1.Development and validation of PhenoRAG: A visualization tool for automated human phenotype ontology term annotation based on large language models and retrieval-augmented generation technology.
Wei ZHONG ; Yousheng YAN ; Kai YANG ; Yan LIU ; Xinyu FU ; Zhengyang YAO ; Chenghong YIN
Chinese Journal of Medical Genetics 2026;43(1):36-43
OBJECTIVE:
To develop a user-friendly visualization application for the automatic annotation of Human Phenotype Ontology (HPO) terms based on large language models and retrieval-augmented generation (RAG) technology, and to validate its performance in an authoritative case dataset.
METHODS:
By integrating the domestic open-source large language model DeepSeek-V3 with RAG technology, an interactive web application was deployed on the Streamlit cloud platform. Using only the latest official HPO dataset as the data source, the lightweight sentence-embedding model BAAI/bge-small-en-v1.5 was employed to construct a FAISS vector index. During the online phase, a four-step closed-loop process is automatically completed: multilingual translation, phenotype phrase extraction, RAG candidate retrieval, term mapping, and official database validation. 121 English case reports publicly released by BMJ Case Reports and Oxford Medical Case Reports (with a gold-standard HPO set of 1 794 terms) were selected for application validation. Precision, recall, and F1 score were calculated and compared horizontally with traditional dictionary tools, standalone large language models, and the similar application "RAG-HPO". Finally, replace the model with the more advanced ChatGPT-5 and evaluate its performance on the newly extracted dataset.
RESULTS:
An HPO term automatic annotation visualization application named PhenoRAG, based on large language models and RAG technology, was successfully developed. Users can access it directly via a web link. Across the 112 cases, a total of 2 150 HPO terms were generated; 2,064 (96.0%) were fully validated by the official database, with a hallucination rate of 1.3% and an HPO ID-name mismatch rate of 2.7%. After deduplication, 1,906 terms remained for testing. The overall precision was 63.65%, recall was 67.34%, and F1 was 65.44%, significantly outperforming traditional annotation tools (F1: 0.45-0.49, P < 0.001). Although PhenoRAG's F1 was lower than that of RAG-HPO (F1 = 0.78, P < 0.001), which relies on a manually constructed synonym database of 54 000 entries plus the HPO dataset, it requires no additional dictionary maintenance and can be used without any background in computer programming. Moreover, after switching to the GPT-5 model, PhenoRAG exhibited no hallucination rate on the new dataset, and its F1 score significantly increased (P = 0.038).
CONCLUSION
Without constructing a synonym database, the PhenoRAG achieved high-accuracy automatic mapping from clinical text to standard HPO terms. It features a low usage threshold, free access, and a Chinese-language interface, and can directly serve rare disease diagnosis, genetic counseling, and research scenarios in China and worldwide, warranting further clinical promotion and multicenter validation.
