from daily life. This mismatch suggests that standard measurement conditions do not sufficiently capture aspects that are relevant for speech perception, such as dip listening and spatial release from masking. In the present study we developed and evaluated a test condition that incorporates these aspects and compared it with a standard condition. Speech intelligibility was measured in 100 participants with normal hearing (NH, N=17) and hearing loss (HL, N=83) ranging from mild to severe. Measurements were conducted using the German matrix sentence test (OLSA) in the standard condition with frontal presentation of stationary noise co-located with the target speech, and the proposed condition with fluctuating, speech-like maskers spatially separated (±60°) from the target. Stimuli were presented via headphones using virtual acoustics. Tests were performed unaided and with individualized amplification. The proposed condition revealed reduced speech intelligibility also for listeners with HL that showed close-to-normal speech intelligibility in the standard condition. With individualized amplification, more listeners with HL showed reduced speech intelligibility compared to NH listeners than in the standard condition. Benefit of amplification varied widely across individuals with similar hearing thresholds, with some listeners showing little or no benefit. The advantages of the proposed condition were driven by masker fluctuations rather than by spatial separation of sound sources. These findings demonstrate that speech intelligibility measurements incorporating fluctuating maskers provide potentially relevant information beyond standard assessments and can support a more individualized assessment of hearing loss.
Serum neurofilament light chain (sNfL) and serum glial fibrillary acidic protein (sGFAP) in neuromyelitis optica spectrum disorder (NMOSD) have recently emerged as potential biomarkers of disease activity, although their clinical relevance remains uncertain. Our aim was to assess longitudinal sNfL and sGFAP dynamics at the individual level in a patient with a severe course of aquaporin-4-antibody-positive (AQP4-IgG) NMOSD who failed common therapeutic approaches and consequently required autologous (ASCT) and ultimately allogeneic (alloSCT) stem cell transplantation. We retrospectively analyzed sNfL and sGFAP levels in 34 longitudinal serum samples collected from a single female patient over an eight-year observation period. Levels of sNfL and sGFAP were measured using the Single Molecule Array (Simoa). Statistical analyses were performed using R software. The median sNfL was 25.21 pg/mL (IQR 32.74), and the median sGFAP level was 358.03 pg/mL (IQR 1954.06). A positive correlation was observed between sNfL levels and the Expanded Disability Status Scale (EDSS) (r = 0.41, p = 0.0149). Elevated sGFAP levels were associated with clinical relapses. This single-patient study provides unique longitudinal data on sNfL and sGFAP in a highly aggressive course of AQP4-IgG-positive NMOSD requiring allogeneic stem cell transplantation. The findings suggest that sNfL may reflect cumulative disability progression, whereas sGFAP appears to more reliably capture disease activity. These observations should be interpreted cautiously as descriptive and hypothesis-generating only.
This study aimed to develop an integrated application that enables the intuitive operation of each radiomics analysis process (image input, feature extraction, dimensionality reduction, model construction, and evaluation) with a graphical user interface (GUI) and no code. Tkinter, a Python standard library, was used to design the GUI. Libraries such as PyRadiomics, scikit-learn, and pydicom were integrated into the GUI for image loading and feature analysis. Feature extraction and selection using the least absolute shrinkage and selection operator, principal component analysis, multiple regression analysis, and linear discriminant analysis-receiver operating characteristic (ROC) analysis were available. Visualization and comma-separated values format outputs were also supported. The two datasets from The Cancer Imaging Archive were used for validation. Additionally, we compared the analysis time with that of conventional methods (e.g., PyRadiomics, RStudio, and Excel). Using the developed application, we were able to reduce the analysis time from approximately 20 min to <1/3 of the 6 min required by the conventional method. Additionally, diagnostic support information was obtained through accuracy evaluation, correlation visualization of features, and ROC analysis. This application improves the efficiency and availability of radiomics analysis and can be easily used by researchers and clinicians with little programming experience. Further enhancements are planned in the future, including the integration of segmentation functions, automatic parameter optimization, and application to other modalities and diseases.
We performed updated subgroup analyses of the EMERALD study to investigate efficacy, safety, and quality of life (QoL) in patients treated with eribulin (E) vs. either docetaxel (DTX) or paclitaxel (PTX). Patients with HER2+ locally advanced/metastatic breast cancer (LABC/MBC) were randomized to receive E or taxane (T) (physician's choice of DTX or PTX; declared in advance at registration) in 21-day cycles, each combined with trastuzumab + pertuzumab. Survival outcomes, treatment patterns, antitumor efficacy, safety, and QoL were examined. The intention-to-treat (and safety) populations comprised 224 (224) patients who received E, 186 (184) who received DTX, and 36 (34) who received PTX. Baseline characteristics were balanced. Progression-free survival (E vs. DTX: 14.0 vs. 13.1 months; hazard ratio [HR], 0.94 [95% CI, 0.74-1.21]; E vs. PTX: 14.1 vs. 10.8 months; HR, 1.00 [95% CI, 0.58-1.73]) and overall survival (E vs. DTX: not reached [NR] vs. NR; E vs. PTX: 75.5 months vs. NR) were similar among the subgroups. Although the incidences of adverse events (AEs) were generally similar, peripheral sensory neuropathy was more frequent with PTX and less frequent with DTX, compared with E. Neutropenia tended to be more frequent with E. Median time to QoL deterioration was longer with E vs. DTX (7.8 vs. 4.7 months) and E vs. PTX (6.1 vs. 3.9 months). This trial revealed comparable efficacy of E vs. either DTX or PTX, when combined with dual HER2 blockade as first-line treatment for HER2+ LABC/MBC. AEs were generally similar between the E and T subgroups. QoL was maintained for longer in the E subgroups vs. T subgroups. ClinicalTrials.gov (NCT03264547; registered: 28 June 2017); University Hospital Medical Information Network (UMIN000027938; registered: 26 June 2017).