RESUMO
Background: Cardiovascular disease (CVD) is a prevalent non-communicable disease globally and holds the position of being the primary cause of mortality worldwide. Consequently, considerable focus has been directed towards the prevention and management of CVD. PCSK9, a frequently targeted element in the treatment and prevention of CVD, can reduce cardiovascular risk by effectively lowering lipid levels even in the context of statin therapy. It also exhibits substantial potential in the diagnosis and treatment of familial hypercholesterolemia from genetic aspects. This bibliometric study aims to analyze and visualize the global trends and emerging hotspots of PCSK9 and CVD researches and provide researchers with new perspectives in further studies. Methods: The data was obtained from the Web of Science Core Collection database. A total of 2,474 publications related to PCSK9 and CVD published between January 2006 and July 2023 were included. The VOSviewer was used to analyze most-cited references, co-authorship, co-citation, co-occurrence and generate a collaborative network map of authors, countries, and institutions. CiteSpace was used to analyze author and institution centroids, keyword bursts, and timeline graphs. Result: A total of 2,474 articles related to CVD and PCSK9 were included. The number of articles and citations show an increasing trend from year to year. Publications were mainly from the United States. The most active institution was Amgen Inc. Watts, Gerald F. was the most prolific author. Atherosclerosis was the most published journal. Literature co-citation and keyword co-occurrence revealed that early studies focused on the lipid-lowering effects of PCSK9 inhibitors in the context of statins therapy, long-term efficacy, adverse effects, LDLR, diagnosis and treatment of familial hypercholesterolemia. In recent years, myocardial ischemic protection, CRISPR-based editing, and new therapeutic strategies for arteriosclerotic cardiovascular disease have gotten wide attention. The protein convertase, inflammation, beta-polyacetate, and inclisiran may be the important future research directions. Conclusion: This study analyses the current status and global trends in the CVD and PCSK9 studies comprehensively, which may provide researchers and policymakers with new and comprehensive perspectives on in this field of research.
RESUMO
Knowledge distillation improves the performance of a small student network by promoting it to learn the knowledge from a pre-trained high-performance but bulky teacher network. Generally, most of the current knowledge distillation methods extract relatively simple features from the middle or bottom layer of teacher network for knowledge transmission. However, the above methods ignore the fusion of features, and the fused features contain richer information. We believe that the richer and better information contained in the knowledge delivered by teachers to students, the easier it is for students to perform better. In this paper, we propose a new method called Multi-feature Fusion Knowledge Distillation (MFKD) to extract and utilize the expressive fusion features of teacher network. Specifically, we extract feature maps from different positions of the network, i.e., the middle layer, the bottom layer, and even the front layer of the network. To properly utilize these features, this method designs a multi-feature fusion scheme to integrate them together. Compared to features extracted from single location of teacher network, the final fusion feature map contains meaningful information. Extensive experiments on image classification tasks demonstrate that the student network trained by our MFKD can learn from the fusion features, leading to superior performance. The results show that MFKD can improve the Top-1 accuracy of ResNet20 and VGG8 by 1.82% and 3.35% respectively on the CIFAR-100 dataset, which is better than state-of-the-art many existing methods.