Humans
;
Phenotype
;
Biological Ontologies
;
Language
;
Software
;
Large Language Models
2.Construction and validation of a clinical predictive model for early neurological deterioration in patients with mild acute ischemic stroke
Weilai LI ; Weihong WU ; Ying JI
Journal of Apoplexy and Nervous Diseases 2025;42(4):321-327
Objective To investigate the risk factors for early neurological deterioration in mild acute ischemic stroke,to construct a clinical predictive model,and to perform internal validation of this model. Methods A retrospective analysis was performed for 739 patients with mild acute ischemic stroke who were admitted to Department of Neurology,Kuntong Hospital of Zunhua,from October 2020 to December 2023,and they were randomly divided into a training set with 534 patients (72.3%) and a validation set with 205 patients (27.7%) at a ratio of 7∶3. Univariate and multivariate logistic regression analyses were performed for the training set to determine the risk factors for early neurological deterioration in mild acute ischemic stroke. A clinical predictive model was constructed,and internal validation was performed in terms of discriminatory ability,calibration,and clinical decision making. A nomogram was plotted. Results The multivariate logistic regression analysis showed that female sex (OR=1.87,95% CI 1.14~3.09,P=0.014),time window ≤6 hours (OR=3.10,95%CI 1.56~6.19,P=0.001),a baseline NIHSS score of 2 points (OR=3.72,95%CI 1.30~10.61,P=0.014),a baseline NIHSS score of 3 points (OR=4.24,95%CI 1.45~12.35,P=0.008),a TOAST classification of large artery atherosclerosis (OR=3.88,95%CI 2.20~6.83,P<0.001),and the responsible arteries of the basilar artery,the middle cerebral artery,and the internal carotid artery (OR=8.39,95%CI 2.28~30.85,P=0.001; OR=6.22,95%CI 1.78~21.71,P=0.004; OR=5.38,95%CI 1.15~25.13,P=0.032) were independent risk factors for early neurological deterioration in mild acute ischemic stroke. The clinical predictive model constructed showed a moderate discriminatory ability (AUC>0.7),good calibration (P>0.05) in the Hosmer-Lemeshow goodness-of-fit test),and good clinical benefits in both the training set and the validation set. Conclusion This clinical predictive model can effectively predict the onset of early neurological deterioration in mild acute ischemic stroke and guide clinicians to make decisions,and therefore,it holds promise for clinical application.
Nomograms
3.Predicting Invasive Non-mucinous Lung Adenocarcinoma IASLC Grading: A Nomogram Based on Dual-energy CT Imaging and Conventional Features.
Kaibo ZHU ; Liangna DENG ; Yue HOU ; Lulu XIONG ; Caixia ZHU ; Haisheng WANG ; Junlin ZHOU
Chinese Journal of Lung Cancer 2025;28(8):585-596
BACKGROUND:
Lung adenocarcinoma is an important pathohistologic subtype of non-small cell lung cancer (NSCLC). Invasive non-mucinous pulmonary adenocarcinomas (INMA) tend to have a poor prognosis due to their significant heterogeneity and diverse histologic components. Establishing a histologic grading system for INMA is crucial for evaluating its malignancy. In 2021, the International Association for the Study of Lung Cancer (IASLC) proposed that a new histological grading system could better stratify the prognosis of INMA patients. The aim of this study was to establish a visualized nomogram model to predict INMA IASLC grading preoperatively by means of dual-energy computed tomography (DECT), fractal dimension (FD), clinical features and conventional CT parameters.
METHODS:
A total of 112 patients with INMA who underwent preoperative DECT were retrospectively enrolled from March 2021 to January 2025. Patients were categorized into low-intermediate grade and high grade groups based on IASLC grading. The clinical characteristics and conventional CT parameters, including baseline features, biochemical markers, and serum tumor markers, were collected. DECT-derived parameters, including iodine concentration (IC), effective atomic number (eff-Z), and normalized IC (NIC), were collected and determined as NIC ratio (NICr) and fractal dimension (FD). Univariate analysis was employed to compare differences in conventional characteristics and DECT parameters between the two groups. Variables demonstrating statistical significance were subsequently incorporated into a multivariate Logistic regression analysis. A nomogram model integrating clinical data, conventional CT parameters, and DECT parameters was developed to identify independent predictors for IASLC grading of INMA. The discriminatory performance of the model was evaluated using receiver operating characteristic (ROC) curve analysis.
RESULTS:
Multivariate analysis identified smoking history [odds ratio (OR)=2.848, P=0.041], lobulation sign (OR=2.163, P=0.004), air bronchogram (OR=7.833, P=0.005), eff-Z in arterial phase (OR=4.266, P<0.001), and IC in arterial phase (OR=1.290, P=0.012) as independent and significant predictors for IASLC grading of INMA. The nomogram model constructed based on these indicators demonstrated optimal predictive performance, achieving an area under the curve (AUC) of 0.804 (95%CI: 0.725-0.883), with specificity and sensitivity of 85.3% and 65.7%, respectively.
CONCLUSIONS
The nomogram model based on clinical features, imaging features and spectral CT parameters have a large potential for application in the preoperative noninvasive assessment of INMA IASLC grading.
