XFXiaolu Fei

Xiaolu Fei

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Xiaolu Fei has a total of 51 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Knowledge-Augmented Spatiotemporal Modeling of Electronic Medical Records for Heart Failure Prognosis Prediction: Development and Validation of KA-STNet (Preprint)

    by Yichen Li, Zhongrang Xie, Muyu Wang, Xin Xin, Jun Huang, Congmin Zhu, Lan Wei, Xiaolu Fei, Hui Chen

    BACKGROUND Heart failure (HF) is a severe and progressive clinical syndrome associated with high mortality, frequent hospitalizations, and substantial healthcare burden worldwide. Identification of critically ill HF patients and accurate assessment of cardiac function grade at discharge are essential for timely intervention, risk stratification, and optimization of post-discharge management. OBJECTIVE This study aims to develop and validate KA-STNet, an interpretable knowledge-augmented spatiotemporal framework for in-hospital mortality prediction and discharge cardiac function grade prediction in patients with HF. METHODS To address the heterogeneity of EMR data, KA-STNet adopts distinct graph representation strategies for static and time-series data. For static data, it integrates external medical knowledge with data-driven statistical associations to learn knowledge-augmented patient representations. For time-series data, it employs a spatiotemporal graph neural network to capture spatial dependencies among laboratory indicators and temporal dynamics in longitudinal laboratory sequences, while incorporating patient-level static representations to guide inter-laboratory association modeling. Interpretability analyses based on Shapley additive explanations (SHAP), GNNExplainer, and Leiden community detection are conducted to identify key predictive factors and clinically meaningful laboratory communities. RESULTS KA-STNet outperformed all baseline models across all evaluated tasks. In the in-hospital mortality risk prediction task, it attained a mean AUROC of 0.968 on the local dataset and 0.897 on the public dataset. In the discharge cardiac function grade prediction task, KA-STNet achieved a mean AUROC of 0.925. Anemia-related variables, indicators of renal function and infection-related markers were identified as important contributors to adverse outcome prediction in patients with HF. Leiden community detection further revealed meaningful multi-organ interaction patterns. Adverse outcomes were associated with network reorganization involving acid–base/electrolyte, inflammatory, and hepatic indicators. CONCLUSIONS KA-STNet provides an effective and interpretable framework for predicting HF outcomes using heterogeneous EMRs. These findings support the value of combining knowledge-augmented static representation learning with spatiotemporal dependency modeling for interpretable prediction of clinically relevant HF outcomes. CLINICALTRIAL Not applicable. This was a retrospective study based on electronic medical records and was not a clinical trial. INTERNATIONAL REGISTERED REPORT RR2-

  • KGC-Explainer: Toward Explainable Knowledge Graph Completion

    by Chengbin Hou, Yufan Song, Jiarui Song, William Cheng-Chung Chu, Xiaolu Fei, Jia Li, Hairong Lv

    Knowledge Graph Completion (KGC) aims to infer missing triples for a given knowledge graph, which can be adopted to various fields ranging from scientific research to real-world applications. Despite the success of numerous KGC methods, the lack of explainability remains a common drawback. However, the explanation of KGC results is crucial in many cases such as providing medical diagnosis and recommending candidates for costly experiments, which could increase the reliability of these techniques to humans. Although a few explainable KGC methods have been proposed, the explainability module is designed for the specific KGC model and cannot be utilized for state-of-the-art KGC models. In this work, we present a post-hoc generic method, namely KGC-Explainer, that can be applied to any KGC model providing triplet scores. KGC-Explainer not only incorporates the input KGC model itself, but also leverages the structural and textual information in knowledge graphs. To demonstrate KGC-Explainer achieving design goals and the superiority of it over other methods, we conduct extensive experiments on two real-world knowledge graphs with different domains and languages, one coming from medical domain in Chinese and another coming from general domain in English. Additionally, we compare the KGC explanation with human explanation, showcasing the practical significance of KGC-Explainer.

