Lung transplantation is a definitive treatment for end-stage lung disease and can significantly improve patient prognosis. However, postoperative complications such as infection, rejection, ischemia-reperfusion injury and chronic lung allograft dysfunction intertwine to form a complex pathological network, posing persistent challenges to long-term patient survival. In 2025, research teams worldwide have made systematic progress in the field of basic lung transplantation research. By integrating cutting-edge technologies including single-cell multi-omics, spatial transcriptomics and novel animal models, significant breakthroughs have been achieved in the evolutionary dynamics of drug-resistant infections, molecular mechanisms of immune regulation, programmed cell death, optimization of donor lung protection strategies and early warning of chronic lung allograft dysfunction. This article systematically reviews the representative advances in basic lung transplantation research worldwide in 2025, and deeply analyzes the implications of mechanistic breakthroughs for optimizing diagnosis and treatment strategies, aiming to anchor the direction for innovative breakthroughs and clinical translation in basic lung transplantation research.

The biological characteristics of hepatocellular carcinoma(HCC)lead to a high incidence of portal vein tumor thrombus.It progresses rapidly,and the opportunity for radical surgical resection can be lost in a short term,resulting in poor prognosis.Reasonable down-staging treatment is a research focus for patients with unresectable HCC complicated with portal vein tumor thrombus to achieve a resectable range.This article reviews the potential population of HCC complicated with portal vein tumor thrombus with conversion therapy for HCC,analyzes the application of hepatic artery infusion chemotherapy,transcatheter arterial chemoembolization,radiotherapy,regional and systemic drug therapy in conversion therapy,and points out that the combination of multidisciplinary comprehensive treatments is the key to improve the prognosis of HCC patients with portal vein tumor thrombus.

Background and Aims:Complex biliary stones are often accompanied by hilar adhesions,distorted anatomy,biliary variations,or biliary-enteric fistulas.Conventional laparoscopic surgery is limited in such settings,with high conversion and complication rates.Robotic surgical systems provide high-definition 3D vision and enhanced dexterity,potentially improving surgical outcomes.This study compared the clinical efficacy of robotic vs.laparoscopic surgery for complex biliary stones.Methods:A retrospective analysis was performed on 145 patients with complex biliary stones who underwent minimally invasive surgery at Fujian Provincial Hospital,Fuzhou University,from January 2023 to January 2025.Patients were divided into a robotic group(n=62)and a laparoscopic group(n=83).Intraoperative outcomes,postoperative complications,and recovery parameters were compared.Stone recurrence was assessed during follow-up.Results:Baseline characteristics were comparable between groups(all P＞0.05).The conversion rate was 0 in the robotic group vs.7.2％in the laparoscopic group(P=0.038).The postoperative complication rate was significantly lower in the robotic group(1.6％vs.10.7％,P=0.044).Time to first oral intake,first flatus,and postoperative hospital stay were all shorter in the robotic group(all P＜0.05).No stone recurrence was observed in either group during follow-up until June 2025.Conclusion:Robotic surgery is a safe and feasible approach for complex biliary stones,offering reduced conversion and complication rates and faster recovery compared with laparoscopy.It holds promise for wider clinical application in complex biliary disease.