Humans
;
Nomograms
;
Female
;
Male
;
Middle Aged
;
Tomography, X-Ray Computed/methods*
;
Lung Neoplasms/pathology*
;
Aged
;
Retrospective Studies
;
Adenocarcinoma of Lung/pathology*
;
Neoplasm Grading
;
Adult
4.Clinical prediction model for patients with early-onset prostate cancer without surgical treatment: Based on the SEER Database.
Han-Dong LIU ; Han-Yu JIA ; Jing WANG ; Li-Ping ZHANG
National Journal of Andrology 2025;31(5):412-420
OBJECTIVE:
The aim of this study is to investigate the risk factors of prognosis in patients with early-onset prostate cancer treated without surgery. A nomogram will be constructed and validated to predict overall survival (OS) of patients with early-onset prostate cancer treated without surgery.
METHODS:
The clinical data was obtained from the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) database on prostate cancer patients aged 18-55 years who were treated without surgery between 2010 and 2015. The clinical data set was divided into training set and validation set according to 7∶3 ratio, including age, race, marital status, Gleason score, prostate specific antigen (PSA) and other 8 factors. And significant variables were screened by univariate Cox regression analysis. Multivariate Cox regression analysis was used to identify the influence factors. Stepwise regression method was used to select the most influential factors on the total OS, and R software was used to build a nomogram model. The accuracy and prediction ability of the model were verified by drawing receiver operating characteristic (ROC) and Calibration Plot. The clinical benefit of the model was evaluated by decision curve analysis (DCA).
RESULTS:
A total of 8 212 patients who met the criteria were randomly assigned to the training set (n=5 752) or validation set (n=2 460), with no statistical difference between the two groups (all P>0.05). Six factors were identified through univariate and multivariate Cox regression analysis including marital status, N stage, M stage, radiotherapy, PSA and Gleason score, which were most closely associated with the OS of prostate cancer patients, and a column graph model was constructed based on these factors. The Consistency index (C-index) of the model in the training set and the verification set were 0.802 and 0.794, respectively. And the apparent diffusion coefficient (AUC) was 0.851, 0.855 and 0.855 for training sets 1, 3 and 5 years, and 0.694, 0.860 and 0.832 for verification sets 1, 3 and 5 years. The calibration chart showed a good agreement between the predicted and actual values of the model. In the analysis of decision curve, the model showed good clinical application value.
CONCLUSION
The prediction model based on marital status, radiotherapy, M stage, N stage, PSA and Gleason score for early-onset prostate cancer patients without surgical treatment has certain reference value which is expected to become an effective tool for clinicians to treat in future prospective studies on large and multi-center samples.
Humans
;
Male
;
Prostatic Neoplasms/diagnosis*
;
Middle Aged
;
Nomograms
;
SEER Program
;
Prognosis
;
Adult
;
Prostate-Specific Antigen
;
Risk Factors
;
Proportional Hazards Models
;
Neoplasm Grading
;
ROC Curve
5.Predictive value of bpMRI for pelvic lymph node metastasis in prostate cancer patients with PSA≤20 μg/L.
