Hasil untuk "Medical technology"

E. Mardis

Yun-Shik Lee

D. Leslie, Anna Waterhouse, Julia B Berthet et al.

W. Drexler

Boyi Xu, Lida Xu, Hongming Cai et al.

A. Koenderink, A. Alú, A. Polman

Matteo Wohlrapp, Niklas Bubeck, Daniel Rueckert et al.

AI-based image reconstruction models are increasingly deployed in clinical workflows to improve image quality from noisy data, such as low-dose X-rays or accelerated MRI scans. However, these models are typically evaluated using pixel-level metrics like PSNR, leaving their impact on downstream diagnostic performance and fairness unclear. We introduce a scalable evaluation framework that applies reconstruction and diagnostic AI models in tandem, which we apply to two tasks (classification, segmentation), three reconstruction approaches (U-Net, GAN, diffusion), and two data types (X-ray, MRI) to assess the potential downstream implications of reconstruction. We find that conventional reconstruction metrics poorly track task performance, where diagnostic accuracy remains largely stable even as reconstruction PSNR declines with increasing image noise. Fairness metrics exhibit greater variability, with reconstruction sometimes amplifying demographic biases, particularly regarding patient sex. However, the overall magnitude of this additional bias is modest compared to the inherent biases already present in diagnostic models. To explore potential bias mitigation, we adapt two strategies from classification literature to the reconstruction setting, but observe limited efficacy. Overall, our findings emphasize the importance of holistic performance and fairness assessments throughout the entire medical imaging workflow, especially as generative reconstruction models are increasingly deployed.

Binesh Sadanandan, Vahid Behzadan

Consistency under paraphrase, the property that semantically equivalent prompts yield identical predictions, is increasingly used as a proxy for reliability when deploying medical vision-language models (VLMs). We show this proxy is fundamentally flawed: a model can achieve perfect consistency by relying on text patterns rather than the input image. We introduce a four-quadrant per-sample safety taxonomy that jointly evaluates consistency (stable predictions across paraphrased prompts) and image reliance (predictions that change when the image is removed). Samples are classified as Ideal (consistent and image-reliant), Fragile (inconsistent but image-reliant), Dangerous (consistent but not image-reliant), or Worst (inconsistent and not image-reliant). Evaluating five medical VLM configurations across two chest X-ray datasets (MIMIC-CXR, PadChest), we find that LoRA fine-tuning dramatically reduces flip rates but shifts a majority of samples into the Dangerous quadrant: LLaVA-Rad Base achieves a 1.5% flip rate on PadChest while 98.5% of its samples are Dangerous. Critically, Dangerous samples exhibit high accuracy (up to 99.6%) and low entropy, making them invisible to standard confidence-based screening. We observe a negative correlation between flip rate and Dangerous fraction (r = -0.89, n=10) and recommend that deployment evaluations always pair consistency checks with a text-only baseline: a single additional forward pass that exposes the false reliability trap.

Chenxin Yuan, Shoupeng Chen, Haojiang Ye et al.

Accurate segmentation of 3D medical scans is crucial for clinical diagnostics and treatment planning, yet existing methods often fail to achieve both high accuracy and computational efficiency across diverse anatomies and imaging modalities. To address these challenges, we propose GCNV-Net, a novel 3D medical segmentation framework that integrates a Tri-directional Dynamic Nonvoid Voxel Transformer (3DNVT), a Geometrical Cross-Attention module (GCA), and Nonvoid Voxelization. The 3DNVT dynamically partitions relevant voxels along the three orthogonal anatomical planes, namely the transverse, sagittal, and coronal planes, enabling effective modeling of complex 3D spatial dependencies. The GCA mechanism explicitly incorporates geometric positional information during multi-scale feature fusion, significantly enhancing fine-grained anatomical segmentation accuracy. Meanwhile, Nonvoid Voxelization processes only informative regions, greatly reducing redundant computation without compromising segmentation quality, and achieves a 56.13% reduction in FLOPs and a 68.49% reduction in inference latency compared to conventional voxelization. We evaluate GCNV-Net on multiple widely used benchmarks: BraTS2021, ACDC, MSD Prostate, MSD Pancreas, and AMOS2022. Our method achieves state-of-the-art segmentation performance across all datasets, outperforming the best existing methods by 0.65% on Dice, 0.63% on IoU, 1% on NSD, and relatively 14.5% on HD95. All results demonstrate that GCNV-Net effectively balances accuracy and efficiency, and its robustness across diverse organs, disease conditions, and imaging modalities highlights strong potential for clinical deployment.