  • Intelligent Head and Neck CTA Report Quality Detection with Large Language Models

    by Liping Tian, Yao Lu, Xiaolu Fei, Jie Lu

    This study aims to identify common errors in head and neck CTA reports using GPT-4, ERNIE Bot, and SparkDesk, evaluating their potential for supporting quality control in Chinese radiological reports. This study collected 10,000 head and neck CTA imaging reports from Xuanwu Hospital (Dataset 1) and 5000 multi-center reports (Dataset 2). We identified six common types of errors and detected them using three large language models: GPT-4, ERNIE Bot, and SparkDesk. The overall quality of the reports was assessed using a 5-point Likert scale. We conducted a Wilcoxon rank-sum test and Friedman test to compare error detection rates and evaluate the models’ performance on different error types and overall scores. For Dataset 2, after manual review, we annotated the six error types and provided overall scoring, while also recording the time taken for manual scoring and model detection. Model performance was evaluated using accuracy, precision, recall, and F1 score. The intraclass correlation coefficient measured consistency between manual and model scores, and ANOVA compared evaluation times. In Dataset 1, the error detection rates for final reports were significantly lower than those for preliminary reports across all three model types. Friedman’s test indicated significant differences in error rates among the three models. In Dataset 2, the detection accuracy of the three LLMs for the six error types was above 95%. GPT-4 had a moderate consistency with manual scores (ICC = 0.517), while ERNIE Bot and SparkDesk showed slightly lower consistency (ICC = 0.431 and 0.456, respectively; P < 0.001). The models evaluated one hundred radiology reports significantly faster than human reviewers. LLMs can differentiate the quality of radiology reports and identify error types, significantly enhancing the efficiency of quality control reviews and providing substantial research and practical value in this field.

  • Open Access

    Effects of computerized cognitive training on functional brain networks in patients with vascular cognitive impairment and no dementia

    by Qiong‐Ge Li, Yi Xing, Zu‐De Zhu, Xiao‐Lu Fei, Yi Tang, Jie Lu

    Aims: Previous neuroimaging studies of vascular cognitive impairment, no dementia (VCIND), have reported functional alterations, but far less is known about the effects of cognitive training on functional connectivity (FC) of intrinsic connectivity networks (ICNs) and how they relate to intervention‐related cognitive improvement. This study provides comprehensive research on the changes in intra‐ and inter‐brain functional networks in patients with VCIND who received computerized cognitive training, with a focus on the underlying mechanisms and potential therapeutic strategies.Methods: We prospectively collected 60 patients with VCIND who were randomly divided into the training group (N = 30) receiving computerized cognitive training and the control group (N = 30) receiving fixed cognitive training. Functional MRI scans and cognitive assessments were performed at baseline, at the 7‐week training, and at the 6‐month follow‐up. Utilizing templates for ICNs, the study employed a linear mixed model to compare intra‐ and inter‐network FC changes between the two groups. Pearson correlation was applied to calculate the relationship between FC and cognitive function.Results: We found significantly decreased intra‐network FC within the default mode network (DMN) following computerized cognitive training at Month 6 (p = 0.034), suggesting a potential loss of functional specialization. Computerized training led to increased functional coupling between the DMN and sensorimotor network (SMN) (p = 0.01) and between the language network (LN) and executive control network (ECN) at Month 6 (p < 0.001), indicating compensatory network adaptations in patients with VCIND. Notably, the intra‐LN exhibited enhanced functional specialization after computerized cognitive training (p = 0.049), with significant FC increases among LN regions, which correlated with improvements in neuropsychological measures (p < 0.05), emphasizing the targeted impact of computerized cognitive training on language abilities.Conclusions: This study provides insights into neuroplasticity and adaptive changes resulting from cognitive training in patients with VCIND, with implications for potential therapeutic strategies.


