Objective To explore the value of machine learning(ML)models based on non-contrast CT(NCCT)radiomics features for predicting the severity of acute phase traumatic brain injury(TBI).Methods Totally 600 TBI patients were retrospectively collected as observation group,other 65 TBI patients were taken as external validation set,while 50 TBI patients were prospectively enrolled as prospective validation set.Patients in observation group were divided into high-risk subgroup(n=240)and low-risk subgroup(n=360)according to Glasgow outcome scale(GOS)at discharge.The severity of acute phase TBI in observation group was assessed by doctor A and B with the same criteria,then an artificial model was established based on clinical and NCCT data at the time of first diagnosis using logistic regression(LR)method for predicting the severity of acute phase TBI.Patients in observation group were divided into training set(n=420,including 168 in high-risk subgroup and 252 in low-risk subgroup)and test set(n=180,including 72 in high-risk subgroup and 108 in low-risk subgroup)at the ratio of 7∶3.Based on NCCT of training set,radiomics features were extracted and selected,and LR,support vector machine(SVM),random forest(RF)and K-nearest neighbor(KNN)were used to establish 4 ML models.The efficacies of the above models were validated in test set,external validation set(including 34 cases of high-risk and 31 cases of low-risk TBI)and prospective validation set(including 21 cases of high-risk and 29 cases of low-risk TBI),respectively.Results The area under the curve(AUC)of doctor A and B for evaluating the severity of acute phase TBI in observation group was 0.606 and 0.771,respectively,of artificial model was 0.824.Based on NCCT in training set,6 optimal radiomics features were selected to construct LR,SVM,RF and KNN ML models,with AUC of 0.983,0.971,0.970 and 0.984 in test set,respectively,while the AUC of artificial model was 0.708.The AUC of LR,SVM,RF,KNN ML models and artificial model in external validation set was 0.879,0.881,0.984,0.863 and 0.733,while in prospective validation set was 0.984,0.873,0.982,0.897 and 0.704,respectively.Conclusion ML models based on CT radiomics could effectively predict the severity of acute phase TBI.

The treatment of hepatobiliary malignant tumor is characterized by the coexistence of multiple treatment methods and multiple disciplines. In order to evaluate the clinical efficacy of different treatment measures or multiple treatment combinations, and to promote the standardized development of comprehensive treatment patterns for hepatobiliary malignant tumor, the Hepatobiliary Center of the First Affiliated Hospital of Nanjing Medical University constructs the registry and follow-up database in hepatobiliary tumor patients based on the information-based platform of the hospital, which will help guide clinicians to make scientific decisions and improve the level of clinical diagnosis and treatment. This study describes the framework design, function modules, data acquisition process and quality control of the database of hepatobiliary malignant tumor. Based on the observational bidirectional cohort study design, the previous clinical data can be sorted to match the current database, on the other hand, the clinical data can be prospectively collected including basic information, admission evaluation, surgical information and postoperative situation, comprehensive treatment measures, regular reexaminations and long-term follow-up, etc. The data quality control system can be improved by formulating standardized operation procedures, regularly personnel training and full-process data management plans. This database will provide high-quality real-world data for clinicians, researchers, and guideline experts, and then provide high-level medical evidence for the standardized development of comprehensive treatment patterns of hepatobiliary malignancies.

