Browse Articles

A clinically applicable and generalizable deep learning model for anterior mediastinal tumors in CT images across multiple institutions

Abstract Rare diseases are often difficult to diagnose, and their scarcity also makes it challenging to develop deep learning models for them due to limited large-scale datasets. Anterior mediastinal tumors—including thymoma and thymic carcinoma—represent such rare entities. A few diagnostic support systems for these tumors have been proposed; however, no prior studies have tested them across multiple institutions, and clinically applicable and generalizable models remain lacking. A total of 711 computed tomography (CT) images were collected from 136 hospitals, each from a different patient with pathologically proven anterior mediastinal tumors (339 males, 372 females). Of these, 485 images were used for training, 62 for tuning, and 164 for external testing. The external testing dataset comprised CT images from 121 unique institutions not involved in the other datasets. A 3D U-Net-based model was trained on the training dataset, and the model with the best performance on the tuning dataset was selected. This model was then evaluated on the external testing dataset for its segmentation and detection performance across different institutions. Based on the reference standards provided by board-certified diagnostic radiologists, the trained model achieved average Dice scores of 0.82, Intersection over Union (IoU) of 0.72, Precision of 0.85, and Recall of 0.82 for tumor segmentation at the CT-image level. The free-response receiver operating characteristic curve—derived from lesion-wise IoU thresholds—demonstrated high sensitivity and a low false-positive rate for tumor detection. Even under a stricter IoU threshold of 0.50, the model maintained a sensitivity of 0.87 with only 0.61 false positives per scan. Our model achieved clinically applicable segmentation and detection performance for anterior mediastinal tumors, demonstrating broad generalizability across 121 institutions and overcoming the data-scarcity challenges inherent to such rare diseases.

Optimization of decompression angles in facial nerve decompression surgery: A decompression model

Facial nerve decompression is a surgical procedure performed for severe facial nerve paralysis associated with conditions such as Bell’s palsy and Ramsay Hunt syndrome. Classical Western studies by Fisch first established the surgical principles of facial nerve decompression, providing the foundation for subsequent work on decompression extent and outcomes. However, the optimal extent of bony decompression of the facial nerve canal remains unclear, and a 180° removal of the surrounding bone has traditionally been performed based on empirical judgment. Nevertheless, more extensive bone removal may increase the risk of surgical complications. This study aimed to evaluate the relationship between the angle of bony decompression around the facial nerve canal and pressure reduction, in order to determine the optimal decompression angle. To achieve this, a simplified experimental model was employed to quantitatively evaluate the relationship between decompression angle and internal pressure reduction, providing mechanical insight into facial nerve decompression rather than clinical data. To evaluate this relationship, a decompression model was created, and pressure changes were measured at opening angles ranging from 30° to 180°. The results revealed that a 150° decompression provided a comparable reduction in pressure to that of a 180° decompression. These findings suggest that the extent of bone removal can be minimized while still achieving sufficient pressure reduction, potentially lowering the risk of nerve injury. We also observed significant pressure reduction at 30°, suggesting utility in regions where extensive bone removal is difficult. The finding that a 150° decompression produced an effect comparable to that of 180° is an important contribution toward improving surgical safety. Moving forward, we aim to refine the decompression model and conduct further investigations using more detailed angle settings, with the goal of establishing a practical surgical technique.

Trustworthy prediction of enzyme commission numbers using a hierarchical interpretable transformer

Dissociable age-dependent effects of emotion on scene and location memory

Morphological traits and microbiome diversity in the free-living nematodes Acrobeles complexus and Zeldia punctata