Lai DONG ; Rong-Jie SHI ; Jin-Wei SHANG ; Zhi-Yi SHEN ; Kai-Yu ZHANG ; Cheng-Long ZHANG ; Bin YANG ; Tian-Bao HUANG ; Ya-Min WANG ; Rui-Zhe ZHAO ; Wei XIA ; Shang-Qian WANG ; Gong CHENG ; Li-Xin HUA
National Journal of Andrology 2025;31(5):426-431
Objective: The aim of this study is to explore the predictive value of biparametric magnetic resonance imaging(bpMRI)for pelvic lymph node metastasis in prostate cancer patients with PSA≤20 μg/L and establish a nomogram. Methods: The imaging data and clinical data of 363 patients undergoing radical prostatectomy and pelvic lymph node dissection in the First Affiliated Hospital of Nanjing Medical University from July 2018 to December 2023 were retrospectively analyzed. Univariate analysis and multivariate logistic regression were used to screen independent risk factors for pelvic lymph node metastasis in prostate cancer, and a nomogram of the clinical prediction model was established. Calibration curves were drawn to evaluate the accuracy of the model. Results: Multivariate logistic regression analysis showed extrocapusular extension (OR=8.08,95%CI=2.62-24.97, P<0.01), enlargement of pelvic lymph nodes (OR=4.45,95%CI=1.16-17.11,P=0.030), and biopsy ISUP grade(OR=1.97,95%CI=1.12-3.46, P=0.018)were independent risk factors for pelvic lymph node metastasis. The C-index of the prediction model was 0.834, which indicated that the model had a good prediction ability. The actual value of the model calibration curve and the prediction probability of the model fitted well, indicating that the model had a good accuracy. Further analysis of DCA curve showed that the model had good clinical application value when the risk threshold ranged from 0.05 to 0.70.Conclusion: For prostate cancer patients with PSA≤20 μg/L, bpMRI has a good predictive value for the pelvic lymph node metastasis of prostate cancer with extrocapusular extension, enlargement of pelvic lymph nodes and ISUP grade≥4.
Humans
;
Male
;
Prostatic Neoplasms/diagnostic imaging*
;
Lymphatic Metastasis
;
Retrospective Studies
;
Nomograms
;
Prostate-Specific Antigen/blood*
;
Lymph Nodes/pathology*
;
Pelvis
;
Predictive Value of Tests
;
Prostatectomy
;
Lymph Node Excision
;
Risk Factors
;
Magnetic Resonance Imaging
;
Logistic Models
;
Middle Aged
;
Aged
6.Quick accomplishment and responsiveness were associated with a lower risk of mortality from cardiovascular disease among Japanese older men: the Japan Collaborative Cohort Study.
Miyu MORIWAKI ; Kokoro SHIRAI ; Hironori IMANO ; Akiko TAMAKOSHI ; Ryo KAWASAKI ; Hiroyasu ISO
Environmental Health and Preventive Medicine 2025;30():15-15
BACKGROUND:
Quick accomplishment and responsiveness are behaviors related to time management by perceived control of time, such as a positive feeling of using one's time well. In recent years, positive psychological states have been associated with a lower risk of cardiovascular disease (CVD). Thus, we investigated the associations of quick accomplishment and responsiveness with CVD mortality in a large cohort study.
METHODS:
The study participants were 75,049 (30,901 men and 44,148 women) aged 40-79 between 1988 and 1990 and followed until the end of 2009. Hazard ratios (HRs) and 95% confidence intervals (CIs) of mortality from CVD according to quick accomplishment, responsiveness, and their combination were calculated after adjustment for potential confounding factors using the Cox proportional hazard model.
RESULTS:
Quick accomplishment was associated with a lower risk of CVD mortality in women; a similar but marginally significant association was observed in men; the respective multivariable HR (95%CI) was 0.91 (0.83-0.99) and 0.93 (0.86-1.01). The presence of both quick accomplishment and responsiveness was associated with lower risk in men, which was confined to men aged 60-79; the respective multivariable HR (95%CI) was 0.88 (0.78-0.99) and 0.83 (0.72-0.96).
CONCLUSIONS
Quick accomplishment was associated with a lower risk of CVD mortality. Quick accomplishment and responsiveness combined were inversely associated with CVD mortality risk among older men.
Adult
;
Aged
;
Female
;
Humans
;
Male
;
Middle Aged
;
Cardiovascular Diseases/psychology*
;
Cohort Studies
;
Japan/epidemiology*
;
Proportional Hazards Models
;
Risk Factors
;
East Asian People/psychology*
7.Relationship between sarcopenia and cardiovascular disease among middle-aged and older adults with normal weight in China: functional limitation plays a mediating role.