Objective:To establish a model C-GALAD for detecting hepatocellular carcinoma (HCC) from the chronic liver disease and healthy people based on the serum markers.Methods:A clinical cohort including 229 hepatocellular carcinoma patients, 2 317 patients with chronic liver disease and 982 healthy people, was retrospectively collected from eight hospitals or physical examination institutions from April 2018 to October 2020. The data were divided into a training set and a testing set by stratified sampling with a 6∶4 ratio. A predictive model was established on the training set using a logistic backward regression method and validated on the testing set. In addition, clinical data from March to July 2021 in Beijing You′ an Hospital affiliated to Capital Medical University, including 84 patients with liver cancer and 204 patients with chronic liver disease collected were used for external independent validation of the model. The receiver operating characteristic curve (ROC) area under curve (AUC), the sensitivity and the specificity were used to evaluate the effectiveness of the model.Results:Through the logistic backward regression method, the seven signatures including age, gender, alpha-fetoprotein (AFP), alpha-fetoprotein alloplasm-3 ratio (AFP-L3%), des-gamma-carboxyprothrombin(DCP), platelet (PLT) and total bilirubin (TBIL) were selected as risk factors in the detection model. The area under the ROC curve (AUC) of the model on the testing set was 0.954, with an 88.04% sensitivity and a 94.85% specificity, and the AUC of model on the external independent validation set was 0.943, with an 89.29% sensitivity and a 90.2% specificity, which were better than other published models.Conclusion:The C-GALAD Ⅱ model can accurately predict the risk of hepatocellular carcinoma occurrence, and thus provide a trustworthy diagnosis method of hepatocellular carcinoma.

Objective:To study the surgical safety and efficacy of preoperative neoadjuvant therapy with immune checkpoint inhibitors combined with anti-angiogenic drugs in patients with China liver cancer staging(CNLC)-Ⅱb and Ⅲa resectable hepatocellular carcinoma.Methods:The data of 129 patients with Ⅱb and Ⅲa hepatocellular carcinoma who underwent surgery at the First Affiliated Hospital of Nanjing Medical University from January 2018 to December 2020 were analyzed. All patients were divided into two groups: the neoadjuvant therapy group( n=14,13 males and 1 female,aged (55.4±12.6)years(range:34 to 75 years)) received immune combined targeted therapy before surgery,immune checkpoint inhibitor camrelizumab was administered intravenously at a dose of 200 mg each time,every 2 weeks for 3 cycles,anti-angiogenesis drug apatinib was taken orally and continuously with a dose of 250 mg for 3 weeks and the conventional surgery group( n=115,103 males and 12 females,aged (55.8±12.0)years(range:21 to 83 years)) did not receive antitumor systemic therapy before surgery. There were 3 patients with CNLC-Ⅱb,11 with CNLC-Ⅲa in the neoadjuvant group;28 patients with CNLC-Ⅱb,87 with CNLC-Ⅲa in the conventional group. Student′s t test or rank-sum test was used to compare the differences between two groups for quantitative data, Fisher′s exact probability method was used to compare the differences of proportions between two groups, and Log-rank test was used to compare survival differences between two groups. Results:The 1-year recurrence rate in the neoadjuvant group was 42.9%,and the 1-year recurrence rate in the conventional group was 64.0%,with a statistically significant difference between the two groups(χ2=3.850, P=0.050);The 1-year survival rate in the neoadjuvant group was 100% and that in the conventional group was 74.2%,with a statistically significant difference between the two groups(χ2=5.170, P=0.023). According to the stratified analysis of the number of tumors,for single tumor,the 1-year recurrence rate in the neoadjuvant group was 25.0%,and that in the conventional surgery group was 71.0%,and the difference between the two groups was statistically significant(χ2=5.280, P=0.022). For multiple tumors, the 1-year recurrence rate in the neoadjuvant group was 66.7%,and the 1-year recurrence rate in the conventional surgery group was 58.9%,with no significant difference between the two groups(χ2=0.110, P=0.736). The operative time,intraoperative blood loss,and postoperative hospital stay in the neoadjuvant group were similar to those in the conventional group,and their differences were not statistically significant. Conclusions:Immune checkpoint inhibitors combined with anti-angiogenic targeted drugs as a neoadjuvant therapy for resectable hepatocellular carcinoma can reduce the 1-year recurrence rate and improve the 1-year survival rate,especially for those with solitary tumor. Limited by the sample size of the neoadjuvant group,the safety of immune combined targeted therapy before surgery cannot be observed more comprehensively,and further studies will be explored.