Objective To investigate the mechanisms by which astaxanthin(AST)alleviates oxaliplatin(OXA)-induced neuropathic pain through antioxidant pathways so as to provide theoretical basis for clinical intervention.Methods Animal experiments:SD rats were divided into five groups(n=6):control group,OXA(4 mg/kg)group,OXA+Oil group,OXA+AST(5 mg/kg)group,and OXA+AST(10 mg/kg)group.Mechanical and cold pain thresholds were measured at day 0,7,14,and 21.Malondialdehyde(MDA)content and superoxide dismutase(SOD)activity in the dorsal root ganglia(DRG)were detected using the thiobarbituric acid(TBA)method and WST-1 assay,respectively.Western blotting was performed to analyze the expressions of Nrf2 and HO-1.Cell experiments:neuro-2a cells were divided into control group,OXA(50 μmol/L)group,AST(10 μmol/L)group,and OXA(50 μmol/L)+AST(10 μmol/L)group.Cells were treated with nerve growth factor(NGF,50 ng/mL)to induce growth,and morphological changes were observed under an inverted microscope.Intracellular reactive oxygen species(ROS)level and mitochondrial superoxide were measured using DCFH-DA fluorescent probe and MitoSOXTM red,respectively.Mitochondrial function was assessed by JC-1 assay.Western blotting was used to detect Nrf2 and HO-1 expressions.Results Animal experiments:① Mechanical and cold pain thresholds were reduced in OXA and OXA+Oil groups(P＜0.05),while AST significantly increased these thresholds in OXA-treated rats(P＜0.05).② SOD activity decreased while MDA content increased in the DRG of OXA-treated rats(P＜0.05).AST restored SOD activity and reduced MDA level(P＜0.05,P＜0.01).③ Western blotting showed elevated Nrf2 and HO-1 expressions in OXA group(P＞0.05),which were further upregulated by AST(P＜0.05,P＜0.01).Cell experiments:① OXA reduced the number of neurite-bearing cells and shortened the average neurite length(P＜0.05).Inverted microscopic observation revealed that AST intervention increased both parameters(P＜0.01,P＜0.001).② OXA increased intracellular and mitochondrial ROS fluorescence intensity(P＜0.05),which was attenuated by AST(P＜0.01).③ JC-1 assay revealed decreased mitochondrial membrane potential in OXA group(P＜0.01),which was partially reversed by AST(P＜0.05).④ Western blotting results showed that OXA upregulated Nrf2 and HO-1 expressions(P＜0.05,P＜0.01),and AST further enhanced their levels(P＜0.01).Conclusion AST alleviates OXA-induced neuropathic pain by promoting Nrf2/HO-1 expression,enhancing SOD activity,reducing lipid peroxidation and ROS production,and improving mitochondrial function.

Objective:To investigate the prognostic factors for esophageal cancer after radical resection and the application value of machine learning prediction model.Methods:The retrospective cohort study was conducted. The clinicopatholigical data of 406 esophageal cancer patients who were admitted to Qilu Hospital of Shandong University from January 2018 to March 2022 were collected. There were 357 males and 49 females, aged (64±8)years. All patients underwent radical resection of esophageal cancer. The 406 patients were randomly divided into a training set of 285 cases and a validation set of 121 cases at a 7∶3 ratio based on a random number table. The training set was used to construct prediction model, and the validation set was used to validate prediction model. Patients were divided into high-risk group and low-risk group based on risk scores. Observation indicators: (1) follow-up of patients and analysis of influencing factors for prognosis; (2) construction and validation of machine learning prediction models. Comparison of measurement data with normal distribution between groups was conducted using the independent sample 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. Comparison of ordinal data between groups was conducted using the rank sum test. The Kaplan-Meier method was used to calculate survival rate and plot survival curve, and the Log-rank test was used for survival analysis. The Cox proportional hazard regression model was used for univariate and multivariate analyses. Independent influencing factors were included, and data processing, machine learning model construction, and visualization were performed using R packages including random survival forest (RSF), gradient boosting machine (GBM), least absolute shrinkage and selection operator Cox regression (LASSO-Cox), Cox proportional hazards model boosting (CoxBoost), survival support vector machine (survivalsvm), extreme gradient boosting (XGBoost), supervised principal component analysis (SuperPC), and Cox partial least squares regression (plsRcox). Receiver operating characteristic (ROC) curves were drawn, and sensitivity, specificity, and area under the curve (AUC) were calculated. The Delong test was used to assess the differences in AUC among different models in the training set, and the time-dependent ROC was used to compare the predictive performance of different models. Calibration curves were used to evaluate model accuracy, and decision curve analysis (DCA) was used to evaluate overall net benefit. Results:(1) Follow-up of patients and analysis of influencing factors for prognosis. All 406 patients were followed up postoperatively for 28(range, 6-36)months, with 1- and 3-year overall survival rate of 86.5% and 40.9%, respectively. The 285 patients in the training set were followed up postoperatively for 