Xiangdong Mu

Dr. Xiangdong Mu


Research Focus & Metrics

Academic Career Length

8Years (2019-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
11 June 2025 in Journal of Medical Internet Research
Verified
Journal Article
Open access
RLRonghao Li
Ronghao Li
SMShuai Mao
Shuai Mao
CZCongmin Zhu
Congmin Zhu
YYYingliang Yang
Yingliang Yang
CTChunting Tan
Chunting Tan
YYYuqing Yang
Yuqing Yang
Background: The rapid advancements in natural language processing, particularly the development of large language models (LLMs), have opened new avenues for managing complex clinical text data. However, the inherent complexity and specificity of medical texts present significant challenges for the practical application of prompt engineering in diagnostic tasks. Objective: This paper explores LLMs with new prompt engineering technology to enhance model interpretability and improve the prediction performance of pulmonary disease based on a traditional deep learning model. Methods: A retrospective dataset including 2965 chest CT radiology reports was constructed. The reports were from 4 cohorts, namely, healthy individuals and patients with pulmonary tuberculosis, lung cancer, and pneumonia. Then, a novel prompt engineering strategy that integrates feature summarization (F-Sum), chain of thought (CoT) reasoning, and a hybrid retrieval-augmented generation (RAG) framework was proposed. A feature summarization approach, leveraging term frequency–inverse document frequency (TF-IDF) and K-means clustering, was used to extract and distill key radiological findings related to 3 diseases. Simultaneously, the hybrid RAG framework combined dense and sparse vector representations to enhance LLMs’ comprehension of disease-related text. In total, 3 state-of-the-art LLMs, GLM-4-Plus, GLM-4-air (Zhipu AI), and GPT-4o (OpenAI), were integrated with the prompt strategy to evaluate the efficiency in recognizing pneumonia, tuberculosis, and lung cancer. The traditional deep learning model, BERT (Bidirectional Encoder Representations from Transformers), was also compared to assess the superiority of LLMs. Finally, the proposed method was tested on an external validation dataset consisted of 343 chest computed tomography (CT) report from another hospital. Results: Compared with BERT-based prediction model and various other prompt engineering techniques, our method with GLM-4-Plus achieved the best performance on test dataset, attaining an F1-score of 0.89 and accuracy of 0.89. On the external validation dataset, F1-score (0.86) and accuracy (0.92) of the proposed method with GPT-4o were the highest. Compared to the popular strategy with manually selected typical samples (few-shot) and CoT designed by doctors (F1-score=0.83 and accuracy=0.83), the proposed method that summarized disease characteristics (F-Sum) based on LLM and automatically generated CoT performed better (F1-score=0.89 and accuracy=0.90). Although the BERT-based model got similar results on the test dataset (F1-score=0.85 and accuracy=0.88), its predictive performance significantly decreased on the external validation set (F1-score=0.48 and accuracy=0.78). Conclusions: These findings highlight the potential of LLMs to revolutionize pulmonary disease prediction, particularly in resource-constrained settings, by surpassing traditional models in both accuracy and flexibility. The proposed prompt engineering strategy not only improves predictive performance but also enhances the adaptability of LLMs in complex medical contexts, offering a promising tool for advancing disease diagnosis and clinical decision-making.

5 December 2019 in Chinese Medical Journal
Verified
Journal Article
Open access
CCCun-Liang Cai
Cun-Liang Cai
MZMing-Qiang Zhang
Ming-Qiang Zhang
JGJun Guo
Jun Guo
LWLi-Wan Wang
Li-Wan Wang
JZJing-Quan Zhao
Jing-Quan Zhao
XMXiangdong Mu
Xiangdong Mu
An abstract is unavailable. This article is available as a PDF only.

1 April 2024 in American Journal of Respiratory and Critical Care Medicine
Verified
Journal Article
Closed access
MZMingqiang Zhang
Mingqiang Zhang
QLQiongzhen Luo
Qiongzhen Luo
XLXuesong Li
Xuesong Li
XMXiangdong Mu
Xiangdong Mu
WWWeiwei Wu
Weiwei Wu

