Purpose: Quality assurance (QA) of auto-segmentations is necessary for safe clinical deployment. Deep learning models are optimised for data similar to their training set, but reliability drops with out-of-distribution (OOD) inputs. This study evaluates whether uncertainty quantification (UQ) can detect OOD inputs arising from distribution shift in head and neck organ of interest (OOI) auto-segmentation. Materials and methods: Monte Carlo dropout-based uncertainty scores and geometric accuracy of nnU-Net models were evaluated in two experiments. First, as a proof-of-concept, we assessed whether uncertainty scores detect metal artefact (N =10) from non-artefact cases (N =10) on a baseline model (mixed training data, N = 620) and a no-metal model (artefact-free training data, N = 620). Second, in a clinically realistic experiment, we split a fine-tuning dataset by uncertainty score to obtain a narrow model (low-uncertainty cases, N = 21) and a broad model (high-uncertainty cases, N = 21). A held-out test set (N = 76) evaluated whether segmentation accuracy differed between uncertainty categories. Results: Both experiments showed that uncertainty scores reflect the learned training distribution. In the supervised experiment, the no-metal model demonstrated greater uncertainty score separation between artefact and non-artefact cases (1.1 & times; 105 vs 0.6 & times; 105) than the baseline model (1.1 & times; 105 vs 0.8 & times; 105). In the unsupervised experiment, the narrow model performed significantly better on certain than uncertain cases (p = 0.04), while no difference was seen for the broad model (p = 0.34), which did include these cases in training. Conclusions: Uncertainty quantification can detect distribution shifts in radiotherapy auto-segmentation. This could support automated QA workflows by identifying in-distribution cases and enabling the application of specialised models.
Purpose To determine whether early radiographic trajectory data can predict the timing of stability in conservatively managed distal radius fractures (DRFs) and to identify which clinical and radiographic features most influence stabilization timing. Methods This retrospective cohort study analyzed data from 1,585 adult patients with conservatively managed DRFs collected from a single-institution fracture registry (2020−2024). Radiographic stability was classified according to predefined parameters of radial height, volar tilt, radial inclination, and ulnar variance at weekly follow-ups. An explainable machine learning model incorporating demographics, fracture characteristics, comorbidities, and early radiographic changes was developed to predict stability at various weeks. Predictive performance was evaluated using area under the receiver operating characteristic curve, calibration plots, Brier scores, decision curve analysis, and SHapley Additive exPlanations. Results Among 1,585 patients, 40% of fractures achieved radiographic stability by week 3. Baseline radiographic measurements did not differ across stability groups, but first-week displacement differed significantly across all parameters. A combined model incorporating early radiographic changes substantially outperformed a baseline-only model. Decision curve analysis demonstrated a superior net benefit compared with routine imaging strategies. SHapley Additive exPlanations analysis identified early changes in volar tilt as the single most influential predictor, followed by changes in radial inclination, patient age, and dorsal comminution. Conclusions Early radiographic trajectory, particularly first-week changes in volar tilt, predicts the timing of stability far more accurately than baseline features alone. A substantial proportion of conservatively managed DRFs stabilize earlier than conventional protocols assume, suggesting that a trajectory-guided approach could safely reduce late follow-up imaging and immobilization duration in selected patients. Prospective validation is needed before clinical implementation. Clinical relevance First-week radiographic displacement, particularly changes in volar tilt, may be more informative than traditional baseline instability criteria for guiding follow-up frequency and immobilization duration in conservatively managed DRFs. Level of evidence Prognostic/III.
Mind-wandering (MW) studies often suffer from extremely small datasets, making it difficult to model condition-specific electroencephalography (EEG) and motivating data augmentation. In these MW studies, EEG is recorded to distinguish between the MW and on-task (OT) state. We examined whether a generative adversarial network (GAN) pretrained on data-rich on-task trials can be adapted to generate physiologically plausible MW EEG under severe data scarcity. To compare the effects of different types of pretraining, we used an event-related potential (ERP)-based Wasserstein GAN with gradient penalty (WGAN-GP) pretrained on OT trials and compared six training approaches: from-scratch training, a no-training baseline that directly reused the OT-pretrained model, full fine-tuning, and three partial fine-tuning variants that updated selected generator and discriminator layers.Models were evaluated using ERP waveforms and power spectra, EEG features (band power, entropy, P3 amplitude and latency, Petrosian fractal dimension, Hjorth complexity), distributional visualization based on principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE), kernel density estimation (KDE), distance- and divergence-based metrics, and training time and number of trainable parameters.Full fine-tuning produced the overall closest match to real MW trials across waveform-, and distribution-level measures on the held-out test set, whereas partial fine-tuning achieved competitive feature-level similarity. In contrast, the no-training baseline kept OT-like rather than MW-specific patterns, and from-scratch training failed to learn MW-like characteristics. Therefore, adapting an OT-pretrained ERP-WGAN-GP model, especially with full or partial fine-tuning approaches, is more effective for MW EEG trial generation than training from scratch or reusing the source model without adaptation.
Identification and development of physiologically relevant in vitro models are critical for the advancement of creating platforms suitable for tissue engineering, drug testing, and mechanistic research on pathologies. This study focuses on skeletal muscle pathologies and establishes a topography-integrated screening platform combining 96-well plates with an aligned surface topography with pattern dimensions ranging from 0 to 23.3 & micro;m in wavelength and 0 to 3.15 & micro;m in height. The system enables simultaneous analysis of myoblast responses on 12 different patterns while preventing cross-talk between cells when they are grown on gradient surfaces for screening. Mechanistic analysis demonstrates topography-dependent modulation of focal adhesion distribution and cytoskeletal reorganization. Surface topography directs myotube alignment through contact guidance, achieving orientation control in optimal configurations and wrinkle-induced polarization mimics native tissue architecture more effectively than planar substrates. Micropattern dimensions differentially regulate cellular morphogenesis, with micro-size topography features producing minimal myotubes (reduced area, quantity, and length) versus enhanced differentiation at larger topography dimensions. This model enables rapid generation of morphologically distinct myotube phenotypes through dimensional tuning, providing a physio-mimetic solution for objectively evaluating muscle therapeutics and is expandable toward other cell culture approaches to identify optimum cell stimulation for specialized culture.
Tibial plateau fractures (TPFs) remain challenging injuries due to their complex three-dimensional morphology, frequent posterior column involvement, and high incidence of associated soft-tissue lesions. Flexion-type fractures represent a distinct and often under recognized entity that is inadequately addressed using traditional classification systems. Flexion-valgus and flexion-varus mechanisms generate fundamentally different fracture configurations and soft-tissue injury profiles, with important implications for surgical management and prognosis. This narrative review provides a practical, literature- and experience-based overview of the distinguishing features of flexion-valgus and flexion-varus tibial plateau fractures, focusing on fracture morphology, associated ligamentous and meniscal injuries, and key treatment principles. Flexion-valgus injuries predominantly involve the posterolateral tibial plateau, commonly presenting as split-depression or rim impaction fractures, and are frequently associated with anterior cruciate ligament (ACL) and lateral meniscal pathology. In contrast, flexion-varus injuries typically result in posteromedial shear fractures with metaphyseal comminution, often extending into the posterolateral central segment, demonstrating significantly higher rates of concomitant ligamentous and meniscal injuries, poorer functional outcomes, and increased risk of conversion to total knee arthroplasty (TKA). Accurate recognition of the underlying injury mechanism and fracture morphology, distinguishing flexion-valgus from flexion-varus injuries, is essential to guide preoperative planning, surgical exposure, fixation strategy, and soft-tissue management, with the goal of optimising clinical outcomes.