30(range, 6-36)months, with 1- and 3-year overall survival rate of 85.1% and 35.5%, respectively. The 121 patients in the validation set were followed up postoperatively for 25(range, 6-36)months, with 1- and 3-year overall survival rate of 87.0% and 43.2%, respectively. There was no significant difference in postoperative overall survival rate between the training set and the validation set ( χ2=3.20, P>0.05). Results of multivariate analysis showed that left thoracic surgical approach, preopera-tive neutrophil count, vascular invasion, perineural invasion, pathological T2-4 stage, pathological N2-3 stage, and postoperative pneumonia were independent risk factors affecting postoperative survival of 285 patients in the training set ( hazard ratio=1.466, 1.037, 1.482, 1.549, 5.268, 7.727, 22.202, 2.539, 2.686, 1.425, 95% confidence interval as 1.026-2.096, 1.003-1.073, 1.008-2.179, 1.105-2.170, 1.201-23.099, 1.833-32.576, 4.734-104.128, 1.577-4.087, 1.631-4.422, 1.018-1.994, P<0.05). (2) Construction and validation of machine learning prediction models. Independent risk factors affecting postoperative survival were included to construct RSF, GBM, LASSO-Cox, CoxBoost, survivalsvm, XGBoost, SuperPC, and plsRcox machine learning prediction models. Results of Delong test showed that there were significant differences in the AUC of RSF and GBM from the other six models ( P<0.05). Results of time-dependent ROC curve showed that all 8 machine learning predic-tion models had good discriminative ability in the training cohort, among which the RSF machine learning prediction model had the best predictive performance. Results of calibration curve showed that the RSF machine learning prediction model fitted well for predicting postoperative 1-, 2-, and 3-year overall survival in the training cohort, with high consistency with actual results. Results of decision curve analysis showed that within a threshold range of 0-0.80, the RSF machine learning prediction model provided a better overall net benefit. Further analysis showed that in the validation set, the AUC of RSF machine learning prediction model for postoperative 1-, 2-, and 3-year survival prediction were 0.786 (95% confidence interval as 0.609-0.962), 0.774 (95% confidence interval as 0.676-0.873), and 0.750 (95% confidence interval as 0.652-0.848), respectively. Results of calibration curve showed that the RSF machine learning prediction model fitted well for predicting postopera-tive 1-, 2-, and 3-year overall survival in the validation set, with high consistency with actual results. In the training set, the optimal cutoff value of the RSF machine learning prediction model risk score was 11.7. Patients with risk score ≥11.7 were classified as the high-risk group, and those with risk score <11.7 as the low-risk group. The median survival times of the two groups were 18.0 months and >36.0 months, respectively, showing a significant difference between them ( χ2=73.30, P<0.05). In the validation set, the optimal cutoff value of the RSF machine learning prediction model risk score was 11.7. Patients with risk score ≥11.7 were classified as the high-risk group, and those with risk score<11.7 as the low-risk group. The median survival times of the two groups were 17.0 months and>36.0 months for the high-risk and low-risk groups, respectively, showing a significant difference between them ( χ2=35.20, P<0.05). Conclusions:Left thoracic surgical approach, preoperative neutrophil count, vascular invasion, perineural invasion, pathological T2-4 stage, pathological N2-3 stage, and postoperative pneumonia are independent risk factors affecting survival of esophageal cancer patients after radical resection. The RSF machine learning prediction model constructed based on these factors can effectively distinguish the survival prognosis of high-risk and low-risk patients.

Prostate cancer is the second most common cancer in men, accounting for 14.1% of new cancer cases in 2020. The aggressiveness of prostate cancer is highly variable, depending on its grade and stage at the time of diagnosis. Despite recent advances in prostate cancer treatment, some patients still experience recurrence or even progression after undergoing radical treatment. Accurate initial staging and monitoring for recurrence determine patient management, which in turn affect patient prognosis and survival. Classical imaging has limitations in the diagnosis and treatment of prostate cancer, but the use of novel molecular probes has improved the detection rate, specificity, and accuracy of prostate cancer detection. Molecular probe-based imaging modalities allow the visualization and quantitative measurement of biological processes at the molecular and cellular levels in living systems. An increased understanding of tumor biology of prostate cancer and the discovery of new tumor biomarkers have allowed the exploration of additional molecular probe targets. The development of novel ligands and advances in nano-based delivery technologies have accelerated the research and development of molecular probes. Here, we summarize the use of molecular probes in positron emission tomography (PET), single-photon emission computed tomography (SPECT), magnetic resonance imaging (MRI), optical imaging, and ultrasound imaging, and provide a brief overview of important target molecules in prostate cancer.