Objective:To study the surgical safety and efficacy of preoperative neoadjuvant therapy with immune checkpoint inhibitors combined with anti-angiogenic drugs in patients with China liver cancer staging(CNLC)-Ⅱb and Ⅲa resectable hepatocellular carcinoma.Methods:The data of 129 patients with Ⅱb and Ⅲa hepatocellular carcinoma who underwent surgery at the First Affiliated Hospital of Nanjing Medical University from January 2018 to December 2020 were analyzed. All patients were divided into two groups: the neoadjuvant therapy group( n=14,13 males and 1 female,aged (55.4±12.6)years(range:34 to 75 years)) received immune combined targeted therapy before surgery,immune checkpoint inhibitor camrelizumab was administered intravenously at a dose of 200 mg each time,every 2 weeks for 3 cycles,anti-angiogenesis drug apatinib was taken orally and continuously with a dose of 250 mg for 3 weeks and the conventional surgery group( n=115,103 males and 12 females,aged (55.8±12.0)years(range:21 to 83 years)) did not receive antitumor systemic therapy before surgery. There were 3 patients with CNLC-Ⅱb,11 with CNLC-Ⅲa in the neoadjuvant group;28 patients with CNLC-Ⅱb,87 with CNLC-Ⅲa in the conventional group. Student′s t test or rank-sum test was used to compare the differences between two groups for quantitative data, Fisher′s exact probability method was used to compare the differences of proportions between two groups, and Log-rank test was used to compare survival differences between two groups. Results:The 1-year recurrence rate in the neoadjuvant group was 42.9%,and the 1-year recurrence rate in the conventional group was 64.0%,with a statistically significant difference between the two groups(χ2=3.850, P=0.050);The 1-year survival rate in the neoadjuvant group was 100% and that in the conventional group was 74.2%,with a statistically significant difference between the two groups(χ2=5.170, P=0.023). According to the stratified analysis of the number of tumors,for single tumor,the 1-year recurrence rate in the neoadjuvant group was 25.0%,and that in the conventional surgery group was 71.0%,and the difference between the two groups was statistically significant(χ2=5.280, P=0.022). For multiple tumors, the 1-year recurrence rate in the neoadjuvant group was 66.7%,and the 1-year recurrence rate in the conventional surgery group was 58.9%,with no significant difference between the two groups(χ2=0.110, P=0.736). The operative time,intraoperative blood loss,and postoperative hospital stay in the neoadjuvant group were similar to those in the conventional group,and their differences were not statistically significant. Conclusions:Immune checkpoint inhibitors combined with anti-angiogenic targeted drugs as a neoadjuvant therapy for resectable hepatocellular carcinoma can reduce the 1-year recurrence rate and improve the 1-year survival rate,especially for those with solitary tumor. Limited by the sample size of the neoadjuvant group,the safety of immune combined targeted therapy before surgery cannot be observed more comprehensively,and further studies will be explored.

Objective:To evaluate the effectiveness of hierarchical management for patients with bronchial asthma.Methods:One hundred and eighty seven patients with bronchial asthma were recruited from January 2018 to November 2019 in Daxing District People′s Hospital. Patients were randomly divided into two groups, 94 patients received disease management education and therapeutic guidance from doctors in the community hospital and district hospital (study group), and 93 patients were followed up in outpatient visits only (control group). After one year, the scores of inhalation technique, treatment adherence, disease management awareness, the Asthma Control Test (ACT), the Mini Asthma Quality of Life Questionnaire (MiniAQLQ) and pulmonary function were evaluated and compared between two groups. The annual acute attack times and time to first exacerbation were also compared between the two groups.Results:After one year of management the treatment adherence rate in study group was higher than that in control group [80.85% (76/94) vs. 51.61% (48/93), χ2=2.834, P=0.02]. The scores of inhaled corticosteroids (ICS) inhalation technique [(6.47±1.28) vs. (4.05±1.37), t=2.241, P=0.04], the correct rates of exhaling before ICS inhalation [94.68% (89/94) vs.56.98% (53/93), χ2=4.436, P=0.01],inhalation [90.43%(85/94) vs.68.82% (64/93),χ2=2.943, P=0.04],holding breath after inhalation [89.36% (84/94) vs.58.06% (54/93),χ2=4.098, P=0.02],rinsing