  • Open Access

    Chinese vascular ultrasound report Generation based on LSTM-RNN-Attention model

    by Rong Wu, Huifang Huang, Beibei Liu, Jianhua Gao, Lan Wei, Xiaolu Fei

    It is a tedious task to summarize the diagnostic impressions from the image description of vascular ultrasound, and it is more possible to make mistakes because it requires rich experience and clear logical thinking ability of ultrasound radiologists. In this paper, we explore the use of a LSTM-RNN-Attention (LRA) model to automatically generate diagnostic impressions based on the description of vascular ultrasound images, and evaluate the performance of the model by ROUGE scores. Also, the paper compares this model with pointer generation network (PGN) and basic encoder-decoder (ED) model. The LRA Model achieved a ROUGE-1 score of 26.353, a ROUGE-2 score of 11.587, a ROUGE-L score of 26.168, each score was almost 8 times higher than it of (PGN). The results showed that LRA model presented better performance than these most commonly used models in the field of chinses radiological text summarization and is useful in clinical practice.

  • Open Access

    Improving the Performance of Outcome Prediction for Inpatients With Acute Myocardial Infarction Based on Embedding Representation Learned From Electronic Medical Records: Development and Validation Study

    by Yanqun Huang, Zhimin Zheng, Moxuan Ma, Xin Xin, Honglei Liu, Xiaolu Fei, Lan Wei, Hui Chen

    Background: The widespread secondary use of electronic medical records (EMRs) promotes health care quality improvement. Representation learning that can automatically extract hidden information from EMR data has gained increasing attention. Objective: We aimed to propose a patient representation with more feature associations and task-specific feature importance to improve the outcome prediction performance for inpatients with acute myocardial infarction (AMI). Methods: Medical concepts, including patients’ age, gender, disease diagnoses, laboratory tests, structured radiological features, procedures, and medications, were first embedded into real-value vectors using the improved skip-gram algorithm, where concepts in the context windows were selected by feature association strengths measured by association rule confidence. Then, each patient was represented as the sum of the feature embeddings weighted by the task-specific feature importance, which was applied to facilitate predictive model prediction from global and local perspectives. We finally applied the proposed patient representation into mortality risk prediction for 3010 and 1671 AMI inpatients from a public data set and a private data set, respectively, and compared it with several reference representation methods in terms of the area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), and F1-score. Results: Compared with the reference methods, the proposed embedding-based representation showed consistently superior predictive performance on the 2 data sets, achieving mean AUROCs of 0.878 and 0.973, AUPRCs of 0.220 and 0.505, and F1-scores of 0.376 and 0.674 for the public and private data sets, respectively, while the greatest AUROCs, AUPRCs, and F1-scores among the reference methods were 0.847 and 0.939, 0.196 and 0.283, and 0.344 and 0.361 for the public and private data sets, respectively. Feature importance integrated in patient representation reflected features that were also critical in prediction tasks and clinical practice. Conclusions: The introduction of feature associations and feature importance facilitated an effective patient representation and contributed to prediction performance improvement and model interpretation.


  • Open Access

    Quality Management of Pulmonary Nodule Radiology Reports Based on Natural Language Processing

    by Xiaolu Fei, Pengyu Chen, Lan Wei, Yue Huang, Yi Xin, Jia Li

    To investigate the feasibility of automated follow-up recommendations based on findings in radiology reports, this paper proposed a Natural Language Processing model specific for Pulmonary Nodule Radiology Reports. Unstructured findings used to describe pulmonary nodules in 48,091 radiology reports were processed in this study. We established an NLP model to extract information entities from findings of radiology reports, using deep learning and conditional random-field algorithms. Subsequently, we constructed a knowledge graph comprising 168 entities and four relationships, based on the export recommendations of the internationally renowned Fleischner Society for pulmonary nodules. These were employed in combination with rule templates to automatically generate follow-up recommendations. The automatically generated recommendations were then compared to the impression part of the reports to evaluate the matching rate of proper follow ups in the current situation. The NLP model identified eight types of entities with a recognition accuracy of up to 94.22%. A total of 43,898 out of 48,091 clinical reports were judged to contain appropriate follow-up recommendations, corresponding to the matching rate of 91.28%. The results show that NLP can be used on Chinese radiology reports to extract structured information at the content level, thereby realizing the prompt and intelligent follow-up suggestion generation or post-quality management of follow-up recommendations.