Humans ; Male ; Prostatic Neoplasms/diagnosis* ; Molecular Probes ; Molecular Imaging/methods* ; Magnetic Resonance Imaging ; Positron-Emission Tomography ; Tomography, Emission-Computed, Single-Photon ; Ultrasonography ; Optical Imaging ; Biomarkers, Tumor ; Precision Medicine/methods*

ObjectiveTo analyze the distribution, drug resistance characteristics, and changing trends of Acinetobacter baumannii (AB) isolated from environmental surfaces and healthcare workers’ hands in a grade Ⅱ level A general hospital in Xuhui District of Shanghai from 2018 to 2023, and to provide reference for infection control in the hospital. MethodsEnvironmental samples were collected quarterly from critical surfaces and healthcare workers’ hands in the intensive care unit (ICU), geriatrics, and respiratory departments from 2018 to 2023. Clinical isolates were obtained from all patients with AB infections in ICU, geriatrics, respiratory department, rehabilitation department, infectious diseases department, emergency department, cardiology department, and orthopedics of the hospital from 2018 to 2023. Retrospective analyses were performed on AB detection rates, strain origins, resistance rates to commonly used antimicrobial agents, and resistance gene features, comparing the antimicrobial resistance between clinically isolated strains and environmentally isolated strains. ResultsFrom 2018 to 2023, a total of 1 416 samples were collected from the hospital and a total of 272 strains of AB were detected, with a positive detection rate of 19.21%. The detection rate gradually decreased year-on-year (χ²_trend=45.290, P<0.001). The majority of samples originated from patient-contacted items (34.56%, 94/272), followed by shared items (26.84%, 73/272) and healthcare worker-contacted items (15.07%, 41/272). From 2018 to 2023, the resistance rate of AB on environmental surfaces and healthcare workers’ hands to commonly tested antibiotics in the hospital ranged from 10% to 40%. The resistance rates to cefotaxime (42.52%) and piperacillin (38.58%) were relative high, while the resistance to polymyxin E (1.57%), polymyxin B (2.36%), and doxycycline (3.94%) maintained low. The annual fluctuations in resistance to cefotaxime, piperacillin, ceftriaxone, tobramycin, doxycycline, minocycline and cotrimoxazole were statistically significant (all P<0.05). There were statistically significant differences in the resistance of clinical and environmental isolates to ampicillin/sulbactam, cefepime, ceftazidime, subamphetamine, meropenem, piperacillin, aztreonam, gentamicin, tobramycin, minocycline, ciprofloxacin, levofloxacin, and cotrimoxazole in the hospital from 2018 to 2023 (all P<0.05). The resistance rate of clinical isolates was generally high, especially to β-lactam and quinolone drugs, which were mostly above 80% [such as cefepime (93.86%), cefotaxime (97.37%), imipenem (98.25%), and ciprofloxacin (99.12%)]. The resistance rate of environmental isolated strains to similar antibiotics was relatively lower, mostly concentrated at 10%‒30%. The whole-genome sequencing of 34 carbapenem-resistant Acinetobacter baumannii (CRAB) strains isolated from the hospital environment in 2023 revealed that the main resistance mechanism was overexpression of efflux pumps (51.97%), followed by changes in target sites (32.46%). Among the 34 CRAB strains, carbapenem resistance genes OXA-23 and OXA-51 were detected in 6 strains (17.65%), while genes such as KPC, IMP, VIM, and SIM were not detected. ConclusionFrom 2018 to 2023, AB in the hospital environment exhibited high resistance rates to certain antimicrobial agents and carried multiple resistance genes, indicating a potential transmission risk. It is necessary to further strengthen bacterial resistance monitoring and hospital infection control, and use antibiotics reasonably.

Alzheimer's disease (AD) is the most prevalent neurodegenerative disorder worldwide, causing dementia and affecting millions of individuals. One prominent characteristic in the brains of AD patients is glucose hypometabolism. In the context of galactose metabolism, intracellular glucose levels are heightened. Galactose mutarotase (GALM) plays a crucial role in maintaining normal galactose metabolism by catalyzing the conversion of β-D-galactose into α-D-galactose (α-D-G). The latter is then converted into glucose-6-phosphate, improving glucose metabolism levels. However, the involvement of GALM in AD progression is still unclear. In the present study, we found that the expression of GALM was significantly increased in AD patients and model mice. Genetic knockdown of GALM using adeno-associated virus did not change the expression of amyloid precursor protein (APP) and APP-cleaving enzymes including a disintegrin and metalloprotease 10 (ADAM10), β-site APP-cleaving enzyme 1 (BACE1), and presenilin-1 (PS1). Interestingly, genetic overexpression of GALM reduced APP and Aβ deposition by increasing the maturation of ADAM10, although it did not alter the expression of BACE1 and PS1. Further electrophysiological and behavioral experiments showed that GALM overexpression significantly ameliorated the deficits in hippocampal CA1 long-term potentiation (LTP) and spatial learning and memory in AD model mice. Importantly, direct α-D-G (20 mg/kg, i.p.) also inhibited Aβ deposition by increasing the maturation of ADAM10, thereby improving hippocampal CA1 LTP and spatial learning and memory in AD model mice. Taken together, our results indicate that GALM shifts APP processing towards α-cleavage, preventing Aβ generation by increasing the level of mature ADAM10. These findings indicate that GALM may be a potential therapeutic target for AD, and α-D-G has the potential to be used as a dietary supplement for the prevention and treatment of AD.