mouth after ICS inhalation [92.55%(87/94) vs.65.59%(61/93),χ2=2.876, P=0.04] in study group were higher than those in control group. The awareness rates of chronic inflammatory airway disease [70.21%(66/94) vs.44.08% (41/93),χ2=2.673, P=0.02], causative factors [85.10% (80/94) vs. 56.99% (53/93),χ2=2.760, P=0.02],treatment misunderstanding [88.29%(83/94) vs.53.76%(50/93),χ2=4.874, P<0.01], therapeutic goal [86.17% (81/94) vs. 49.46% (46/93),χ2=4.491, P<0.01] and requiring long-term treatment [90.43% (85/94) vs.48.38% (45/93),χ2=4.503, P<0.01] in study group were higher than those in control group. The scores of ACT [(22.71±2.81) vs. (19.50±5.34), t=2.041, P=0.04] and miniAQLQ [(84.28±11.16) vs. (64.23±14.38), t=3.298, P<0.01] in study group were higher than those in control group. The number of annual acute exacerbation was less [0(0, 1) vs.2(1, 3), Z=-3.237, P<0.01] and the time to first exacerbation was longer [184(96, 284)d vs. 96(59, 177)d, Z=3.873, P<0.01] in study group than those in the control group after one year of management. Conclusion:The hierarchical management can effectively enhance the inhalation technique and treatment adherence of the patients with bronchial asthma, and improve the quality of life of patients.

Objective:To summarize the experience of surgical treatment of primary liver cancer.Methods:The clinical data of 10 966 surgically managed cases with primary liver cancer, from January 1986 to December 2019 at Hepatobiliary Center, the First Affiliated Hospital of Nanjing Medical University, were retrospectively analyzed. The life table method was used to calculate the survival rate and postoperative recurrence rate. Log‐rank test was used to compare the survival process of different groups, and the Cox regression model was used for multivariate analysis. In addition, 2 884 cases of hepatocellular carcinoma(HCC) with more detailed follow‐up data from 2009 to 2019 were selected for survival analysis. Among 2 549 patients treated with hepatectomy, there were 2 107 males and 442 females, with an age of (56.6±11.1) years (range: 20 to 86 years). Among 335 patients treated with liver transplantation, there were 292 males and 43 females, with an age of (51.0±9.7) years (range: 21 to 73 years). The outcomes of hepatectomy versus liver transplantation, anatomic versus non-anatomic hepatectomy were compared, respectively.Results:Of the 10 966 patients with primary liver cancer, 10 331 patients underwent hepatectomy and 635 patients underwent liver transplantation. Patients with liver resection were categorized into three groups: 1986－1995(712 cases), 1996－2008(3 988 cases), 2009?2019(5 631 cases). The 5‐year overall survival rate was 32.9% in the first group(1986－1995). The 5‐year overall survival rate of resected primary liver cancer was 51.7% in the third group(2009‐2019), among which the 5‐year overal survival rates of hepatocellular carcinoma, intrahepatic cholangiocarcinoma and mixed liver cancer were 57.4%, 26.6% and 50.6%, respectively. Further analysis was performed on 2 549 HCC patients with primary hepatectomy. The 1‐, 3‐, 5‐, and 10‐year overall survival rates were 88.1%, 71.9%, 60.0%, and 41.0%, respectively, and the perioperative mortality rate was 1.0%. Two hundred and forty‐seven HCC patients underwent primary liver transplantation, with 1‐, 3‐, 5‐, and 10‐year overall survival rates of 84.0%, 64.8%, 61.9%, and 57.6%, respectively. Eighty‐eight HCC patients underwent salvage liver transplantation, with the 1‐, 3‐, 5‐, and 10‐year overall survival rates of 86.8%, 65.2%, 52.5%, and 52.5%, respectively. There was no significant difference in survival rates between the two groups with liver transplantation ( P>0.05). Comparing the overall survival rates and recurrence rates of primary hepatectomy (2 549 cases) with primary liver transplantation (247 cases), the 1‐, 3‐, 5‐, and 10‐year overall survival rates in patients within Milan criteria treated with hepatectomy and transplantation were 96.3%, 87.1%, 76.9%, 54.7%, and 95.4%, 79.4%, 77.4%, 71.7%, respectively ( P=0.754). The 1‐, 3‐, 5‐year recurrence rates were 16.3%, 35.9%, 47.6% and 8.1%, 11.7%, 13.9%, respectively( P<0.01). The 1‐, 3‐, 5‐, 10‐year overall survival rates in patients with no large vessels invasion beyond the Milan criteria treated with liver resection and transplantation were 87.2%, 65.9%, 53.0%, 33.0% and 87.6%, 71.8%, 71.8%, 69.3%, respectively( P=0.003); the 1‐, 3‐, 5‐year recurrence rate were 39.2%, 57.8%, 