Hui CHENG ; Zhihui JIA ; Jiaheng CHEN ; Yao Jie XIE ; Jose HERNANDEZ ; Harry H X WANG
Environmental Health and Preventive Medicine 2025;30():46-46
BACKGROUND:
Cardiovascular disease (CVD) is the predominant cause of mortality in China. However, the mechanisms linking sarcopenia to CVD remain poorly understood, particularly in normal-weight populations. Individuals with the absence of overweight or obesity may tend to experience missed opportunities for timely intervention. This study aimed to investigate the longitudinal association between sarcopenia and incidence of new-onset CVD in a normal-weight population, and to examine the mediating effect of functional limitation in this relationship.
METHODS:
We conducted a closed-cohort analysis using a nationwide sample of 4,147 middle-aged and older adults with normal weight in China. We performed Cox proportional hazards regression analysis to explore the associations of baseline sarcopenia with incident CVD. The difference method was applied to estimate the mediation proportion of functional limitation in this association.
RESULTS:
Over a mean follow-up period of 7.62 years, CVD occurred in 835 participants. In the multivariable-adjusted Cox model, individuals with sarcopenia exhibited a significantly higher likelihood of developing incident CVD compared to those without sarcopenia (adjusted hazard ratio [aHR] = 1.45, 95% confidence interval [CI]: 1.21-1.73, P < 0.001). Similar associations were observed for the incidence of heart disease and stroke. Functional limitation accounted for approximately 15.0% of the total effect of sarcopenia on incident CVD (P < 0.001).
CONCLUSIONS
Sarcopenia exerts both direct and indirect effects on incident CVD among middle-aged and older adults who are normal weight, with functional limitation serving as a significant mediator. Interventions targeting both sarcopenia and functional limitation may offer a promising strategy for enhancing cardiovascular health in this population.
Humans
;
Sarcopenia/complications*
;
China/epidemiology*
;
Male
;
Female
;
Middle Aged
;
Cardiovascular Diseases/etiology*
;
Aged
;
Incidence
;
Cohort Studies
;
Proportional Hazards Models
;
Risk Factors
;
Aged, 80 and over
;
Longitudinal Studies
8.Expression and regulatory mechanism of miR-34a in neonatal rat model of bron-chopulmonary dysplasia induced by hyperoxia.
Mengyue HUO ; Hua MEI ; Yuheng ZHANG ; Yanbo ZHANG ; Chunli LIU
Journal of Peking University(Health Sciences) 2025;57(2):237-244
OBJECTIVE:
To investigate the expression and possible regulatory mechanism of miR-34a in the lung tissue of neonatal rat model of bronchopulmonary dysplasia (BPD) induced by hyperoxia.
METHODS:
In the study, 80 newborn SD rats were randomly divided into hyperoxia group (FiO2=60%) and air group (FiO2=21%) within 2 hours after birth, 40 rats per group. Lung tissue samples of the SD rats in each group were extracted on the 1st, 7th, 14th and 21st days after birth, and the pathological changes of lung tissue were observed under light microscope after HE staining. The number of radial alveolar counts (RAC) and the mean alveolar diameter (MAD) and the thickness of alveolar septal thickness (AST) were measured to evaluate the development of alveoli. Real-time fluorescence quantitative PCR was used to detect the expression of miR-34a, angiopoietin-1 (Ang-1) and tyrosine kinase receptor-2 (Tie-2) in lung tissue of rats in hyperoxia group and air group at different time points. Enzyme-linked immunosorbent assay (ELISA) was used to detect the proteins expression of Ang-1 and Tie-2 in the lung tissues of the two groups at different time points.
