ZTE (United States)

The diagnostic work-up for vestibular pathologies involves a battery of tests designed to quantify the functioning of the otolith organs and semicircular canals. Clinical data from video head-impulse tests, vestibular-evoked myogenic potentials, subjective visual verticality, and caloric tests are usually collected. Our study applied regression analyses to predict the affected side of a patient group with vestibular schwannoma, learning from laboratory vestibular tests, to assess their relative predictive capacity in predicting the tumor side. Technology or Method: The dataset was pre-processed to handle missing values, outliers, and differences in the measurement scales. The mean asymmetry values and their direction (either negative = left-side asymmetry or positive = right-side asymmetry) were calculated. The classifiers' ability to accurately predict the tumor side was evaluated. Finally, both logistic and multiple regression analyses were conducted. The regression models' binary output (i.e., right or left side affected) was compared to the true labels of the affected side given by magnetic resonance imaging to estimate the model's accuracy. Linear regression analysis showed that caloric, cVEMP and RLLL reached AUCs >0.9; multiple regression revealed an AUC of 0.96 for caloric and cVEMP combined. Our study demonstrated that combining caloric and vestibular-evoked myogenic potential tests provides the most accurate identification of the vestibular schwannoma-affected side, achieving the highest predictive capacity. Furthermore, our findings align with previous studies revealing that the monocular video head-impulse test introduces a gain bias for all three semicircular canals that must be adjusted to correctly estimate semicircular canal function. Clinical and Impact-This study addresses the clinical challenge of finding the affected side in unilateral vestibular schwannoma patients by using machine learning to vestibular tests linking computational methods with clinical practice.

Brown adipose tissue (BAT) plays a key role in energy metabolism and is closely linked to metabolic disorders such as obesity. While PET imaging provides precise BAT quantification via standardized uptake values (SUV), its use remains costly and invasive. A recent approach using convolutional neural networks (CNNs) to predict BAT activity from CT scans has shown potential, but it relies on large datasets and sometimes suffers from limited generalizability. In this work, we investigate a novel, data-efficient alternative using Gaussian Splatting-a neural scene representation technique originally developed for real-time rendering and recently adapted to medical imaging. We first train a dual-channel Gaussian model on a single CT/PET pair. For new patients, this Gaussian representation is fine-tuned using only the CT image. Preliminary results suggest that the model captures general PET structure and some BAT regions, despite limitations due to CT-only supervision. Upcoming work will incorporate prior-guided learning to improve anatomical accuracy and generalization.

(Abstracted from Cancer 2025;131(6):e35769) Goals of several organizations around the world include the equality and empowerment of women and girls, including the United Nations Sustainable Development. A large contributor to health care disparities is the low priority of research in topics surrounding women’s health and in diseases that affect mostly female individuals.

Dawud al-Muqammas, the elder colleague of the much better known Sadia Gaon, the first prominent figure in medieval Jewish philosophy, wrote his major work Isrún-Maqalat (Twenty Chapters) at the end of the 9th century, which was preserved in a manuscript discovered in the National Library of St Petersburg at the end of the 19th century, although not in its entirety. The work was written in Arabic, as was the custom of the time, and forms an integral part of the intellectual-historical fluid known as the (mutazilite) kalam. The world over languages and religions, which was the golden age of Muslim and Jewish philosophy – approximately the 10th-12th centuries – could perhaps be called rational theology in scholastic terms. A particular feature of the genre is that its representatives, from the Pyrenees to modern-day Afghanistan, pondered the same problems: the uniqueness of God, his transcendence, his purely pneumatic character, in short, the borderline questions of revelation and Hellenic philosophy. The main work of Al-Muqammas, especially its 11th chapter, interprets a very complex issue, the relations between predestination and free will in a Jewish, Muslim and Syriac Christian context, attempting to overcome the contradictions that these three religions, and especially Islam, brought to the surface again and again at the turn of the 9th and 10th centuries.

