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

In recent years, with continuous advancements in molecular biology and gene testing technologies, the diagnosis and treatment of colorectal cancer have been rapidly transitioning toward precision medicine. The application of molecular classification, target detection, and liquid biopsy technologies has driven ongoing updates to clinical guidelines. Multidisciplinary team colla-boration, innovations in precision surgical techniques, and the widespread adoption of neoadjuvant combination therapies have collectively promoted more individualized and scientific management of colorectal cancer. Looking ahead,the authors believe that as multi-omics biomarkers, organoid models, and artificial intelligence are increasingly integrated into clinical practice, precision diagnosis and treatment of colorectal cancer will deepen further, offering patients more efficient and personalized therapeutic options.

Objective:To analyze the clinical characteristics of locally advanced rectal cancer patients with pathological complete response (pCR) after neoadjuvant chemoradiotherapy combined with immunotherapy.Methods:The retrospective cohort study was conducted. The clinicopatholo-gical data of 46 patients with locally advanced rectal cancer who were admitted to 6 medical centers, including Beijing Friendship Hospital of Capital Medical University et al, from June 2021 to November 2022 were collected. There were 29 males and 17 females, aged (61±4)years. Patients received neoadjuvant chemoradiotherapy combined with immune checkpoint inhibitor therapy, and under-went radical total mesorectal excision during 6-12 weeks after radiotherapy. Observation indicators: (1) comparison of clinical characteristics between pCR and non-pCR patients;(2) postoperative complications and adverse reactions of pCR and non-pCR patients. Comparison of measurement data with normal distribution between groups was conducted using the t test. Comparison of measurement data with skewed distribution between groups was conducted using the Mann-Whitney U test. Comparison of count data between groups was conducted using the chi-square test or Fisher exact probability. Comparison of ordinal data between groups was conducted using the Mann-Whitney U test. Results:(1) Comparison of clinical characteristics between pCR and non-pCR patients. Before neoadjuvant therapy, there were 14 cases aged ≥50 years and 6 cases aged <50 years in pCR patients, versus 25 cases and 1 case in non-pCR patients, showing a significant difference between the two groups ( P<0.05). After neoadjuvant therapy, cases in clinical stage T0, T1, T2, T3, T4 were 11, 1, 5, 3, 0 for pCR patients versus 7, 4, 2, 11, 2 for non-pCR patients, cases of tumor regression grade 1, 2, 3, 4 were 11, 8, 1, 0 for pCR patients versus 7, 14, 4, 1 for non-pCR patients, cases in low-risk, medium-risk, high-risk of neoadjuvant rectal scoring and grading were 20, 0, 0 for pCR patients versus 4, 18, 4 for non-pCR patients, respectively, showing significant differences in above indicators between the two groups ( Z=-2.256, -2.104, -5.458, P<0.05). (2) Postoperative complications and adverse reactions of pCR and non-pCR patients. Postoperative complications occurred in 2 cases of pCR patients and 5 cases of non-pCR patients, postoperative adverse reactions occurred in 11 cases of pCR patients and 10 cases of non-pCR patients, showing no significant difference between the two groups ( P>0.05). Conclusion:Compared with locally advanced rectal cancer patients aged ≥50 years, those aged <50 years have significant benefits from neoadjuvant chemoradiotherapy combined with immunotherapy. Clinical T staging and magnetic resonance imaging-detected tumor regression grade after neoadjuvant therapy have predictive value for patients with pCR .

