Enhancing Target Trial Emulation with Natural Language Processing: A Case Study of Corticosteroids for Sepsis Patients in Critical Care

Presentation Time: 03:30 PM - 03:45 PM

Abstract Keywords: Bioinformatics, Critical Care, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF

Target trial emulation using large-scale real-world patient data (RWD) has attracted more attention from researchers to assess the effectiveness of drug treatments. Controlling confounding factors is important for accurately estimating treatment effects from RWD. Here, we propose a framework that leverages Natural Language Processing techniques to analyze clinical notes and integrates this information with structured data to enhance target trial emulation. Experiment results show the efficiency of our framework regarding balancing confounding variables.

Speaker(s):
Suraj Rajendran, PhD
Weill Cornell Medicine

Author(s):
Suraj Rajendran, BS - Cornell University; Zhenxing Xu - Weill Cornell Medical College; Edward Schenck, MD - Weill Cornell Medicine; Fei Wang, PhD - Weill Cornell Medicine;

Unifying Efficient Attention Algorithms for Extreme Long Medical Text Summarization

Presentation Time: 03:45 PM - 04:00 PM

Abstract Keywords: Large Language Models (LLMs), Deep Learning, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF

Effective summarization of medical text such as electronic health records can significantly alleviate the heavy burdens of clinicians in terms of time allocation. However, unlike traditional natural language summarization, medical text summarization experiences an extremely large sequence length, making it infeasible to perform a memory-intensive attention mechanism. In our study, we unify two mainstream accelerated attention algorithms, namely FlashAttention and PagedAttention, to minimize memory accesses and mitigate memory fragmentation. As a result, our unified attention algorithm demonstrates better memory management with 10\% speedup in exact, long medical text summarization.

Speaker(s):
Zhaozhuo Xu, Ph.D.
Stevens Institute of Technology

Author(s):

A New Approach to Detecting Semantic Novelty of Biomedical Literature

Presentation Time: 04:00 PM - 04:15 PM

Abstract Keywords: Data Mining, Natural Language Processing, Knowledge Representation and Information Modeling
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF

Rapid biomedical publication growth challenges comprehending the evolving scientific landscape. Traditional measures of novelty are usually based on the occurrence of new words or the analysis of citation networks, while the actual text contents of publications are less utilized and underestimated. To address this gap, we propose a novel embedding-based methodology to detect semantic novelty in literature and analyze the research trend evolution. The preliminary evaluation results indicate that our semantic novelty metric is significantly correlated with citations and exhibited higher scores for Nobel-winning papers within PubMed. Moreover, by visualizing the semantic novelty evolution, our method shows the potential of tracing and analyzing the emerging research trends and trajectories in specific fields.

Speaker(s):
Xueqing Peng, PhD
Yale University

Author(s):
Xueqing Peng, PhD - Yale University; Yutong Xie, BA - University of Michigan; Huan He, Ph.D. - Yale University; Yan Hu - UTHealth Science Center Houston; Kalpana Raja, PhD, MRSB, CSci - School of Medicine, Yale University; Fongci Lin, PhD - Yale University; Qijia Liu, Student - University of Michigan; Jeffrey Zhang, PhD - Yale University; Qingyu Chen, PhD - Yale University; Qiaozhu Mei, Ph.D - University of Michigan; Hua Xu, Ph.D - Yale University;

Student Behavior Analysis using YOLOv5 and OpenPose in Smart Classroom Environment

Presentation Time: 04:15 PM - 04:30 PM

Abstract Keywords: Education and Training, Governance of Artificial Intelligence, Evaluation
Primary Track: Foundations
Programmatic Theme: Academic Informatics / LIEAF

The Patient-Reported Outcome Measurement Information System (PROMIS) is used to assess various health domains with electronic health records and neighborhood data for subgroup clustering analysis in 11,525 rehabilitation patients (2020-2023), resulting in four symptom profiles: Normal, Mild, Moderate, and Severe. Factors like Medi-Cal insurance, more physical therapy visits per episode, higher Functional Comorbidity Index (FCI) score, and more cancellations correlate with severe symptom profiles, aiding in developing predictive models for rehabilitation outcomes.

Speaker(s):
Chengcheng Ma, College
Montessori

Author(s):
Xiang Li, Graduate - Chongqing Normal University; Yucheng Ji, Graduate - Chongqing Normal University; Jiayi Yang, Graduate - Chongqing Normal University; Mingyong Li, Doctor；Senior Experimental Engineer - Chongqing Normal University;

"The winding journey of human-machine symbiosis": Nurse researchers' experiences and perceptions of generative artificial intelligence in China: A qualitative study

Presentation Time: 04:30 PM - 04:45 PM

Abstract Keywords: Large Language Models (LLMs), Qualitative Methods, Governance of Artificial Intelligence, Nursing Informatics, Education and Training
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF

The application of generative artificial intelligence (GAI) within the field of nursing research is thriving, yet no studies have explored the experiences of nursing researchers with GAI. This study employed qualitative research methodologies to explore nursing researchers' perceptions and experiences of using GAI within the field of nursing research, providing reference for the judicious application, development, and education of GAI in future nursing studies.

Speaker(s):
Ruifu Kang, BsN
Capital Medical University

Bohan Zhang, MM
The Hong Kong Polytechnic University

Author(s):
Ruifu Kang, BsN - Capital Medical University; Zehui Xuan, BsN - Capital Medical University; Ling Tong, Ph.D - Capital Medical University; Bohan Zhang, MM - The Hong Kong Polytechnic University; QIAN XIAO, PhD - Capital Medical University;