Jessica L. Rosenberg, Nancy Holincheck

Medical technologies, including quantum machine learning (QML) and quantum sensing, represent transformative tools for addressing some of the most pressing challenges in healthcare and drug discovery today. We discuss the ways that these topics engage the interest of high-school, undergraduate, and graduate students in understanding quantum information science and engineering. We describe how students built their understanding of these areas through a research project that allowed them to gain an understanding of the technology, its limitations, and the associated ethical considerations. We also consider the challenges of building this kind of work into the curriculum and of bringing students with interests in the biological and medical areas into quantum science and engineering.

Eric Lyimo, Neema B. Kulaya, Lembris Njotto et al.

Abstract Background Malaria, which affects over half of the world’s population, is controlled through clinical interventions and vector control strategies. However, these efforts are threatened by resistance to anti-malarial drugs and insecticides, as well as affected by environmental, ecological, and climatic changes. This study examined changes in malaria prevalence and related factors based on data from 18 cross-sectional surveys conducted in two villages in northeastern Tanzania. Methods From 2003 to 2021, annual cross-sectional malariometric surveys were conducted in two study villages, Mkokola (lowland) and Kwamasimba (highland), samples collected to determine Plasmodium falciparum infection and human exposure to malaria vector Anopheles. Pearson's chi-squared test was used for comparing proportions, logistic and linear regressions test were used analyse associations. Generalized Estimating Equations (GEE) was used to analyse the relationship between malaria prevalence and climatic variables. Results Malaria prevalence in Kwamasimba and Mkokola dropped from ~ 25% and ~ 80% to 0% and 1%, respectively, between 2003 and 2011, reaching 0% in both villages by 2014. This decline was associated with increased bed net use and reduced exposure to Anopheles bites. However, between 2018 and 2021, prevalence resurged, with Kwamasimba reaching 2003–2004 levels despite high bed net use. Between 2003 and 2021 there was an increasing trend in average monthly maximum temperatures (R2 = 0.1253 and 0.2005), and precipitation (R2 = 0.125 and 0.110) as well as minimum relative humidity (R2 = 0.141 and 0.1162) in Kwamasimba and Mkokola villages, respectively, while maximum relative humidity slightly decreased. Furthermore, during 2003–2011, malaria prevalence was positively associated with temperature, maximum temperature, and relative humidity, while precipitation showed a negative association (Estimate:− 0.0005, p < 0.001). Between 2012–2021, all climatic factors, including temperature (Estimate: 0.0256, p < 0.001), maximum temperature (Estimate: 0.0121, p < 0.001), relative humidity (Estimate: 0.00829, p < 0.001), and precipitation (Estimate: 0.000105, p < 0.001), showed positive associations. Conclusion From 2003 to 2014, malaria prevalence declined in two Tanzanian villages but resurged after 2018, particularly in highland Kwamasimba. Most likely, vector dynamics affected by changing climatic conditions drove this resurgence, emphasizing the need for adaptive, climate-informed malaria control strategies.

Xiaofei Huang, Mengru Xie, Yixuan Wang et al.