Objective To observe the value of machine learning(ML)models based on ultrasound radiomics for preoperatively distinguishing atypical parathyroid tumor(APT)/parathyroid carcinoma(PC)and parathyroid adenoma(PA).Methods Totally 330 primary hyperparathyroidism patients who underwent surgical treatments were retrospectively enrolled and categorized into APT/PC group(n=78)and PA group(n=252)according to surgical pathology and clinical follow-up results,also divided into training set(n=231)and test set(n=99)at the ratio of 7∶3.Based on preoperative ultrasound,545 radiomics features were extracted,and recursive feature elimination(RFE),Kruskal-Wallis or analysis of variance methods were used to screen the features,respectively.Support vector machine(SVM),linear discriminant analysis(LDA),least absolute shrinkage and selection operator logistic regression(LRLASSO),also random forest(RF)and decision tree(DT)algorithms were adopted to construct ML models for differentiating APT/PC and PA,respectively.Then the models were trained in training set,their performance were verified in test set,and a 5-fold cross-validation was adopted to screen out the better combinations.Results Compared with Kruskal-Wallis and analysis of variance methods,the distinguishing efficacy of SVM,LDA,LRLASSO,RF and DT models constructed based on features screened out using RFE method in training set(area under the curve[AUC]=0.870,0.878,0.850,0.847,1.000)and test set(AUC=0.856,0.842,0.827,0.847 and 0.704)were all relatively higher.In test set,the AUC of SVM,LDA,LRLASSO and RF models constructed based on the features screened out using RFE method(included 25,23,17 and 23 features)were all higher than that of DT model(8 features)(all P＜0.001).No significant difference of AUC was found between SVM,LRLASSO or RF models and LDA model(all P＞0.05).The AUC of SVM and RF models were higher than that of LRLASSO model(both P＜0.05),while of SVM and RF models were not significantly different(P＞0.05),indicating that SVM,LDA and RF models were better ones.Conclusion SVM,LDA,LRLASSO,RF and DT models based on ultrasound radiomics could effectively distinguish APT/PC and PA preoperatively,among which SVM,LDA and RF models had better diagnostic efficacy.

In recent years, with continuous advancements in molecular biology and gene testing technologies, the diagnosis and treatment of colorectal cancer have been rapidly transitioning toward precision medicine. The application of molecular classification, target detection, and liquid biopsy technologies has driven ongoing updates to clinical guidelines. Multidisciplinary team colla-boration, innovations in precision surgical techniques, and the widespread adoption of neoadjuvant combination therapies have collectively promoted more individualized and scientific management of colorectal cancer. Looking ahead,the authors believe that as multi-omics biomarkers, organoid models, and artificial intelligence are increasingly integrated into clinical practice, precision diagnosis and treatment of colorectal cancer will deepen further, offering patients more efficient and personalized therapeutic options.

Objective:To analyze the clinical characteristics of locally advanced rectal cancer patients with pathological complete response (pCR) after neoadjuvant chemoradiotherapy combined with immunotherapy.Methods:The retrospective cohort study was conducted. The clinicopatholo-gical data of 46 patients with locally advanced rectal cancer who were admitted to 6 medical centers, including Beijing Friendship Hospital of Capital Medical University et al, from June 2021 to November 2022 were collected. There were 29 males and 17 females, aged (61±4)years. Patients received neoadjuvant chemoradiotherapy combined with immune checkpoint inhibitor therapy, and under-went radical total mesorectal excision during 6-12 weeks after radiotherapy. Observation indicators: (1) comparison of clinical characteristics between pCR and non-pCR patients;(2) postoperative complications and adverse reactions of pCR and non-pCR patients. Comparison of measurement data with normal distribution between groups was conducted using the t test. Comparison of measurement data with skewed distribution between groups was conducted using the Mann-Whitney U test. Comparison of count data between groups was conducted using the chi-square test or Fisher exact probability. Comparison of ordinal data between groups was conducted using the Mann-Whitney U test. Results:(1) Comparison of clinical characteristics between pCR and non-pCR patients. Before neoadjuvant therapy, there were 14 cases aged ≥50 years and 6 cases aged <50 years in pCR patients, versus 25 cases and 1 case in non-pCR patients, showing a significant difference between the two groups ( P<0.05). After neoadjuvant therapy, cases in clinical stage T0, T1, T2, T3, T4 were 11, 1, 5, 3, 0 for pCR patients versus 7, 4, 2, 11, 2 for non-pCR patients, cases of tumor regression grade 1, 2, 3, 4 were 11, 8, 1, 0 for pCR patients versus 7, 14, 4, 1 for non-pCR patients, cases in low-risk, medium-risk, high-risk of neoadjuvant rectal scoring and grading were 20, 0, 0 for pCR patients versus 4, 18, 4 for non-pCR patients, respectively, showing significant differences in above indicators between the two groups ( Z=-2.256, -2.104, -5.458, P<0.05). (2) Postoperative complications and adverse reactions of pCR and non-pCR patients. Postoperative complications occurred in 2 cases of pCR patients and 5 cases of non-pCR patients, postoperative adverse reactions occurred in 11 cases of pCR patients and 10 cases of non-pCR patients, showing no significant difference between the two groups ( P>0.05). Conclusion:Compared with locally advanced rectal cancer patients aged ≥50 years, those aged <50 years have significant benefits from neoadjuvant chemoradiotherapy combined with immunotherapy. Clinical T staging and magnetic resonance imaging-detected tumor regression grade after neoadjuvant therapy have predictive value for patients with pCR .

