国家预印本平台

针对电化学溶解氢传感器铂电极对氢的吸附仍然偏强、铍青铜基底上贵金属敏感膜结合力差且易受铜、铍溶出污染的问题，本文采用脉冲电沉积法在铍青铜基底上制备了不同Pt/Ru/Ir摩尔比的Pt-Ru-Ir三元合金敏感膜，并以纯Pt膜为对照，系统研究了其在40 g/L NaF介质（pH≈9.7）中的氢气氧化反应（HOR）性能。借助X射线衍射（XRD）、扫描电子显微镜与能谱（SEM/EDS）表征了膜的晶相结构、微观形貌与元素分布，并通过循环伏安（CV）、旋转圆盘电极线性扫描伏安（RDE-LSV）结合Koutecky-Levich（K-L）、微极化区Butler-Volmer拟合与Tafel分析，定量评价合金化对HOR动力学的提升。结果表明：Ir的引入可进一步细化晶粒并引起（111）衍射峰高角侧偏移，晶格畸变加剧；在所考察的六种配比中，Pt:Ru:Ir = 4:1:1敏感膜兼具细小均匀的颗粒、致密的膜层与适中的晶格畸变，HOR性能最优。机理上，HOR增强源于Ru、Ir诱导的Pt d带中心下移（电子效应）、Ru/Ir亲氧位点邻近供给OH*（双功能协同）以及晶粒细化与（111）择优取向（几何效应）的三重耦合。本研究为铍青铜基底上低成本、高性能三元合金敏感膜的设计及其在溶解氢传感器中的应用提供了实验依据。

针对联盟链协作中敏感信息披露与业务验证并存的需求，构建基于零知识证明的可验证隐私保护模型，并在 Hyperledger Fabric 上实现原型。模型从内容、身份及元数据与操作关系三个维度组织问题，以消息可验证承诺、交易内容机密性、接收方身份保护和发送方身份保护四个子方案协调既有密码学机制。完整 Marlin 验证置于链下，链码检查授权服务签名及状态条件，状态更新以有效提交为前提。历史功能报告记录 11 类测试、41 个有效用例；系统测试中基础流程与完整隐私流程的平均采样时间分别为 130 ms 和 352 ms。独立协议实验描述了三种所测配置的成本差异。结果支持限定场景下的流程可执行性，但采样时间不能确认为有效提交延迟，也不能分解为单一组件开销。该模型为联盟链中减少直接披露并保留验证证据提供了原型实现，不构成绝对匿名或普遍安全保证。

DeepLearning (DL)-based Computer-Aided Diagnostic (CAD) systems have shown great promise in supporting clinicians, but their performance heavily depends on the availability of large, well-annotated datasets. In medical lesion detection, the scarcity of comprehensive, stan dardized annotations and the heterogeneity of existing datasets limit the development of robust foundation models. In this work, we introduce a novel approach to building a multi-domain foundation model for le sion detection by integrating diverse medical image datasets annotated for both anatomical structures and abnormalities. We first propose a student-teacher training strategy to effectively combine datasets with heterogeneous label spaces, mitigating catastrophic forgetting, and im proving feature learning. In addition, building on our previously pro posed example-based EM-DETR framework, we adapt it to jointly learn from multiple domain-specific datasets, enabling fine-grained, example driven detection across modalities. Our model achieves state-of-the-art (SOTA) results, with image-level AUC scores exceeding previous SOTA by 1% point on four key findings in CXRs. We achieve a leading mAP50 in CXRs and mammography, surpassing EM-DETR trained on a single dataset by 2-3% points. This approach paves the way for scalable foundation models in medical imaging that leverage heterogeneous data while maintaining robust and generalizable performance.

Foundation models pretrained on large natural-image collections are increasingly adapted for medical image segmentation, in the expectation that they will replace models trained for each task. Most comparisons, however, use data from the training distribution and a single task-specific baseline whose configuration is rarely justified, and therefore provide limited evidence of performance at new clinical sites. It also remains unclear how much of a foundation model must be adapted, and what determines this. We adapted a foundation model, DINOv2, with strategies ranging from tuning only its bias terms to full fine-tuning, and compared it with U-Nets and ResU-Nets trained from scratch. Experiments covered colonoscopy, breast ultrasound and dermoscopy, with testing on held-out data from the training source and on independent datasets from other institutions. The adapted foundation model generalised better on every external dataset. Within the training distribution, the task-specific models remained competitive when training data were abundant, and delineated boundaries more precisely, while the foundation model more often missed small lesions. Only a small fraction of the foundation model required adaptation, and the capacity needed was determined by the imaging domain rather than by the adaptation method. A widely used polyp test set was also found to duplicate images from another benchmark. Foundation-model adaptation is therefore warranted where models must work across sites, with capacity selected on the target domain.

Let C_h be the closure of the nonreal roots of integer polynomials with coefficients of modulus at most h. With the nonzero real roots included, the closure is connected at every height; without them, connectedness becomes arithmetic. We prove that C_h is connected exactly when h=q(q-1) is a pronic number 2, 6, 12, 20, ..., and has at least five components otherwise. At height two this confirms a twenty-year-old numerical observation of Shmerkin and Solomyak on double zeros of power series: apart from 1/2 they form the interval $[1/ρ,1)=C_2\cap(0,1)$, with $ρ$ the tribonacci constant. Disconnection is almost always infinite. For (p-1)^2<h<=p^2-3, $h\neq p(p-1)$, infinitely many components with nonempty interior accumulate at $\pm p$ and $\pm 1/p$, and when h<p(p-1) these four points are themselves components. Next to each square q^2>=16, at least two of the three heights q^2-2, q^2-1, q^2 have infinitely many components accumulating at a common point of (1/(q+1),1/q). Explicit series give infinitely many components at every nonpronic height below 167 other than 1, 3, 4, 9, 25, 63 and 81, and in general at most sqrt(X)+1 nonpronic heights h<=X can have finitely many components, although C_h is a Peano compactum at every height. The proofs compare values and derivatives of all admissible power series near a real double zero; the comparison sets are open and closed in the whole core, so components separated near the axis cannot reconnect elsewhere. On the real axis, the double zeros at heights 1, 3, 4, 7 form sets with infinitely many components; at height one this proves a conjecture of Shmerkin and Solomyak and one of Solomyak on the connectedness locus for pairs of affine maps.