Abstract At advanced phases of atherosclerosis, the rupture and thrombogenesis of vulnerable plaques emerge as primary triggers for acute cardiovascular events and fatalities. Pathogenic infection such as periodontitis-associated Porphyromonas gingivalis (Pg) has been suspected of increasing the risks of atherosclerotic cardiovascular disease, but its relationship with atherosclerotic plaque destabilization remains elusive. Here we demonstrated that the level of Pg-positive clusters positively correlated with the ratio of necrotic core area to total atherosclerotic plaque area in human clinical samples, which indicates plaque instability. In rabbits and Apoe −/− mice, Pg promoted atherosclerotic plaque necrosis and aggravated plaque instability by triggering oxidative stress, which led to macrophage necroptosis. This process was accompanied by the decreased protein level of forkhead box O3 (FOXO3) in macrophages. The mechanistic dissection showed that Pg lipopolysaccharide (LPS) evoked macrophage oxidative stress via the TLR4 signaling pathway, which subsequently activated MAPK/ERK-mediated FOXO3 phosphorylation and following degradation. While the gingipains, a class of proteases produced by Pg, could effectively hydrolyze FOXO3 in the cytoplasm of macrophages. Both of them decreased the nuclear level of FOXO3, followed by the release of histone deacetylase 2 (HDAC2) from the macrophage scavenger receptor 1 (Msr1) promoter, thus promoting Msr1 transcription. This enhanced MSR1-mediated lipid uptake further amplified oxidative stress-induced necroptosis in lipid-laden macrophages. In summary, Pg exacerbates macrophage oxidative stress-dependent necroptosis, thus enlarges the atherosclerotic plaque necrotic core and ultimately promotes plaque destabilization.

Hao Jin, Jiandong Ding, Xiaoguo Zhang et al.

Abstract Modulating the balance between pro- and anti-inflammatory monocyte subsets holds therapeutic promise in acute myocardial infarction (AMI); however, effective and selective strategies are still lacking. In this study, we are the first to identify Ten-Eleven-Translocation 3 (TET3) expression in circulating monocytes as an independent predictor of AMI occurrence and patient prognosis in a clinical cohort. Building on this novel insight, we engineered a monocyte-targeted RNAi delivery system designed to silence TET3 expression selectively. The platform employs periodic mesoporous silica nanoparticles (PMS) loaded with siTET3, and is surface-modified with polyethylenimine (PEI) and polyethylene glycol (PEG) to enhance cellular uptake. Critically, we further functionalized the system with a CD14 receptor-recognizing transmembrane peptide (Cys-Gly-Trp-Arg-Arg-Arg-NH₂), enabling precise monocyte targeting and internalization. Our targeted nanotherapeutic successfully reprogrammed inflammatory monocytes in vitro, leading to attenuated pro-inflammatory phenotypes. In vivo, treatment with siTET3-loaded nanoparticles markedly reduced infarct size and myocardial fibrosis in murine AMI models. Importantly, translational validation in a porcine AMI model demonstrated substantial suppression of cardiac inflammation and improved post-infarction outcomes following systemic administration of the nanotherapeutic. Graphical abstract

Arwin Nemani, Schahryar Kananian, Annabelle Starck et al.

Abstract Background Refugees and asylum seekers encounter numerous post-migration living difficulties (PMLDs) that can substantially affect their mental health. However, the role of PMLDs remains insufficiently explored, particularly in clinical refugee populations. This study aimed to identify subgroups based on patterns of PMLD by examining their relationship with depressive symptoms and determining which stressors function as key bridges. Methods This study reports a secondary analysis of baseline data from the ReTreat trial. Data were collected from 141 refugees and asylum seekers enrolled in a multicentre randomized controlled trial of a culturally adapted CBT program in Germany. Participants completed measures of depressive symptoms (PHQ-9) and post-migration stressors (27-item checklist). Latent Profile Analysis (LPA) was used to identify distinct burden profiles. Exploratory Factor Analysis (EFA) examined the dimensionality of PMLDs. Network analysis was conducted to investigate symptom–stressor connectivity. Results Three latent profiles emerged: Class 1 showed elevated distress across all domains; Class 2 was characterized by family separation and homesickness; and Class 3 exhibited minimal post-migration stress. EFA of PMLDS supported a four-factor solution: institutional/legal stressors, structural hardship, health/service access, and emotional/family-related strain. Depressive symptoms differed significantly across profiles, with highest scores in the high burden group (Class 1). Network analysis identified institutional/legal and emotional/family-related stressors as central bridge nodes linking PMLDs to depressive symptoms. Conclusions PMLDs are multidimensional and heterogeneously distributed among forcibly displaced individuals. Legal insecurity and emotional strain are particularly influential in connecting environmental hardship to depressive symptoms. Trial registration This study uses baseline data from a registered randomized controlled trial (DRKS00021536).