Objective To observe the value of machine learning(ML)models based on ultrasound radiomics for preoperatively distinguishing atypical parathyroid tumor(APT)/parathyroid carcinoma(PC)and parathyroid adenoma(PA).Methods Totally 330 primary hyperparathyroidism patients who underwent surgical treatments were retrospectively enrolled and categorized into APT/PC group(n=78)and PA group(n=252)according to surgical pathology and clinical follow-up results,also divided into training set(n=231)and test set(n=99)at the ratio of 7∶3.Based on preoperative ultrasound,545 radiomics features were extracted,and recursive feature elimination(RFE),Kruskal-Wallis or analysis of variance methods were used to screen the features,respectively.Support vector machine(SVM),linear discriminant analysis(LDA),least absolute shrinkage and selection operator logistic regression(LRLASSO),also random forest(RF)and decision tree(DT)algorithms were adopted to construct ML models for differentiating APT/PC and PA,respectively.Then the models were trained in training set,their performance were verified in test set,and a 5-fold cross-validation was adopted to screen out the better combinations.Results Compared with Kruskal-Wallis and analysis of variance methods,the distinguishing efficacy of SVM,LDA,LRLASSO,RF and DT models constructed based on features screened out using RFE method in training set(area under the curve[AUC]=0.870,0.878,0.850,0.847,1.000)and test set(AUC=0.856,0.842,0.827,0.847 and 0.704)were all relatively higher.In test set,the AUC of SVM,LDA,LRLASSO and RF models constructed based on the features screened out using RFE method(included 25,23,17 and 23 features)were all higher than that of DT model(8 features)(all P＜0.001).No significant difference of AUC was found between SVM,LRLASSO or RF models and LDA model(all P＞0.05).The AUC of SVM and RF models were higher than that of LRLASSO model(both P＜0.05),while of SVM and RF models were not significantly different(P＞0.05),indicating that SVM,LDA and RF models were better ones.Conclusion SVM,LDA,LRLASSO,RF and DT models based on ultrasound radiomics could effectively distinguish APT/PC and PA preoperatively,among which SVM,LDA and RF models had better diagnostic efficacy.

Objective:To establish a nomogram for preoperative differentiating intrahepatic cholangiocarcinoma (ICC) from hepatocellular carcinoma (HCC) based on clinical, ultrasound, and contrast-enhanced ultrasound (CEUS) data.Methods:A retrospective analysis was conducted on ultrasound and CEUS data of 462 patients who underwent hepatectomy in Nanjing Drum Tower Hospital, the Affiliated Hospital of Nanjing University Medical School from January 2016 to December 2023, including 262 cases of HCC (56.7%) and 200 cases of ICC (43.3%). The data were randomly divided into training set ( n=324) and validation set ( n=138) in a 7∶3 ratio. Univariate analysis was used to initially screen for variables with statistically significant differences between HCC and ICC groups in the training set, and LASSO regression was performed to select the variables with higher coefficients. Logistic regression analyses were then used to predict independent risk factors for ICC. A nomogram was drawn using R software. The performance of the nomogram was then validated using ROC curve, calibration curve, and decision curve analysis (DCA). Results:Univariate analysis showed that there were significant differences in age, gender, liver cirrhosis, HBsAg (+ ), ALP >185 U/L, CA19-9 >27 kU/L, CA242>10 kU/L, irregular shape, border, cholangiectasis, portal vein tumor thrombus, enhanced pattern in arterial phase, clearance time <60 s, intra-tumoral vein between ICC and HCC groups (all P<0.05). The top 10 features were selected for LASSO regression analysis. Logistic regression analysis revealed that gender, cirrhosis, CA19-9>27 kU/L, CA242>10 kU/L, cholangiectasis, clearance time <60 s, intra-tumoral vein and enhanced pattern in arterial phase were risk factors for ICC (all P<0.05). The area under the ROC curve in the training and validation groups were 0.963 and 0.914, respectively. In the training group, the specificity and sensitivity of the nomogram were 0.926 and 0.917, respectively, and in the validation group, they were 0.875 and 0.871, respectively. The calibration curve showed that the prediction effect of the model was in good agreement with the actual situation. DCA showed that the nomogram could increase the net benefit to the different diagnosis of ICC in patients. Conclusions:The nomogram based on clinical, ultrasound and CEUS features has a good predictive value for preoperative identification of ICC and provides reliable evidence for clinical practice.

