Multimodal electronic health record (EHR) data are widely used in clinical applications. Conventional methods usually assume that each sample (patient) is associated with the unified observed modalities, and all modalities available for sample. However, missing modality caused by various social reasons a common issue real-world scenarios. Existing mostly rely on solving generative model learns mapping from latent space to original input space, which an unstable ill-posed inverse problem. To...

UNet and its variants have been widely used in medical image segmentation. However, these models, especially those based on Transformer architectures, pose challenges due to their large number of parameters computational loads, making them unsuitable for mobile health applications. Recently, State Space Models (SSMs), exemplified by Mamba, emerged as competitive alternatives CNN architectures. Building upon this, we employ Mamba a lightweight substitute within UNet, aiming at tackling...

The COVID-19 pandemic highlighted the need for predictive deep-learning models in health care. However, practical prediction task design, fair comparison, and model selection clinical applications remain a challenge. To address this, we introduce evaluate two new tasks-outcome-specific length-of-stay early-mortality patients intensive care-which better reflect realities. We developed evaluation metrics, adaptation designs, open-source data preprocessing pipelines these tasks while also...

The integration of multimodal Electronic Health Records (EHR) data has significantly improved clinical predictive capabilities. Leveraging notes and multivariate time-series EHR, existing models often lack the medical context relevent to tasks, prompting incorporation external knowledge, particularly from knowledge graph (KG). Previous approaches with KG have primarily focused on structured extraction, neglecting unstructured modalities semantic high dimensional knowledge. In response, we...

We propose TAMER, a Test-time Adaptive MoE-driven framework for EHR Representation learning. TAMER combines Mixture-of-Experts (MoE) with Test-Time Adaptation (TTA) to address two critical challenges in modeling: patient population heterogeneity and distribution shifts. The MoE component handles diverse subgroups, while TTA enables real-time adaptation evolving health status distributions when new samples are introduced. Extensive experiments across four real-world datasets demonstrate that...

The lack of standardized techniques for processing complex health data from COVID-19 patients hinders the development accurate predictive models in healthcare. To address this, we present a protocol utilizing real-world multivariate time-series electronic records patients. We describe steps covering necessary setup, standardization, and formatting. then provide detailed instructions creating datasets training evaluating AI designed to predict two key outcomes: in-hospital mortality length...

ABSTRACT In recent years, researchers have increasingly sought batteries as an efficient and cost‐effective solution for energy storage supply, owing to their high density, low cost, environmental resilience. However, the issue of dendrite growth has emerged a significant obstacle in battery development. Excessive during charging discharging processes can lead short‐circuiting, degradation electrochemical performance, reduced cycle life, abnormal exothermic events. Consequently,...

Aircraft detection in remote sensing images has become an attractive research topic, which plays essential role various military and civil applications. In this letter, we develop a novel method for aircraft based on deep residual network (ResNet) Super-Vector (SV) coding. First, variant of ResNet with fewer layers is designed to increase the resolution feature map, multi-level convolutional features are merged into informative description region proposal. Meanwhile, extract histogram...

In datacenter networks, bandwidth-demanding elephant flows without deadline and delay-sensitive mice with strict coexist. They compete each other for limited network resources, the effective scheduling of such mix-flows is extremely challenging. We propose a deep reinforcement learning private link approach (DRL-PLink), which combines software-defined (DRL) to schedule mix-flows. DRL-PLink divides bandwidth establishes some corresponding private-links different types isolate them that...

The study aims to develop AICare, an interpretable mortality prediction model, using electronic medical records (EMR) from follow-up visits for end-stage renal disease (ESRD) patients. AICare includes a multichannel feature extraction module and adaptive importance recalibration module. It integrates dynamic static features perform personalized health context representation learning. dataset encompasses 13,091 demographic data of 656 peritoneal dialysis (PD) patients spanning 12 years. An...

The inherent complexity of structured longitudinal Electronic Health Records (EHR) data poses a significant challenge when integrated with Large Language Models (LLMs), which are traditionally tailored for natural language processing. Motivated by the urgent need swift decision-making during new disease outbreaks, where traditional predictive models often fail due to lack historical data, this research investigates adaptability LLMs, like GPT-4, EHR data. We particularly focus on their...

In intramedullary nail (IMN) surgical operations, one of the main efforts for surgeons is to find axes two distal holes. Two holes on an IMN, which are inside canal a patient's femur, can only be seen in lateral X-ray view. For standard procedure, localization hole trial-and-error process results long time and large dose exposure. this paper, algorithm derive three-dimensional position orientation axis was developed. The first derives through images. Then calculated projecting back boundary...

Abstract to promote the construction of intelligent and information technology in field road traffic is conducive smart cities formulation macro strategies plans for urban development. In view shortcomings current system, based on convolution neural network, situation awareness technology, database other technologies, using CNN, R-CNN, fast R-CNN this paper constructs a deep network model big data image, carries out system demand analysis framework design implementation. Through actual case...

The COVID-19 pandemic highlighted the need for predictive deep learning models in healthcare. However, practical prediction task design, fair comparison and model selection clinical applications remain a challenge. To address this, we introduced evaluated two new tasks - Outcome-specific length-of-stay Early mortality patients intensive care which better reflect realities. We developed evaluation metrics, adaptation designs, open-source data preprocessing pipelines these tasks, while also...

Retracted.

Accurate network modeling can be used to help optimize load balancing or routing/flow scheduling strategies ensure Quality of Service (QoS). However, existing methods have some disadvantages, such as, not being suitable for actual networks and low generalization. This article proposes a Link Delay Model (LDM) based on graph neural (GNN). The key idea is inspired from the following observations: there an inherent correlation between delay, jitter, packet loss, throughput each link (this calls...

Analyzing the health status of patients based on Electronic Health Records (EHR) is a fundamental research problem in medical informatics. The presence extensive missing values EHR makes it challenging for deep neural networks to directly model patient's EHR. Existing learning training protocols require use statistical information or imputation models reconstruct values; however, inject non-realistic data into downstream analysis models, significantly limiting performance. This paper...

<title>Abstract</title> This study investigates the relationship between dietary intake and mortality risk among patients with End-Stage Renal Disease (ESRD), a population for whom nutritional management is crucial yet challenging due to disease's complexity limitations of existing guidelines. Recognizing gaps in current research, particularly lack detailed evidence-based recommendations micronutrient focus predominantly on early stages chronic kidney disease, our work seeks provide guidance...

The integration of multimodal Electronic Health Records (EHR) data has notably advanced clinical predictive capabilities. However, current models that utilize notes and multivariate time-series EHR often lack the necessary medical context for precise tasks. Previous methods using knowledge graphs (KGs) primarily focus on structured extraction. To address this, we propose EMERGE, a Retrieval-Augmented Generation (RAG) driven framework aimed at enhancing modeling. Our approach extracts...