Cyber threats have become increasingly prevalent and sophisticated. Prior work has extracted actionable cyber threat intelligence (CTI), such as indicators of compromise, tactics, techniques, and procedures (TTPs), or threat feeds from various sources: open source data (e.g., social networks), internal intelligence (e.g., log data), and “first-hand” communications from cybercriminals (e.g., underground forums, chats, darknet websites). However, “first-hand” data sources remain underutilized because it is difficult to access or scrape their data. In this work, we analyze (i) 6.6 million posts, (ii) 3.4 million messages, and (iii) 120,000 darknet websites. We combine NLP tools to address several challenges in analyzing such data. First, even on dedicated platforms, only some content is CTI-relevant, requiring effective filtering. Second, “first-hand” data can be CTI-relevant from a technical or strategic viewpoint. We demonstrate how to organize content along this distinction. Third, we describe the topics discussed and how “first-hand” data sources differ from each other. According to our filtering, 20% of our sample is CTI-relevant. Most of the CTI-relevant data focuses on strategic rather than technical discussions. Credit card-related crime is the most prevalent topic on darknet websites. On underground forums and chat channels, account and subscription selling is discussed most. Topic diversity is higher on underground forums and chat channels than on darknet websites. Our analyses suggest that different platforms may be used for activities with varying complexity and risks for criminals.

Transparency and immutability of blockchains can expose metadata and raise concerns about its classification as personal data under privacy regulations. This paper investigates privacy risks associated with metadata in blockchain-based identity systems. Additionally, two privacy-preserving mechanism designs, namely Zero-Knowledge Proof (ZKP) and Homomorphic Encryption (HE), to protect metadata are proposed. As a result, this work introduces the first use case of HE privacy-preserving mechanism in the context of Decentralized Identity (DI) and Self-Sovereign Identity (SSI) systems.

With the breakthrough of Transformer-based pre-trained models, the demand for fine-tuning (FT) to adapt the base pre-trained models to downstream applications continues to grow, so it is essential for service providers to reduce the cost of processing FT requests. Low-rank adaption (LoRA) is a widely used FT technique that only trains small-scale adapters and keeps the base model unaltered, conveying the possibility of processing multiple FT tasks by jointly training different LoRA adapters with a shared base model. Nevertheless, through in-depth analysis, we reveal the efficiency of joint FT is dampened by two heterogeneity issues in the training data — the sequence length variation and skewness. To tackle these issues, we develop LobRA, a brand new framework that supports processing multiple FT tasks by jointly training LoRA adapters. Two innovative designs are introduced. Firstly, LobRA deploys the FT replicas (i.e., model replicas for FT) with heterogeneous resource usages and parallel configurations, matching the diverse workloads caused by the sequence length variation. Secondly, for each training step, LobRA takes account of the sequence length skewness and dispatches the training data among the heterogeneous FT replicas to achieve workload balance. We conduct experiments to assess the performance of LobRA, validating that it significantly reduces the GPU seconds required for joint FT by 45.03%-60.67%.

Hоzirgi kundа о‘smirlik dаvridа bulling muаmmоsi vа ungа qаrshi kurаshishdа qо‘llаnilаdigаn kорing strаtеgiyаlаrining о‘rni muhimdir. О‘smirlik dаvri ijtimоiy vа еmоtsiоnаl rivоjlаnishidа muhim bоsqiсh bо‘lib, bu dаvrdа bulling о‘z nаvbаtidа о‘smirlаrning ruhiy sаlоmаtligigа sаlbiy tа’sir kо‘rsаtishi mumkin, shuning uсhun bu muаmmоni hаl qilishdа sаmаrаli strаtеgiyаlаrni qо‘llаsh zаrur. Mаqоlаdа о‘smirlik dаvridа bullinggа qаrshi kurаshishdа kорing strаtеgiyаlаrining аhаmiyаti hаqidаgi mа’lumоtlаr kеltirilgаn.

(Abstracted from Ultrasound Obstet Gynecol 2024;63:15–23 Congenital anomalies can be diagnosed prenatally through genetic testing, including exome sequencing (ES), quantitative fluorescence polymerase chain reaction (QF-PCR), and chromosomal microarray analysis (CMA), in addition to whole-genome sequencing (WGS). Each method has anomalies it can better detect, with WGS having the greatest diagnostic capability.