Yankun Wu, Yuta Nakashima, Noa Garcia

Social biases in generative models have gained increasing attention. This paper proposes an automatic evaluation protocol for text-to-image generation, examining how gender bias originates and perpetuates in the generation process of Stable Diffusion. Using triplet prompts that vary by gender indicators, we trace presentations at several stages of the generation process and explore dependencies between prompts and images. Our findings reveal the bias persists throughout all internal stages of the generating process and manifests in the entire images. For instance, differences in object presence, such as different instruments and outfit preferences, are observed across genders and extend to overall image layouts. Moreover, our experiments demonstrate that neutral prompts tend to produce images more closely aligned with those from masculine prompts than with their female counterparts. We also investigate prompt-image dependencies to further understand how bias is embedded in the generated content. Finally, we offer recommendations for developers and users to mitigate this effect in text-to-image generation.

Nisreen Salama, Rebhi Bsharat, Abdallah Alwawi et al.

Abstract Background AI can improve medical practice, address staff shortages, and enhance diagnostic efficiency. The ChatGPT of Open AI, launched in 2022, uses AI in medical education. However, the long-term impact is uncertain, and integration varies globally, particularly in the Middle East. Aim To explore the knowledge, practices, and attitudes of nursing students in Palestinian universities regarding AI, specifically the use of ChatGPT. Methodology A cross-sectional design was used to conduct this study. The study was performed at 8 private and governmental universities in the West Bank, Palestine, from 1st May 2024 to 30 May 2024, and 304 nursing students participated. Results The study revealed that 84.5% of nursing students at Palestinian universities were aware of AI technology, yet 69.9% lacked formal education or training related to ChatGPT. Despite this gap, 79% supported the integration of AI into nursing curricula and specialized training programs, reflecting strong optimism about its role in education and healthcare. While 58.6% had used AI in their coursework and 68.1% felt comfortable with technology, disparities in proficiency and access remain key barriers to effective AI integration. Major challenges to AI adoption in Palestine include insufficient training, the absence of AI-focused curricula, and financial constraints, underscoring the need for institutional and pedagogical reforms. Concerns about AI’s reliability, costs, and potential diagnostic errors persist, emphasizing the complexities of its integration into nursing education and practice. Conclusion This study highlights the knowledge, attitudes, and practices of Palestinian nursing students regarding AI and ChatGPT. It reveals that, despite growing awareness, the lack of formal education on AI underscores the need for comprehensive curricula. While students’ express optimism about AI’s potential in healthcare, concerns about its reliability and integration persist. The study also reveals that barriers such as inadequate training, limited curricula, and financial constraints must be addressed to effectively integrate AI into nursing education and prepare students for its expanding role in healthcare.

Xiangyu Wu, Hailiang Zhang, Yang Yang et al.

In this paper, we present our champion solution to the Global Artificial Intelligence Technology Innovation Competition Track 1: Medical Imaging Diagnosis Report Generation. We select CPT-BASE as our base model for the text generation task. During the pre-training stage, we delete the mask language modeling task of CPT-BASE and instead reconstruct the vocabulary, adopting a span mask strategy and gradually increasing the number of masking ratios to perform the denoising auto-encoder pre-training task. In the fine-tuning stage, we design iterative retrieval augmentation and noise-aware similarity bucket prompt strategies. The retrieval augmentation constructs a mini-knowledge base, enriching the input information of the model, while the similarity bucket further perceives the noise information within the mini-knowledge base, guiding the model to generate higher-quality diagnostic reports based on the similarity prompts. Surprisingly, our single model has achieved a score of 2.321 on leaderboard A, and the multiple model fusion scores are 2.362 and 2.320 on the A and B leaderboards respectively, securing first place in the rankings.