RESULTS:
The weight of rats in the hyperoxia group on the 7th, 14th and 21st days after birth was significantly lower than that in the air group (P all < 0.05). With the prolongation of oxygen exposure, the number of alveoli decreased, the volume increased, the structure simplified, the alveolar cavity enlarged obviously and the alveolar septum thickened in the hyperoxia group. On the 7th, 14th and 21st days after birth, the RAC in the hyperoxia group was significantly lower than that in the air group (P all < 0.05). Compared with the air group, MAD and AST increased significantly on the 7th, 14th and 21st days after birth in the hyperoxia group, and the difference was statistically significant (P all < 0.05). The expression level of miR-34a in lung tissue of hyperoxia group was significantly higher than that of air group on the 7th, 14th and 21st days after birth, and the difference was statistically significant (P all < 0.05). Compared with the air group at the same time point, the expression levels of Ang-1 and Tie-2 mRNA and protein in the hyperoxia group were lower than those in the air group on the 14th and 21st days after birth (P all < 0.05).
CONCLUSION
The new BPD model of newborn SD rats can be successfully established by continuous exposure to 60% hyperoxia. The expression of miR-34a was up-regulated in the lung tissue of the new BPD model of neonatal rats. MiR-34a may play an important role in the occurrence and development of BPD by regulating Ang-1/Tie-2 signal pathway.
Animals
;
MicroRNAs/metabolism*
;
Bronchopulmonary Dysplasia/genetics*
;
Hyperoxia/metabolism*
;
Rats, Sprague-Dawley
;
Animals, Newborn
;
Rats
;
Angiopoietin-1/genetics*
;
Disease Models, Animal
;
Receptor, TIE-2/genetics*
;
Lung/pathology*
;
Male
9.Application Status of Machine Learning in Assisted Diagnosis Techniques of Cardiovascular Diseases.
Pinliang LIAO ; Zihong WANG ; Miao TIAN ; Hong CHAI ; Xiaoyu CHEN
Chinese Journal of Medical Instrumentation 2025;49(1):24-34
In recent years, cardiovascular disease has become a common disease. With the development of machine learning and big data technologies, the processing ability of electrocardiogram (ECG) signals has been greatly enhanced through new computer technologies, enabling the auxiliary diagnosis technology for cardiovascular disease (CVD) to achieve new improvements. This article discusses the application of machine learning in ECG processing, especially in the auxiliary diagnosis of diseases. Firstly, the conventional signal preprocessing methods are introduced, and then the EEG signal processing methods based on feature extraction and fuzzy classification are explored. Secondly, the application of auxiliary diagnosis in CVD is further summarized. Finally, the advantages and disadvantages of the two methods are analyzed, and based on this, a design of an auxiliary diagnostic system compatible with the two methods is proposed, providing a new perspective for similar applied researches in the future.
Machine Learning
;
Cardiovascular Diseases/diagnosis*
;
Humans
;
Electrocardiography
;
Signal Processing, Computer-Assisted
;
Diagnosis, Computer-Assisted
;
Fuzzy Logic
;
Electroencephalography
10.Comparative Study of Seven New Dressings in Promoting Chronic Wound Healing in db/db Mice.
Qiuyun FENG ; Jia KE ; Danning QI ; Lei ZHOU ; Haiguang CHAI
Chinese Journal of Medical Instrumentation 2025;49(3):295-301
This study evaluated the healing-promoting effect and applicability of seven new dressings in chronic wounds. A chronic wound model was established using 48 db/db diabetic mice, which were randomly divided into 8 groups (control, polymer film, alginate, foam, hydrocolloid, hydrogel, carbon fiber, and silver dressing groups). Regular monitoring was conducted on the 5, 10, 15, and 20 days after surgery, and a comprehensive evaluation was performed based on healing rate, characteristic of histopathology, and semi-quantitative scoring. The results showed that, except for the polymer film dressing group, all other dressing groups had significantly better healing-promoting effect than the control group ( P<0.05), with the hydrocolloid, carbon fiber, and silver dressing groups demonstrated particularly outstanding efficacy. This study systematically compared the efficacy differences of seven dressings, and combined them with the adhesion, exudate volume and infection risks to provide a scientific basis for clinical dressing selection.
Animals
;
Mice
;
Wound Healing
;
Bandages
;
Male
;
Diabetes Mellitus, Experimental
;
Disease Models, Animal

Result Analysis
Print
Save
E-mail