  • Open Access

    Improving Performance of Outcome Prediction for In-patients with Acute Myocardial Infarction Based on Embedding Representation Learned From Electronic Medical Records: Development and Validation Study (Preprint)

    by Yanqun Huang, Zhimin Zheng, Moxuan Ma, Xin Xin, Honglei Liu, Xiaolu Fei, Lan Wei, Hui Chen

    BACKGROUND The widespread secondary utilization of electronic medical records (EMRs) promotes healthcare quality improvement. Representation learning that can automatically extract hidden information from EMR data has gained increasing attention. OBJECTIVE We aimed to propose a patient representation containing more feature associations and task-specific feature importance to improve outcome prediction performance for in-patients with acute myocardial infarction (AMI). METHODS Medical concepts including patients’ age, gender, diagnosis diseases, laboratory tests, structured radiological features, procedures and medications were firstly embedded into real-value vectors using the improved skip-gram algorithm where concepts in the context windows were selected by feature association strengths measured by association rules’ confidence. Then each patient was represented as the sum of the feature embeddings weighted by the task-specific feature importance, which was applied to facilitate predictive model prediction from global and local perspectives. We finally applied the proposed patient representation into mortality risk prediction for 3010 and 1671 AMI in-patients from a public dataset and a private dataset, respectively, comparing with several reference representation methods in terms of the areas under the receiver operator curve (AUC). RESULTS Compared with the reference methods, the proposed embedding-based representation showed consistently superior predictive performance on two datasets, achieving the mean AUCs of 0.861 and 0.980, while the greatest AUCs among reference methods were 0.852 and 0.942 on the public and private datasets, respectively. Feature importance integrated in patient representation also reflected features that were consistently critical in prediction tasks and clinical practice. CONCLUSIONS The introduction of feature associations and feature importance facilitated an effective patient representation and contributed to prediction performance improvement and model interpretation.

  • Open Access

    Sequential Data–Based Patient Similarity Framework for Patient Outcome Prediction: Algorithm Development

    by Ni Wang, Muyu Wang, Yang Zhou, Honglei Liu, Lan Wei, Xiaolu Fei, Hui Chen

    Background: Sequential information in electronic medical records is valuable and helpful for patient outcome prediction but is rarely used for patient similarity measurement because of its unevenness, irregularity, and heterogeneity. Objective: We aimed to develop a patient similarity framework for patient outcome prediction that makes use of sequential and cross-sectional information in electronic medical record systems. Methods: Sequence similarity was calculated from timestamped event sequences using edit distance, and trend similarity was calculated from time series using dynamic time warping and Haar decomposition. We also extracted cross-sectional information, namely, demographic, laboratory test, and radiological report data, for additional similarity calculations. We validated the effectiveness of the framework by constructing k–nearest neighbors classifiers to predict mortality and readmission for acute myocardial infarction patients, using data from (1) a public data set and (2) a private data set, at 3 time points—at admission, on Day 7, and at discharge—to provide early warning patient outcomes. We also constructed state-of-the-art Euclidean-distance k–nearest neighbor, logistic regression, random forest, long short-term memory network, and recurrent neural network models, which were used for comparison. Results: With all available information during a hospitalization episode, predictive models using the similarity model outperformed baseline models based on both public and private data sets. For mortality predictions, all models except for the logistic regression model showed improved performances over time. There were no such increasing trends in predictive performances for readmission predictions. The random forest and logistic regression models performed best for mortality and readmission predictions, respectively, when using information from the first week after admission. Conclusions: For patient outcome predictions, the patient similarity framework facilitated sequential similarity calculations for uneven electronic medical record data and helped improve predictive performance.