Objective:To establish and evaluate a clinical and ultrasound parameters-based nomogram for the preoperative differentiating diagnosis of intrahepatic cholangiocarcinoma (ICC).Methods:A total of 723 patients undergoing hepatectomy in Nanjing Drum Tower Hospital, Affiliated Hospital of Medical School, Nanjing University from January 2016 to August 2022 were retrospectively screened. A total of 399 patients with hepatocellular carcinoma (HCC, 198 cases) or ICC (201 cases) were enrolled in this study, including 284 males and 115 females, aged (60.5±10.5) years. Through random sampling using computer-generated random numbers, patients were divided into training ( n=279) and validation groups ( n=120) in a ratio of 7∶3. Univariate and multivariate logistic regression were performed to identify factors differentiating ICC, and a nomogram was established using R software based on independent risk factors for ICC. The accuracy of the nomogram was evaluated by receiver operating characteristic curve and calibration curves. Decision curve analysis was performed to assess the net benefit of the model. Results:Multivariate logistic regression analysis showed that irregular shape, cholangiectasis, female, cirrhosis, carbohydrate antigen 242 >10 U/ml, carbohydrate antigen 125 >30 U/ml and alpha-fetoprotein >10 μg/L were independent differentiating factors for ICC (all P<0.05). A nomogram was constructed based on those factors. The nomogram showed a better discrimination between ICC and HCC. The area under the curve of the training group and the validation group were 0.966 and 0.956, respectively. The calibration curve showed that the prediction effect of the model is in good agreement with the actual situation. Decision curve analysis showed that the nomogram was more effective than diagnosing all patients as either HCC or ICC, which yielded a net benefit at the most reasonable threshold probabilities. Conclusion:The nomogram for the preoperative diagnosis of ICC based on clinical and ultrasound features showed a good diagnostic performance.