69.7% and 29.7%, 36.7%, 36.7%, respectively ( P<0.01). The 1‐, 3‐, 5‐, and 10‐year overall survival rates in patients with large vessels invasion treated with liver resection and transplantation were 62.1%, 36.1%, 22.2%, 15.0% and 62.9%, 31.8%,19.9%, 0, respectively ( P=0.387); the 1‐, 3‐, 5‐year recurrence rates were 61.5%, 74.7%, 80.8% and 59.7%, 82.9%, 87.2%, respectively( P=0.909). Independent prognostic factors for both overall survival and recurrence‐free survival rates of HCC patients treated with liver resection included gender, neoadjuvant therapy, symptoms, AST, intraoperative or postoperative blood transfusion, tumor number, tumor size, cirrhosis, macrovascular invasion, microvascular invasion, and pathological differentiation. Propensity score matching analysis of 443 pairs further showed that there was no significant difference in overall survival rate between anatomical liver resection and non‐anatomical liver resection( P=0.895), but the recurrence rate of non‐anatomical liver resection was higher than that of anatomical liver resection( P=0.035). Conclusions:In the past decade, the overall survival rate of HCC undergoing surgical treatment is significantly higher than before. For HCC patients with good liver function reservation, surgical resection can be performed first, and salvage liver transplantation can be performed after recurrence. The effect of salvage liver transplantation is comparable to that of primary liver transplantation. As for the choice of liver resection approaches, non‐anatomical resection can reserve more liver tissue and can be selected as long as the negative margin is guaranteed.

Objective:To compare the performance of multiple machine learning algorithms in predicting recurrence after resection of early-stage hepatocellular carcinoma(HCC).Methods:Clinical data of 882 early-stage HCC patients who were admitted to the First Affiliated Hospital of Nanjing Medical University from May 2009 to December 2019 and treated with curative surgical resection were retrospectively collected. There were 701 males and 181 females,with an age of (57.3±10.5)years(range:21 to 86 years). All patients were randomly assigned in a 2∶1 ratio, the training dataset consisted of 588 patients and the test dataset consisted of 294 patients. The construction of machine learning-based prediction models included random survival forest(RSF),gradient boosting machine,elastic net regression and Cox regression model. The prediction accuracy of the model was measured by the concordance index(C-index). The prediction error of the model was measured by the integrated Brier score. Model fit was assessed by the calibration plot. The performance of machine learning models with that of rival model and HCC staging systems was compared. All models were validated in the independent test dataset.Results:Median recurrence-free survival was 61.7 months in the training dataset while median recurrence-free survival was 61.9 months in the validation dataset, there was no significant difference between two datasets in terms of recurrence-free survival( χ2=0.029, P=0.865). The RSF model consisted of 5 commonly used clinicopathological characteristics, including albumin-bilirubin grade,serum alpha fetoprotein,tumor number,type of hepatectomy and microvascular invasion. In both training and test datasets,the RSF model provided the best prediction accuracy,with respective C-index of 0.758(95% CI:0.725 to 0.791) and 0.749(95% CI:0.700 to 0.797),and the lowest prediction error,with respective integrated Brier score of 0.171 and 0.151. The prediction accuracy of RSF model for recurrence after resection of early-stage HCC was superior to that of other machine learning models,rival model(ERASL model) as well as HCC staging systems(BCLC,CNLC and TNM staging),with statistically significant difference( P<0.01). Calibration curves demonstrated good agreement between RSF model-predicted probabilities and observed outcomes.All patients could be stratified into low-risk,intermediate-risk or high-risk group based on RSF model;statistically significant differences among three risk groups were observed in both training and test datasets(all P<0.01). The risk stratification of RSF model was superior to that of TNM staging. Conclusion:The proposed RSF model assembled with 5 commonly used clinicopathological characteristics in this study can predict the recurrence risk with favorable accuracy that may facilitate clinical decision-support for patients with early-stage HCC.

