Background: Understanding how artificial intelligence (AI) is employed to predict, diagnose, and perform relevant analyses in Alzheimer’s disease research is a rapidly evolving field. This study integrated and analyzed the relevant literature from the Science Citation Index (SCI) and Social Science Citation Index (SSCI) on the application of AI in Alzheimer’s disease (AD), covering publications from 2004 to 2023.Objective: This study aims to identify the key research hotspots and trends of the application of AI in AD over the past 20 years through a bibliometric analysis.Methods: Using the Web of Science Core Collection database, we conducted a comprehensive visual analysis of literature on AI and AD published between January 1, 2004, and December 31, 2023. The study utilized Excel, Scimago Graphica, VOSviewer, and CiteSpace software to visualize trends in annual publications and the distribution of research by countries, institutions, journals, references, authors, and keywords related to this topic.Results: A total of 2,316 papers were obtained through the research process, with a significant increase in publications observed since 2018, signaling notable growth in this field. The United States, China, and the United Kingdom made notable contributions to this research area. The University of London led in institutional productivity with 80 publications, followed by the University of California System with 74 publications. Regarding total publications, the Journal of Alzheimer’s Disease was the most prolific while Neuroimage ranked as the most cited journal. Shen Dinggang was the top author in both total publications and average citations. Analysis of reference and keyword highlighted research hotspots, including the identification of various stages of AD, early diagnostic screening, risk prediction, and prediction of disease progression. The “task analysis” keyword emerged as a research frontier from 2021 to 2023.Conclusion: Research on AI applications in AD holds significant potential for practical advancements, attracting increasing attention from scholars. Deep learning (DL) techniques have emerged as a key research focus for AD diagnosis. Future research will explore AI methods, particularly task analysis, emphasizing integrating multimodal data and utilizing deep neural networks. These approaches aim to identify emerging risk factors, such as environmental influences on AD onset, predict disease progression with high accuracy, and support the development of prevention strategies. Ultimately, AI-driven innovations will transform AD management from a progressive, incurable state to a more manageable and potentially reversible condition, thereby improving healthcare, rehabilitation, and long-term care solutions.

Treatments on neoplastic diseases and cancer using genotoxic drugs often cause long-term health problems related to premature aging. The underlying mechanism is poorly understood. Based on the study of a long-lasting senescence-like growth arrest (10-12 weeks) of human dermal fibroblasts induced by psoralen plus UVA (PUVA) treatment, we here revealed that slowly repaired bulky DNA damages can serve as a “molecular scar” leading to reduced cell proliferation through persistent endogenous production of reactive oxygen species (ROS) that caused accelerated telomere erosion. The elevated levels of ROS were the results of mitochondrial dysfunction and the activation of NADPH oxidase (NOX). A combined inhibition of DNA-PK and PARP1 could suppress the level of ROS. Together with a reduced expression level of BRCA1 as well as the upregulation of PP2A and 53BP1, these data suggest that the NHEJ repair of DNA double-strand breaks may be the initial trigger of metabolic changes leading to ROS production. Further study showed that stimulation of the pentose phosphate pathway played an important role for NOX activation, and ROS could be efficiently suppressed by modulating the NADP/NADPH ratio. Interestingly, feeding cells with ribose-5-phosphate, a precursor for nucleotide biosynthesis that produced through the PPP, could evidently suppress the ROS level and prevent the cell enlargement related to mitochondrial biogenesis. Taken together, these results revealed an important signaling pathway between DNA damage repair and the cell metabolism, which contributed to the premature aging effects of PUVA, and may be generally applicable for a large category of chemotherapeutic reagents including many cancer drugs.

As a pivotal research method in the field of granular computing (GrC), fuzzy rough sets (FRSs) have garnered significant attention due to their successful overcoming of the limitations of traditional rough sets in handling continuous data. This paper is dedicated to exploring the application potential of FRS models within the framework of multi-source complex information systems, which undoubtedly holds profound research significance. Firstly, a novel multi-source mixed information system (MsMIS), encompassing five distinct data types, is introduced, thereby enriching the dimensions of data processing. Subsequently, a similarity function, designed based on the unique attributes of the data, is utilized to accurately quantify the similarity relations among objects. Building on this foundation, fuzzy T-norm operators are employed to integrate the similarity matrices derived from different data types into a cohesive whole. This integration not only lays a solid foundation for subsequent model construction but also highlights the value of multi-source information fusion in the analysis of the MsMIS. The integrated results are subsequently utilized to develop FRS models. Through rigorous examination from the perspective of information granularity, the rationality of the FRS model is proven, and its mathematical properties are explored. This paper contributes to the theoretical advancement of FRS models in GrC and offers promising prospects for their practical implementation.

Inadequate force feedback and collision warnings in teleoperated surgical instruments elevate risks during intravascular cannulation. This study introduces an innovative master haptic interface that utilizes magnetorheological (MR) fluid to enhance surgeons’ operational perception during robot-assisted intervention surgery. The system delivers real-time haptic feedback to enhance surgical operational safety and automatically amplifies the feedback force when the contact force on the slave side surpasses the predefined threshold, enabling timely collision alerts. A series of preliminary experiments has been carried out to validate the efficacy of this particular type of haptic interface. The experimental results clearly indicate that the master haptic interface based on MR fluid and carefully designed can effectively enhance the operator’s haptic perception and provide collision alarms in a timely manner with haptic clues, improving the safety and operability of robot intravascular intervention. This research provides some insights into the functional improvements of safe and reliable robot-assisted catheter systems.

This study applies a Few-shot learning algorithm (Met-FSNet) to the multi-target semantic segmentation of renal parenchyma, perirenal fat, liver, spleen, gallbladder, subcutaneous fat, ribs, and vertebrae in CT images. The proposed Fewshot segmentation model incorporates both a ST module and a Fusion module, which enhance the model's spatial adaptability and ability to integrate multi-scale features, resulting in improved robustness. The model demonstrates strong generalization capabilities, performing well not only for parenchymal organs (kidney, liver, and spleen) but also for hollow organs (gallbladder), bony structures (ribs and vertebrae), and adipose tissues (perirenal and subcutaneous fat). Experiments were conducted using a private multi-target segmentation dataset, achieving excellent segmentation results across all eight categories. Subcutaneous fat yielded the highest IOU of 93 %, followed by the liver at 91.1 %. All other categories showed improved IOU metrics compared to the baseline network. This work holds significant clinical value for physicians in the early diagnosis of metabolic syndrome.

This study proposes an explainable few-shot learning framework for cervical cancer nest classification, addressing challenges of limited samples and dense cellular overlap. Integrating Model-Agnostic Meta-Learning (MAML) with a dual-attention feature extraction module (spatial-focused CBAM and frequency-domain FECAM), our method achieves efficient pathological feature representation in Euclidean space. Evaluated on 3,931 cervical cytopathology images (2,000 normal; 1,931 abnormal across four subtypes), the framework attained 92.3 %, $\mathbf{8 7. 5 0 \%, ~} \mathbf{9 7. 8 \%}$, and $\mathbf{9 1. 2 \%}$ on average in the normal category for Precision, Recall, Specificity, F1-Score, and Accuracy respectively. The average accuracy and variance of the four replicate experiments is $\mathbf{0. 6 9 7}$ and $6.02 \mathrm{e}-5$, respectively. It significantly outperformed 18 benchmark models. Meanwhile, the mathematical formalization validates model interpretability, while the dualattention mechanism enhances discriminative feature capture of nuclear abnormalities.