Animals ; ADAM10 Protein/metabolism* ; Alzheimer Disease/pathology* ; Amyloid Precursor Protein Secretases/metabolism* ; Disease Models, Animal ; Humans ; Mice ; Amyloid beta-Peptides/metabolism* ; Male ; Mice, Transgenic ; Membrane Proteins/metabolism* ; Cognitive Dysfunction/pathology* ; Mice, Inbred C57BL ; Amyloid beta-Protein Precursor/metabolism* ; Female ; Hippocampus/metabolism* ; Long-Term Potentiation/physiology*

Objective To investigate the effects of different anesthesia depths on stress states and inflammatory mediators in patients undergoing video-assisted thoracoscopic lobectomy.Methods A total of 89 lung cancer patients who underwent video-assisted thoracoscopic lobectomy were selected as study subjects.Based on intraoperative bispectral index(BIS)range,the patients were divided into deep anesthesia group(BIS of 40 to＜50,n=45)and shallow anesthesia group(BIS of 50 to＜60,n=44).Vital signs(mean arterial pressure,heart rate and blood oxygen saturation),anesthesia re-covery time,extubation time,dosage of vasoactive drugs,postoperative pain intensity[Visual Ana-logue Scale(VAS)],postoperative analgesic dosage,perioperative stress state[prostaglandin E2(PGE2),nerve growth factor(NGF)and substance P(SP)],levels of inflammatory mediators[neuron-specific enolase(NSE),tumor necrosis factor-α(TNF-α)and S100β protein]at different time points(before anesthesia induction,immediately after intubation,before lesion resection and at the end of surgery)and the incidence of anesthesia-related adverse reactions were compared between the two groups.Results Before lesion resection and at the end of surgery,the mean arterial pressure and heart rate in the deep anesthesia group were significantly lower than those in the shallow anesthe-sia group(P＜0.05).The anesthesia recovery time and extubation time in the deep anesthesia group were significantly longer than those in the shallow anesthesia group(P＜0.05).At the end of surgery and on postoperative day one,the levels of PGE2,NGF and SP in the deep anesthesia group were significantly lower than those in the shallow anesthesia group,while the levels f NSE,TNF-α and S100β protein were significantly higher than those in the shallow anesthesia group(P＜0.05).There were no significant differences in the dosage of vasoactive drugs,VAS scores,sufentanil dos-age and the incidence of anesthesia-related adverse reactions between thetwo groups(P＞0.05).Conclusion During one-lung ventilation in patients undergoing video-assisted thoracoscopic surgery lobectomy,deep anesthesia can effectively control surgical stress and maintain stability of intraopera-tive hemodynamics,but it is associated with delayed postoperative awakening and more pronounced inflammatory response.Shallow anesthesia results in faster postoperative awakening and lower levels of inflammatory mediators,but it is associated with more significant intraoperative stress response and unstable hemodynamics.