1 January 2021 in Annals of Translational Medicine
Verified
Journal Article
Open access
PDF file
CCCunliang Cai
Cunliang Cai
QZQianli Zeng
Qianli Zeng
GZGuiliang Zhou
Guiliang Zhou
XMXiangdong Mu
Xiangdong Mu
Background: Large cell neuroendocrine carcinoma (LCNEC) of the lung is a rare neuroendocrine neoplasm. Previous studies have shown that microRNAs (miRNAs) are widely involved in tumor regulation through targeting critical genes. However, it is unclear which miRNAs play vital roles in the pathogenesis of LCNEC, and how they interact with transcription factors (TFs) to regulate cancer-related genes. Methods: To determine the novel TF-miRNA-target gene feed-forward loop (FFL) model of LCNEC, we integrated multi-omics data from Gene Expression Omnibus (GEO), Transcriptional Regulatory Relationships Unraveled by Sentence-Based Text Mining (TRRUST), Transcriptional Regulatory Element Database (TRED), and The experimentally validated microRNA-target interactions database (miRTarBase database). First, expression profile datasets for mRNAs (GSE1037) and miRNAs (GSE19945) were downloaded from the GEO database. Overlapping differentially expressed genes (DEGs) and differentially expressed miRNAs (DEMs) were identified through integrative analysis. The target genes of the FFL were obtained from the miRTarBase database, and the Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) functional enrichment analyses were performed on the target genes. Then, we screened for key miRNAs in the FFL and performed gene regulatory network analysis based on key miRNAs. Finally, the TF-miRNA-target gene FFLs were constructed by the hypergeometric test. Results: A total of 343 DEGs and 60 DEMs were identified in LCNEC tissues compared to normal tissues, including 210 down-regulated and 133 up-regulated genes, and 29 down-regulated and 31 up-regulated miRNAs. Additionally, 55 DEMs were predicted through miRanda, mirDB, and TargetScan to be associated with the 77 LCNEC key genes. Finally, ETS1-miR195-CD36 and E2F3-miR195-CD36 from the TF-miRNA-mRNA FFLs were constructed, and were found to be significantly enriched in focal adhesion, cytokine-receptor interaction, hematopoietic cell lineage, and transforming growth factor (TGF)-β signaling pathways. Survival analysis showed that the differential expression of the multi-effect factors ETS1, CD36, and hsa-miR-195 in the FFL significantly affected the survival time of LCNEC patients. Conclusions: In summary, we constructed the TF-miRNA-target gene regulatory network, which is helpful for understanding complex LCNEC regulatory mechanisms. The expression differences of the regulated multi-effect target factors ETS1, CD36, and hsa-miR-195 had a significant impact on the survival and development of LCNEC.

1 May 2025 in International Journal of Chronic Obstructive Pulmonary Disease
Verified
Journal Article
Open access
PDF file
NLNa Li
Na Li
WWWenjing Wang
Wenjing Wang
YLYinyin Lv
Yinyin Lv
CLCaihong Li
Caihong Li
XMXiangdong Mu
Xiangdong Mu
Purpose: This study aims to investigate the effects of case management on elderly patients with chronic obstructive pulmonary disease (COPD) and compare changes in pulmonary function indicators, quality of life, dyspnea, and inhaled medication adherence before and after case management. Patients and Methods: A total of 199 elderly COPD patients were enrolled from Beijing Tsinghua Changgung Hospital between January 2020 and November 2023. We implemented case management programs including regular patient education, personalized interventions, continuous support systems, and timely feedback mechanisms, and subsequently collected patient data at 1, 3, 6, and 12 months after initiating case management. Results: After one year of case management, pulmonary function indicators (FEV1/FVC, FEV1, FEV1%pred, MMEF75/25 and MMEF75/25%) in elderly patients with COPD showed significant improvement (P < 0.001). The CAT score decreased from 17.96± 6.57 (mean ± standard deviation) to 11.80± 6.50, and mMRC score decreased from 2.35± 0.66 to 1.14± 0.92 (p< 0.001). Additionally, 77.9% (155/199) of patients were able to maintain their use of the inhaler for 1 year. Conclusion: This study confirmed the significant role of case management in improving respiratory function, quality of life, and dyspnea in elderly patients with COPD. These findings not only provide a valuable reference for clinicians and patients, but also provide a strong foundation for further optimization of management strategies for elderly patients with COPD.

19 August 2025 in Scientific Reports
Verified
Journal Article
Open access
PDF file
QLQiongzhen Luo
Qiongzhen Luo
XHXinwei He
Xinwei He
JXJiao Xu
Jiao Xu
LLLi Li
Li Li
LZLina Zhao
Lina Zhao
XMXiangdong Mu
Xiangdong Mu
Pulmonary mucormycosis (PM) and invasive pulmonary aspergillosis (IPA) are both invasive pulmonary fungal infections with similar clinical manifestations. Few studies have focused on the expression of serum iron in patients with PM and IPA. A prospective, case‒control study was conducted between January 2023 and December 2024. Patients diagnosed with PM or IPA were assigned to the PM group or the IPA group, respectively. Clinical data were recorded for all enrolled patients. Serum iron, serum ferritin, unsaturated iron-binding capacity (UIBC), total iron-binding capacity (TIBC), and percentage transferrin saturation levels were determined via an automatic biochemical detector. A total of 58 patients with PM and 49 patients with IPA were enrolled. Nine patients in the PM group and one patient in the IPA group died during hospitalization. The serum iron, UIBC, and TIBC levels in the PM group were significantly lower than those in the IPA group (all p < 0.05). The serum ferritin level was significantly greater in the PM group than in the IPA group (p = 0.021). Binary logistic regression analysis revealed that only the TIBC level was an independent risk factor for PM [OR (95% CI) = 0.892 (0.839–0.947), p < 0.001]. Cox proportional hazards regression analysis revealed that the serum iron concentration [HR (95% CI) = 0.872 (0.784–0.969); p = 0.011] was an independent predictor of mortality risk in PM patients. Lower TIBC levels at admission were associated with a greater risk of PM, and reduced serum iron levels predict poor prognosis in patients with PM. Using IPA patients as controls, we preliminarily revealed aberrant iron metabolism in PM patients. Trial registration: ClinicalTrials.gov ID NCT06495905, Date of registration July 11, 2024 (retrospectively registered).