Morphological adaptations play a key role in shaping the feeding behavior and microbiome associations of Cephalobidae nematodes. To investigate how morphology influences nematode-associated microbiomes, we selected two widely distributed species: Acrobeles complexus , exhibiting elaborated oral structures, and Zeldia punctata , with simpler oral morphology. Unlike earlier studies that reported the microbiomes of A. complexus and Z. punctata independently, this study is the first to directly compare the two species. By integrating in silico re-analysis of our previously published microbiome datasets with new light microscopy and scanning electron microscopy (SEM) observations, we demonstrate how morphological adaptations, such as labial probolae and cuticle structures, shape associated bacterial communities. Our results revealed that A. complexus harbored a more diverse bacterial community than Z. punctata . Morphology showed that the complex oral structures of A. complexus facilitated selective bacterial capture, supporting greater microbial diversity compared to the simpler morphology of Z. punctata . Although statistical significance was not observed, the two species showed distinct patterns of microbial richness and diversity. Principal Coordinate Analysis (PCoA) revealed clearly separated bacterial community structures between the species. Linear discriminant analysis effect size identified potential microbial biomarkers at the genus level, including Firmicutes and Clostridium in A. complexus and Actinobacteria and Pseudomonas in Z. punctata . Predicted functional pathway analysis revealed notable differences in microbial metabolism, such as enrichment of bacterial secretion systems in A. complexus and amoebiasis and lipid metabolism pathways in Z. punctata . This study highlights the role of morphological adaptations in shaping microbiome composition in Cephalobidae nematodes and provides insights into the contribution of free-living bacterivorous nematodes to soil microbial balance. These findings lay the groundwork for further studies on nematode-mediated microbial interactions in soil ecosystems.

Understanding alkali metal promotion in hydrogenation catalysis through Strong Metal–Base Interaction

Synergistic protective and regenerative effects of hyaluronic acid and polynucleotides against UVA-induced oxidative stress in dermal fibroblasts

Abstract Ultraviolet A (UVA) radiation, a principal driver of skin photoaging, generates excessive reactive oxygen species (ROS) in dermal fibroblasts, causing oxidative stress, loss of viability, inflammatory signaling, and extracellular matrix (ECM) degradation. Hyaluronic acid (HA) and polynucleotides (PN) are clinically used dermal biomaterials; however, their protection against UVA injury remains insufficiently defined. We evaluated HA, PN, and their combination in human dermal fibroblasts (HDFs) subjected to photodamage. HDFs were pretreated with HA, PN, or both, irradiated with 20 J/cm 2 UVA, and then maintained in treated media to mimic therapeutic recovery. UVA reduced viability and proliferation, downregulated ECM genes ( COL1A1 , FN1 ), and increased intracellular and mitochondrial ROS and proinflammatory cytokine gene ( TNF-α ). Monotherapy partially alleviated these changes. In contrast, combined HA + PN synergistically improved survival and proliferation, lowered ROS to near baseline, restored ECM transcription, and upregulated antioxidant enzymes ( GPX1 , S OD2 ). HA + PN also increased fibroblast invasion, indicating regenerative activity beyond cytoprotective effects. Under basal conditions, neither HA nor PN showed cytotoxicity or prooxidant effects, while modestly enhancing ECM transcription. These findings demonstrate that HA and PN act synergistically to counter UVA-induced oxidative stress and support dermal regeneration, highlighting a combinatorial bioactive strategy for photoaged skin.

Use of Kaplan-Meier and Cox regressions in the distribution of length of stay in animal shelters for pre-specified calendar periods: Definition, computation, and examples of dog length of stay in orange county California

Computations of length of stay in animal shelters rely on fixed animal cohorts. This is appropriate for research studies that pre-select cohorts, but it is problematic for operational assessments of animal shelters in fixed calendar periods or for comparisons among periods or shelters. Considering only the length of stay of animals whose stay ended within the study period leads to misinterpretation. The use of the Kaplan-Meier and Cox proportional hazards methods with left-truncation and right-censoring is proposed to correctly account for all animals present in the shelter for any fraction of a study period, including those that were present at the beginning and those that remain in care at the end of the period. Examples of dog length of stay in Orange County Animal Care in California show that this computation method corrects the misleading view of historically used calculations of length of stay. Statistically significant changes in length of stay are observed in 8 out of 23 quarterly periods. In a comparison of length of stay before and after the COVID-19 pandemic, the observed significant change in length of stay cannot be explained by variations in sizes and ages of incoming dogs and may be connected to operational policies that restricted visitor access. The proposed approach enables timely tracking of length of stay, accurate comparisons, and assessment of shelter practices and resource needs.