Objective:To investigate the application value of magnetic resonance imaging(MRI) in evaluating the efficacy of anti-PD-1 combined with neoadjuvant therapy for microsatellite stability (MSS)/proficient mismatch repair (pMMR) locally advanced rectal cancer (LARC).Methods:The prospective single-arm phase Ⅱ study was conducted. The clinicopathological data of 37 patients with MSS/pMMR LARC who were admitted to Beijing Friendship Hospital of Capital Medical University from April 2021 to September 2022 were collected. All patients underwent anti-PD-1 combined with neoadjuvant therapy and radical total mesorectal excision. Observation indicators: (1) enrolled pati-ents; (2) MRI and pathological examination; (3) concordance analysis of MRI examination reading; (4) evaluation of MRI examination. Measurement data with normal distribution were represented as Mean± SD. Count data were expressed as absolute numbers or percentages. Linear weighted κ value was used to evaluate the concordance of radiologist assessment. Sensitivity, negative predictive value, accuracy, overstaging rate and understaging rate were used to evaluate the predictive value. Results:(1) Enrolled patients. A total of 37 eligible patients were screened out, including 21 males and 16 females, aged (61±11)years. MRI examination was performed before and after combined therapy, and pathological examination was performed after radical resection. (2) MRI and pathological examination of patients. Among the 37 patients, MRI before combined therapy showed 0, 0, 5, 24 and 8 cases in stage T0, T1, T2, T3 and T4, 10, 17 and 10 cases in stage N0, N1 and N2, 28 and 9 cases of positive and negative extramural vascular invasion (EMVI), 4 and 33 cases of positive and negative mesorectal fascia (MRF), respectively. MRI examination after combined therapy showed 15, 4, 7, 10 and 1 cases in stage T0, T1, T2, T3 and T4, 34, 2 and 1 cases in stage N0, N1 and N2, 9 and 28 cases of positive and negative EMVI, 1 and 36 cases of positive and negative MRF. There were 16, 13, 8 and 0 cases of tumor regression grading (TRG) 0, 1, 2 and 3, respectively. Postoperative pathological examination showed 18, 4, 3, 11, 1 cases in stage T0, T1, T2, T3, T4, 33, 3, 1 cases in stage N0, N1, N2, positive and negative EMVI and unknown data in 1, 35, 1 cases, positive and negative circumferential margin in 0 and 37 cases, grade 0, grade 1, grade 2, grade 3 of American Joint Committee on Cancer TRG in 18, 9, 8, 2 cases, respectively. Pathological complete response rate was 48.6%(18/37) and approximate pathological complete response rate was 24.3%(9/37). (3)Concordance analysis of MRI examination reading. The κ value of T staging and N staging on MRI before combined therapy was 0.839 ( P<0.05) and 0.838 ( P<0.05), respectively. The κ value of T staging and N staging on MRI after combined therapy was 0.531 ( P<0.05) and 0.846 ( P<0.05), respectively. The κ value of EMVI and MRF was 0.708 ( P<0.05) and 0.680 ( P<0.05) before combined therapy, and they were 0.561 ( P<0.05) and 1.000 ( P<0.05) after combined therapy, respectively. The κ value of TRG 3-round reading for TRG was 0.448 ( P<0.05). (4) Evaluation of MRI examination. ① MRI evaluation of T and N staging. The accuracy of MRI examination after combined therapy for distinguishing stage T0 was 75.7%[28/37, 95% confidence interval ( CI) as 62.2%-89.2%], the understaging rate was 8.1%(3/37, 95% CI as 0-18.9%), the overstaging rate was 16.2%(6/37, 95% CI as 5.4%-29.7%). The accuracy of MRI examination for distinguishing stage T0-T2 was 86.5%(32/37, 95% CI as 73.0%-97.3%), its understaging rate and overstaging rate were 8.1%(3/37, 95% CI as 0-18.9%) and 5.4% (2/37, 95% CI as 0-13.5%), respectively. The accuracy of MRI examination for distinguishing N staging was 91.9%(34/37, 95% CI was 81.1%-100.0%), its understaging rate and overstaging rate were 5.4%(2/37, 95% CI as 0-13.5%) and 2.7%(1/37, 95% CI as 0-8.1%), respectively. Among 18 patients in pathological stage T0, the overstaging rate of MRI was 33.3%(6/18). All the 4 patients in pathological stage T1 and 3 pati-ents in pathological stage T2 had correct diagnosis. There were 3 cases with understaging among 12 patients in pathological stage T3-T4. Among the 37 patients in pathological stage N0-N2, 34 cases had correct diagnosis, 1 case was overstaged as stage N1 due to a round mesorectal lymph node with short diameter as 6 mm, and 2 cases were diagnosed as stage N0 due to the small lymph nodes with the maximum short diameter as 3 mm. ② MRI evaluation of EMVI and MRF. The accuracy, sensitivity and negative predictive value of MRI for evaluating EMVI were 86.5%(32/37, 95% CI as 75.0%-97.2%), 100.0% and 100.0%, respectively, and the overestimation rate of EMVI was 13.9%(5/36, 95% CI as 2.8%-25.0%), and no underestimation occurred. Of 35 pathologically negative EMVI patients, a rate of 14.3%(5/35) of patients were positive on MRI. The main reason for overestaging was that thickened fibrous tissue outside the rectal wall was mistaken for vascular invasion. The accuracy of MRI for evaluating MRF was 97.3%(36/37, 95% CI as 91.9%-100.0%), and 1 case (1/37, 2.7%, 95% CI as 0-8.1%) was overestimated as positive MRF due to misdiagnosis of pararectal MRF lymph nodes. The negative predictive value of MRI for assessing MRF was 100.0%. ③ MRI evaluation of TRG. The accuracy, understaging and overstaging rates of MRI for evaluating pathological TRG 0 were 78.4%(29/37, 95% CI as 64.9%-91.9%), 8.1%(3/37, 95% CI as 0-18.9%), 13.5%(5/37, 95% CI as 5.4%-27.0%), respectively. The accuracy, understaging and overstaging rates of MRI for evaluating pathological TRG 0-1 were 89.2%(33/37, 95% CI as 78.4%-97.3%), 8.1%(3/37, 95% CI as 0-18.9%), 2.7%(1/37, 95% CI as 0-8.1%), respectively. Of the 18 patients with pathologic complete response, 5 cases were diagnosed as pathological TRG 1 and 13 cases as pathological TRG 0. One near-pCR patient was assessed as pathological TRG 2. Two patients with pathological TRG 3 were incorrectly diagnosed on MRI. Conclusions:Anti-PD-1 combined with neoadjuvant therapy can downstage the LARC pati-ents with MSS/pMMR. MRI is effective in predicting T staging, N staging, EMVI, MRF and TRG. However, overstaging should be prevented.