Objective:To investigate the prognostic factors for esophageal cancer after radical resection and the application value of machine learning prediction model.Methods:The retrospective cohort study was conducted. The clinicopatholigical data of 406 esophageal cancer patients who were admitted to Qilu Hospital of Shandong University from January 2018 to March 2022 were collected. There were 357 males and 49 females, aged (64±8)years. All patients underwent radical resection of esophageal cancer. The 406 patients were randomly divided into a training set of 285 cases and a validation set of 121 cases at a 7∶3 ratio based on a random number table. The training set was used to construct prediction model, and the validation set was used to validate prediction model. Patients were divided into high-risk group and low-risk group based on risk scores. Observation indicators: (1) follow-up of patients and analysis of influencing factors for prognosis; (2) construction and validation of machine learning prediction models. Comparison of measurement data with normal distribution between groups was conducted using the independent sample 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. Comparison of ordinal data between groups was conducted using the rank sum test. The Kaplan-Meier method was used to calculate survival rate and plot survival curve, and the Log-rank test was used for survival analysis. The Cox proportional hazard regression model was used for univariate and multivariate analyses. Independent influencing factors were included, and data processing, machine learning model construction, and visualization were performed using R packages including random survival forest (RSF), gradient boosting machine (GBM), least absolute shrinkage and selection operator Cox regression (LASSO-Cox), Cox proportional hazards model boosting (CoxBoost), survival support vector machine (survivalsvm), extreme gradient boosting (XGBoost), supervised principal component analysis (SuperPC), and Cox partial least squares regression (plsRcox). Receiver operating characteristic (ROC) curves were drawn, and sensitivity, specificity, and area under the curve (AUC) were calculated. The Delong test was used to assess the differences in AUC among different models in the training set, and the time-dependent ROC was used to compare the predictive performance of different models. Calibration curves were used to evaluate model accuracy, and decision curve analysis (DCA) was used to evaluate overall net benefit. Results:(1) Follow-up of patients and analysis of influencing factors for prognosis. All 406 patients were followed up postoperatively for 28(range, 6-36)months, with 1- and 3-year overall survival rate of 86.5% and 40.9%, respectively. The 285 patients in the training set were followed up postoperatively for 30(range, 6-36)months, with 1- and 3-year overall survival rate of 85.1% and 35.5%, respectively. The 121 patients in the validation set were followed up postoperatively for 25(range, 6-36)months, with 1- and 3-year overall survival rate of 87.0% and 43.2%, respectively. There was no significant difference in postoperative overall survival rate between the training set and the validation set ( χ2=3.20, P>0.05). Results of multivariate analysis showed that left thoracic surgical approach, preopera-tive neutrophil count, vascular invasion, perineural invasion, pathological T2-4 stage, pathological N2-3 stage, and postoperative pneumonia were independent risk factors affecting postoperative survival of 285 patients in the training set ( hazard ratio=1.466, 1.037, 1.482, 1.549, 5.268, 7.727, 22.202, 2.539, 2.686, 1.425, 95% confidence interval as 1.026-2.096, 1.003-1.073, 1.008-2.179, 1.105-2.170, 1.201-23.099, 1.833-32.576, 4.734-104.128, 1.577-4.087, 1.631-4.422, 1.018-1.994, P<0.05). (2) Construction and validation of machine learning prediction models. Independent risk factors affecting postoperative survival were included to construct RSF, GBM, LASSO-Cox, CoxBoost, survivalsvm, XGBoost, SuperPC, and plsRcox machine learning prediction models. Results of Delong test showed that there were significant differences in the AUC of RSF and GBM from the other six models ( P<0.05). Results of time-dependent ROC curve showed that all 8 machine learning predic-tion models had good discriminative ability in the training cohort, among which the RSF machine learning prediction model had the best predictive performance. Results of calibration curve showed that the RSF machine learning prediction model fitted well for predicting postoperative 1-, 2-, and 3-year overall survival in the training cohort, with high consistency with actual results. Results of decision curve analysis showed that within a threshold range of 0-0.80, the RSF machine learning prediction model provided a better overall net benefit. Further analysis showed that in the validation set, the AUC of RSF machine learning prediction model for postoperative 1-, 2-, and 3-year survival prediction were 0.786 (95% confidence interval as 0.609-0.962), 0.774 (95% confidence interval as 0.676-0.873), and 0.750 (95% confidence interval as 0.652-0.848), respectively. Results of calibration curve showed that the RSF machine learning prediction model fitted well for predicting postopera-tive 1-, 2-, and 3-year overall survival in the validation set, with high consistency with actual results. In the training set, the optimal cutoff value of the RSF machine learning prediction model risk score was 11.7. Patients with risk score ≥11.7 were classified as the high-risk group, and those with risk score <11.7 as the low-risk group. The median survival times of the two groups were 18.0 months and >36.0 months, respectively, showing a significant difference between them ( χ2=73.30, P<0.05). In the validation set, the optimal cutoff value of the RSF machine learning prediction model risk score was 11.7. Patients with risk score ≥11.7 were classified as the high-risk group, and those with risk score<11.7 as the low-risk group. The median survival times of the two groups were 17.0 months and>36.0 months for the high-risk and low-risk groups, respectively, showing a significant difference between them ( χ2=35.20, P<0.05). Conclusions:Left thoracic surgical approach, preoperative neutrophil count, vascular invasion, perineural invasion, pathological T2-4 stage, pathological N2-3 stage, and postoperative pneumonia are independent risk factors affecting survival of esophageal cancer patients after radical resection. The RSF machine learning prediction model constructed based on these factors can effectively distinguish the survival prognosis of high-risk and low-risk patients.