Dual promoter–enhancer activities reflect a unified regulatory logic

Early Permian terrestrial apex predator regurgitalite indicates opportunistic feeding behaviour

Abstract Fossilised digestive remains (bromalites) provide unique insights into extinct animals’ behavioural ecology, physiology and diet. We describe fossilised regurgitated stomach content from the early Permian Bromacker locality (Thuringia, Germany) using micro-CT, osteological, chemical and taphonomical analyses. The regurgitalite consists of a compact cluster of 41 bones with a unique taphonomic signature, including sub-articulated, aligned long bones, an irregular overall shape, and low phosphorus contents in the near-bone matrix. The multitaxic elements comprise a maxilla attributed to the captorhinomorph Thuringothyris mahlendorffae , postcranial elements of the bolosaurid Eudibamus cursoris and an unidentified diadectid, along with several unassignable elements, indicating opportunistic feeding behaviour. The regurgitalite size and composition suggest an apex predator as producer, such as the sphenacodontid Dimetrodon teutonis or the varanopid Tambacarnifex unguifalcatus , both known from Bromacker. This specimen represents the geologically oldest terrestrial regurgitalite and reveals novel insights into the feeding behaviours and the trophic network in a late Palaeozoic continental ecosystem.

Ectopic cambia in wisteria vines are associated with the expression of conserved KNOX genes

The mTOR signaling pathway regulates key steps of mammary gland organoid genesis in a temporal manner

Inverse palladocenes

Developing digital biomarker for predicting cognitive response to multi-domain intervention

Abstract Computerized cognitive training allows real-time tracking of performance metrics that may serve as digital biomarkers. This study investigated the value of a novel in-game digital biomarker, RTACC (Reaction Time-Accuracy Correlation), the correlation between reaction time and accuracy, using data from 130 participants with mild cognitive impairment enrolled in the intervention arm of the SUPERBRAIN-MEET randomized controlled trial. Participants underwent a 24-week multi-domain intervention, consisting of computerized cognitive training, physical exercise, nutritional education, vascular/metabolic risk management, and motivation enhancement. RTACC was derived from task-level RT and accuracy and examined in relation to cognitive and biomarker outcomes. Linear regression analysis revealed a significant association between RTACC and changes in Repeatable Battery for the Assessment of Neuropsychological Status scores from baseline to 24 weeks (beta coefficient = -11.90 ± 3.78, T = − 3.14, P  = 0.002). RTACC also showed a marginal effect on changes in brain-derived neurotrophic factor levels (beta coefficient = − 3.13 ± 1.64, P  = 0.057). Logistic regression analysis demonstrated that RTACC combined with clinical information identified good responders with an area under the receiver operating characteristic curve of 0.73 (95% CI: 0.62–0.84). These findings suggest that this in-game digital biomarker (RTACC) may help identify individuals likely to benefit from multi-domain intervention.

Growth of rhombohedral-stacked single-crystal WS2/MoS2 vertical heterostructures

Correction: Systematic review and meta-analysis of virome profiles and quantification of Torque teno virus load in blood of acute febrile illness patients

Multistep receptor binding of the hepatitis B virus preS1 domain

Clinical relevance of tissue copper, selenium, and cadmium alterations in colorectal cancer

Anti-TLR2 immunotherapy modulates neuron-to-oligodendrocyte propagation of α-synuclein in mouse and human models

Dynamic community detection using class preserving time series generation with Fourier Markov diffusion

Abstract Generating class-consistent time series necessitates the maintenance of both overarching structure and detailed temporal dynamics–an endeavor that current GAN and diffusion models find challenging. We introduce FMD-GAN, a Fourier–Markov diffusion framework that integrates spectral clustering, state-conditioned frequency-domain noise modulation, and a dual-branch temporal–spectral discriminator to generate realistic and class-consistent sequences. In four UCR datasets (ECG200, GunPoint, FordA, ChlorineConc), FMD-GAN attains state-of-the-art or competitive outcomes, with up to a 50% reduction in FID and consistent enhancements in DTW, class consistency accuracy (CCA), and spectral distance (SD) compared to six representative baselines. Ablation studies validate the roles of spectrum masking, Markov-guided diffusion, and adversarial learning, whilst sensitivity analysis illustrates resilience to hyperparameters. Qualitative visualizations demonstrate significant semantic congruence between actual and produced samples. These findings indicate that the integration of spectral priors with probabilistic diffusion facilitates the production of time series that preserve structure and are cognizant of class distinctions, pertinent to biomedical monitoring, sensor analytics, and Tiny AI systems.