Objective:To summarize the experience of surgical treatment of primary liver cancer.Methods:The clinical data of 10 966 surgically managed cases with primary liver cancer, from January 1986 to December 2019 at Hepatobiliary Center, the First Affiliated Hospital of Nanjing Medical University, were retrospectively analyzed. The life table method was used to calculate the survival rate and postoperative recurrence rate. Log‐rank test was used to compare the survival process of different groups, and the Cox regression model was used for multivariate analysis. In addition, 2 884 cases of hepatocellular carcinoma(HCC) with more detailed follow‐up data from 2009 to 2019 were selected for survival analysis. Among 2 549 patients treated with hepatectomy, there were 2 107 males and 442 females, with an age of (56.6±11.1) years (range: 20 to 86 years). Among 335 patients treated with liver transplantation, there were 292 males and 43 females, with an age of (51.0±9.7) years (range: 21 to 73 years). The outcomes of hepatectomy versus liver transplantation, anatomic versus non-anatomic hepatectomy were compared, respectively.Results:Of the 10 966 patients with primary liver cancer, 10 331 patients underwent hepatectomy and 635 patients underwent liver transplantation. Patients with liver resection were categorized into three groups: 1986－1995(712 cases), 1996－2008(3 988 cases), 2009?2019(5 631 cases). The 5‐year overall survival rate was 32.9% in the first group(1986－1995). The 5‐year overall survival rate of resected primary liver cancer was 51.7% in the third group(2009‐2019), among which the 5‐year overal survival rates of hepatocellular carcinoma, intrahepatic cholangiocarcinoma and mixed liver cancer were 57.4%, 26.6% and 50.6%, respectively. Further analysis was performed on 2 549 HCC patients with primary hepatectomy. The 1‐, 3‐, 5‐, and 10‐year overall survival rates were 88.1%, 71.9%, 60.0%, and 41.0%, respectively, and the perioperative mortality rate was 1.0%. Two hundred and forty‐seven HCC patients underwent primary liver transplantation, with 1‐, 3‐, 5‐, and 10‐year overall survival rates of 84.0%, 64.8%, 61.9%, and 57.6%, respectively. Eighty‐eight HCC patients underwent salvage liver transplantation, with the 1‐, 3‐, 5‐, and 10‐year overall survival rates of 86.8%, 65.2%, 52.5%, and 52.5%, respectively. There was no significant difference in survival rates between the two groups with liver transplantation ( P>0.05). Comparing the overall survival rates and recurrence rates of primary hepatectomy (2 549 cases) with primary liver transplantation (247 cases), the 1‐, 3‐, 5‐, and 10‐year overall survival rates in patients within Milan criteria treated with hepatectomy and transplantation were 96.3%, 87.1%, 76.9%, 54.7%, and 95.4%, 79.4%, 77.4%, 71.7%, respectively ( P=0.754). The 1‐, 3‐, 5‐year recurrence rates were 16.3%, 35.9%, 47.6% and 8.1%, 11.7%, 13.9%, respectively( P<0.01). The 1‐, 3‐, 5‐, 10‐year overall survival rates in patients with no large vessels invasion beyond the Milan criteria treated with liver resection and transplantation were 87.2%, 65.9%, 53.0%, 33.0% and 87.6%, 71.8%, 71.8%, 69.3%, respectively( P=0.003); the 1‐, 3‐, 5‐year recurrence rate were 39.2%, 57.8%, 69.7% and 29.7%, 36.7%, 36.7%, respectively ( P<0.01). The 1‐, 3‐, 5‐, and 10‐year overall survival rates in patients with large vessels invasion treated with liver resection and transplantation were 62.1%, 36.1%, 22.2%, 15.0% and 62.9%, 31.8%,19.9%, 0, respectively ( P=0.387); the 1‐, 3‐, 5‐year recurrence rates were 61.5%, 74.7%, 80.8% and 59.7%, 82.9%, 87.2%, respectively( P=0.909). Independent prognostic factors for both overall survival and recurrence‐free survival rates of HCC patients treated with liver resection included gender, neoadjuvant therapy, symptoms, AST, intraoperative or postoperative blood transfusion, tumor