Andrew Hoopes, Neel Dey, Victor Ion Butoi et al.

We present VoxelPrompt, an end-to-end image analysis agent that tackles free-form radiological tasks. Given any number of volumetric medical images and a natural language prompt, VoxelPrompt integrates a language model that generates executable code to invoke a jointly-trained, adaptable vision network. This code further carries out analytical steps to address practical quantitative aims, such as measuring the growth of a tumor across visits. The pipelines generated by VoxelPrompt automate analyses that currently require practitioners to painstakingly combine multiple specialized vision and statistical tools. We evaluate VoxelPrompt using diverse neuroimaging tasks and show that it can delineate hundreds of anatomical and pathological features, measure complex morphological properties, and perform open-language analysis of lesion characteristics. VoxelPrompt performs these objectives with an accuracy similar to that of specialist single-task models for image analysis, while facilitating a broad range of compositional biomedical workflows.

Daniel Duenias, Brennan Nichyporuk, Tal Arbel et al.

The integration of diverse clinical modalities such as medical imaging and the tabular data extracted from patients' Electronic Health Records (EHRs) is a crucial aspect of modern healthcare. Integrative analysis of multiple sources can provide a comprehensive understanding of the clinical condition of a patient, improving diagnosis and treatment decision. Deep Neural Networks (DNNs) consistently demonstrate outstanding performance in a wide range of multimodal tasks in the medical domain. However, the complex endeavor of effectively merging medical imaging with clinical, demographic and genetic information represented as numerical tabular data remains a highly active and ongoing research pursuit. We present a novel framework based on hypernetworks to fuse clinical imaging and tabular data by conditioning the image processing on the EHR's values and measurements. This approach aims to leverage the complementary information present in these modalities to enhance the accuracy of various medical applications. We demonstrate the strength and generality of our method on two different brain Magnetic Resonance Imaging (MRI) analysis tasks, namely, brain age prediction conditioned by subject's sex and multi-class Alzheimer's Disease (AD) classification conditioned by tabular data. We show that our framework outperforms both single-modality models and state-of-the-art MRI tabular data fusion methods. A link to our code can be found at https://github.com/daniel4725/HyperFusion

Manqin Chen, Xinbin Chen, Huaxiang Ling et al.

BackgroundFibrinogen plays a pivotal role in the inflammatory cascade and is intricately linked to the pathogenesis of sepsis. Nevertheless, its significance as a prognostic marker for sepsis-associated acute kidney injury (SA-AKI) remains uncertain. This study aimed to investigate the association between fibrinogen levels and 28-day mortality with sepsis-associated acute kidney injury.MethodThe fibrinogen levels of patients admitted to the intensive care unit of Beth Israel Deaconess Medical Center between 2008 and 2019 were retrospectively assessed, and those diagnosed with SA-AKI were divided into low, middle and high fibrinogen level groups according to tertiles. Multivariate Cox proportional hazards model was used to assess the 28-day mortality risk of the SA-AKI patients.ResultsA total of 3,479 patients with SA-AKI were included in the study. Fibrinogen demonstrated an independent association with 28-day mortality, yielding a hazard ratio (HR) of 0.961 (95% confidence interval [CI]: 0.923-0.999, P = 0.0471). Notably, a non-linear relationship between fibrinogen levels and 28-day mortality was observed, with the threshold observed at approximately 1.6 g/l. The effect sizes and corresponding CIs below and above this threshold were 0.509 (0.367, 0.707) and 1.011 (0.961, 1.064), respectively. Specifically, the risk of mortality among SA-AKI patients decreased by 49.1% for every 1 g/l increment in fibrinogen, provided that fibrinogen levels were less than 1.6 g/l.ConclusionIn patients with SA-AKI, a non-linear relationship was identified between fibrinogen levels and 28-day mortality. Particularly, when their fibrinogen levels were less than 1.6 g/l, a concomitant decrease in 28-day mortality was observed as fibrinogen levels increased.