LM-PCVMNet: Pediatric Cervical Vertebral Maturation Analysis with Deep Fusion of Landmarks and Metadata
本文提出LM-PCVMNet，一种结合解剖标志和元数据的深度学习框架，以提高儿童颈椎成熟度自动分期的准确性。
0 citationsRead paper
Selected work
Representative Papers
A machine-learning-assisted progressive digit-randomness screening framework for detecting non-random patterns in raw numerical research data
This study addresses the lack of effective methods for detecting potential fabrication or non-random patterns in raw numerical research data. To this end, it proposes a progressive digital randomness screening framework (FDRS) that integrates statistical tests with interpretable machine learning models—specifically elastic net logistic regression and random forests—to enable fine-grained risk assessment and stratification of datasets. The framework uniquely combines multidimensional features of decimal-digit structure, a progressive subsampling strategy, and metrics including univariate and joint digit tests, Cramér’s V, entropy, Kullback–Leibler divergence, and a digit preference index. In internal validation, FDRS achieves an AUC of 0.984 and accuracy of 92.7%, and demonstrates strong performance in distinguishing high- from low-risk real-world datasets, substantially enhancing the efficiency of manual verification.
0 citationsRead paper
Ideological Isolation in Online Social Networks: A Survey of Computational Definitions, Metrics, and Mitigation Strategies
This study addresses the growing concern that content homogenization and selective exposure in online social networks intensify ideological segregation, thereby diminishing information diversity and exacerbating societal polarization. To tackle this issue, the work proposes a unified computational framework that systematically integrates four dimensions—network structure, content semantics, user interaction, and cognitive bias—to formally define, measure, and intervene in ideological segregation. By leveraging network topology analysis, semantic modeling, user behavior mining, and recommendation system calibration, the research elucidates the interplay between technical mechanisms and sociopolitical outcomes. The findings offer both a theoretical foundation and actionable technical pathways for designing social media platforms that balance information diversity with user experience.
0 citationsRead paper
A bibliometric analysis on the current situation and hot trends of the impact of microplastics on soil based on CiteSpace
This study addresses the growing concern of soil microplastic (MP) contamination by systematically reviewing research advances from 2013 to 2024 to clarify pollution mechanisms, thematic evolution, and future directions. Using bibliometric analysis via CiteSpace and VOSviewer, the authors integrated keyword co-occurrence, clustering, burst detection, and collaboration networks to identify ten core themes—including plastic pollution dynamics and microbial community responses—and proposed, for the first time, a three-stage evolutionary trajectory: “initial exploration → expansion and deepening → integration and interdisciplinarity,” underpinning a domain-specific knowledge map. Results show a steady annual increase in publications, peaking at 956 papers in 2024; strong collaborative ties among core authors and institutions; and an urgent need for multi-scale ecological risk assessment and targeted mitigation strategies. This work provides both theoretical foundations and methodological frameworks for soil MP monitoring, risk management, and policy-informed intervention.
0 citationsRead paper
Recent publications
Latest Papers
LM-PCVMNet: Pediatric Cervical Vertebral Maturation Analysis with Deep Fusion of Landmarks and Metadata
本文提出LM-PCVMNet，一种结合解剖标志和元数据的深度学习框架，以提高儿童颈椎成熟度自动分期的准确性。
0 citationsRead paper
A machine-learning-assisted progressive digit-randomness screening framework for detecting non-random patterns in raw numerical research data
This study addresses the lack of effective methods for detecting potential fabrication or non-random patterns in raw numerical research data. To this end, it proposes a progressive digital randomness screening framework (FDRS) that integrates statistical tests with interpretable machine learning models—specifically elastic net logistic regression and random forests—to enable fine-grained risk assessment and stratification of datasets. The framework uniquely combines multidimensional features of decimal-digit structure, a progressive subsampling strategy, and metrics including univariate and joint digit tests, Cramér’s V, entropy, Kullback–Leibler divergence, and a digit preference index. In internal validation, FDRS achieves an AUC of 0.984 and accuracy of 92.7%, and demonstrates strong performance in distinguishing high- from low-risk real-world datasets, substantially enhancing the efficiency of manual verification.
0 citationsRead paper
Ideological Isolation in Online Social Networks: A Survey of Computational Definitions, Metrics, and Mitigation Strategies
This study addresses the growing concern that content homogenization and selective exposure in online social networks intensify ideological segregation, thereby diminishing information diversity and exacerbating societal polarization. To tackle this issue, the work proposes a unified computational framework that systematically integrates four dimensions—network structure, content semantics, user interaction, and cognitive bias—to formally define, measure, and intervene in ideological segregation. By leveraging network topology analysis, semantic modeling, user behavior mining, and recommendation system calibration, the research elucidates the interplay between technical mechanisms and sociopolitical outcomes. The findings offer both a theoretical foundation and actionable technical pathways for designing social media platforms that balance information diversity with user experience.
0 citationsRead paper
A bibliometric analysis on the current situation and hot trends of the impact of microplastics on soil based on CiteSpace
This study addresses the growing concern of soil microplastic (MP) contamination by systematically reviewing research advances from 2013 to 2024 to clarify pollution mechanisms, thematic evolution, and future directions. Using bibliometric analysis via CiteSpace and VOSviewer, the authors integrated keyword co-occurrence, clustering, burst detection, and collaboration networks to identify ten core themes—including plastic pollution dynamics and microbial community responses—and proposed, for the first time, a three-stage evolutionary trajectory: “initial exploration → expansion and deepening → integration and interdisciplinarity,” underpinning a domain-specific knowledge map. Results show a steady annual increase in publications, peaking at 956 papers in 2024; strong collaborative ties among core authors and institutions; and an urgent need for multi-scale ecological risk assessment and targeted mitigation strategies. This work provides both theoretical foundations and methodological frameworks for soil MP monitoring, risk management, and policy-informed intervention.
0 citationsRead paper