number, tumor size, cirrhosis, macrovascular invasion, microvascular invasion, and pathological differentiation. Propensity score matching analysis of 443 pairs further showed that there was no significant difference in overall survival rate between anatomical liver resection and non‐anatomical liver resection( P=0.895), but the recurrence rate of non‐anatomical liver resection was higher than that of anatomical liver resection( P=0.035). Conclusions:In the past decade, the overall survival rate of HCC undergoing surgical treatment is significantly higher than before. For HCC patients with good liver function reservation, surgical resection can be performed first, and salvage liver transplantation can be performed after recurrence. The effect of salvage liver transplantation is comparable to that of primary liver transplantation. As for the choice of liver resection approaches, non‐anatomical resection can reserve more liver tissue and can be selected as long as the negative margin is guaranteed.

Objective:To compare the performance of multiple machine learning algorithms in predicting recurrence after resection of early-stage hepatocellular carcinoma(HCC).Methods:Clinical data of 882 early-stage HCC patients who were admitted to the First Affiliated Hospital of Nanjing Medical University from May 2009 to December 2019 and treated with curative surgical resection were retrospectively collected. There were 701 males and 181 females,with an age of (57.3±10.5)years(range:21 to 86 years). All patients were randomly assigned in a 2∶1 ratio, the training dataset consisted of 588 patients and the test dataset consisted of 294 patients. The construction of machine learning-based prediction models included random survival forest(RSF),gradient boosting machine,elastic net regression and Cox regression model. The prediction accuracy of the model was measured by the concordance index(C-index). The prediction error of the model was measured by the integrated Brier score. Model fit was assessed by the calibration plot. The performance of machine learning models with that of rival model and HCC staging systems was compared. All models were validated in the independent test dataset.Results:Median recurrence-free survival was 61.7 months in the training dataset while median recurrence-free survival was 61.9 months in the validation dataset, there was no significant difference between two datasets in terms of recurrence-free survival( χ2=0.029, P=0.865). The RSF model consisted of 5 commonly used clinicopathological characteristics, including albumin-bilirubin grade,serum alpha fetoprotein,tumor number,type of hepatectomy and microvascular invasion. In both training and test datasets,the RSF model provided the best prediction accuracy,with respective C-index of 0.758(95% CI:0.725 to 0.791) and 0.749(95% CI:0.700 to 0.797),and the lowest prediction error,with respective integrated Brier score of 0.171 and 0.151. The prediction accuracy of RSF model for recurrence after resection of early-stage HCC was superior to that of other machine learning models,rival model(ERASL model) as well as HCC staging systems(BCLC,CNLC and TNM staging),with statistically significant difference( P<0.01). Calibration curves demonstrated good agreement between RSF model-predicted probabilities and observed outcomes.All patients could be stratified into low-risk,intermediate-risk or high-risk group based on RSF model;statistically significant differences among three risk groups were observed in both training and test datasets(all P<0.01). The risk stratification of RSF model was superior to that of TNM staging. Conclusion:The proposed RSF model assembled with 5 commonly used clinicopathological characteristics in this study can predict the recurrence risk with favorable accuracy that may facilitate clinical decision-support for patients with early-stage HCC.