Citations

Abstract Background The cerebellar dentate is a known site of iron accumulation and tau deposition in progressive supranuclear palsy (PSP); however, detailed iron distribution and localized volume and shape changes remain unclear across PSP clinical variants. Objective The study aimed to clarify details about regional susceptibility distribution, volume, and shape deformation in the cerebellar dentate across PSP clinical variants. Methods Sixty‐seven PSP patients, including 31 PSP‐Richardson syndrome, 12 PSP‐parkinsonism, 10 PSP‐progressive gait freezing, and 14 PSP‐cortical variants (PSP‐frontal, PSP‐speech/language, and PSP‐corticobasal syndrome), and 31 healthy controls underwent 3 Tesla magnetic resonance imaging to reconstruct quantitative susceptibility maps. Manual dentate segmentation was performed to extract magnetic susceptibility and volume adjusted for total intracranial volume. Shape deformation was assessed using Deformetrica atlas construction. Group comparisons were evaluated with Kruskal–Wallis tests and receiver operating characteristic analysis. Results Patients with PSP‐Richardson syndrome, PSP‐parkinsonism, and PSP‐cortical variants showed significantly increased dentate susceptibility compared with controls ( P < 0.05). Patients with PSP‐progressive gait freezing showed no significant susceptibility differences. Volume loss was most prominent in PSP‐cortical variants, bilaterally, and PSP‐Richardson syndrome, left hemisphere only ( P < 0.05). Shape analysis revealed subregion‐specific deformation patterns, especially in the anterior and posterior deformation in PSP‐progressive gait freezing and the ventroanterior deformation in PSP‐cortical variants. Conclusions The PSP clinical variants had different patterns of susceptibility, volume, and shape metrics, reflecting distinct neurodegenerative patterns of the cerebellar dentate. These findings contribute to our understanding of the neurobiology of PSP and may help to improve phenotypic differentiation of PSP clinical variants. © 2026 International Parkinson and Movement Disorder Society.

Abstract Radiation therapy (RT) has long been included in the treatment of glioblastoma (GBM). However, radioresistance in cancer cells as well as toxicity in normal tissues are major obstacles to clinical efficacy. Improved understanding of the mechanisms of tumor microenvironment-induced radioresistance during and after radiation therapy can provide fundamental insights to improve clinical outcomes in GBM. Here, using three-dimensional engineered hydrogel models in vitro, we report the influence of extracellular matrix, hypoxia, and adjacent neuronal cells in radiotherapeutic sensitivity. We find that mechanical cues and oxygen availability regulate cellular response to radiation, with softer matrices allowing for more DNA damage. Hyaluronan fragments from the extracellular matrix also modulate rapid metabolic response to radiation, especially in hypoxic environments. We show that neuronal networks influence tumor metabolic activity and the inflammatory response. Overall, we demonstrate here that alternative radiation strategies, such as low dose rate radiation therapy and microenvironmental regulation, have the potential to be more effective in a specific subset of radiosensitive GBM tumors.

Abstract Error analysis is a critical step in evaluating and improving clinical concept extraction models, the most common clinical natural language processing (NLP) task. Unlike corpus annotation, which follows standardized protocols for creating gold-standard datasets, error analysis requires nuanced judgment grounded in both clinical expertise and NLP knowledge. This is especially important given the heterogeneity of clinical text, where variations in documentation style, note structure, and terminology can substantially influence model behavior. Despite its importance, there is currently no standardized, user-level framework to support systematic error analysis in clinical concept extraction task. In this study, we developed and validated MedError, a machine-assisted, human-in-the-loop framework designed to standardize and enhance error analysis for clinical concept extraction tasks. We collected and manually curated a corpus of 1,187 unique errors from a total of 4,237 notes across three different distinct hospitals. The error categories were defined using our previously validated error taxonomy and included 480 false negatives and 707 false positives across 25 error types and 48 clinical concept categories. We evaluated the performance of three proprietary and three open-source large language models (LLMs) in automatically classifying these errors into 26 and 15 predefined categories. We further developed a machine-assisted framework, MedError, which integrates best practices in error analysis, LLM-assisted classification and reasoning, and a user-friendly interface to enable more efficient, reproducible, and context-aware error analysis. The framework supports both single-site and federated multisite error analysis, facilitating the effective translation of clinical NLP systems into real-world settings.