| Part 1.The Trends of Laparoscopic Surgery:Real World Data from the single High-Volume Hospital in China,1999-2019Objective:The epidemiologic data describing the trend of laparoscopic surgery in real-world was lack to support its wide use and clinical benefit in China because of the singledomain studies and unreliable coding.This study aimed to report trends of laparoscopic surgery in a Chinese hospital from 1999 to 2019.Methods:The original data from electronic medical record(EMR)were processed by natural language processing(NLP)and extracted from the Nanfang Hospital’s Intelligence Platform,and then joinpoint regression was used to evaluate the trends with time and segmented time into some periods by the annual number of laparoscopic surgeries.Subsequently,the demographic,surgical data,complications,and healthcare utilization in different periods were described and compared.Results:The cases increased steadily across the whole study period and 3 joinpoints were detected in 2009,2012 and 2015 respectively.There were fewer complications(1.9%,210 of 11229)and delayed healing of incision(1.1%,118 of 11229)in 2013-2015(all P<0.05).The mortalities in 4 periods were similar(P>0.05).After adjustment,the least overall morbidities appeared in 2013-2015(odd ratio(OR),0.617;95%CI,0.504-0.756)and the biggest changes took place in the infection(OR,0.640;95%CI 0.514-0.795).In addition,with time going by,the length of stay decreased and the hospital charge increased.Conclusion:Our findings indicate an upward trend in laparoscopic surgery with more indications,safety,and efficacy between 1999 to 2019,and suggest that more realworld evidence about cost-effectiveness should be provided.Part 2.The Changes of Laparoscopy’s Diagnostic Efficacy in Staging of Gastric CancerObjective:It’s still hard to stage gastric cancer accurately while laparoscopy plays an important role.The purpose of this part is to describe the trend of laparoscopy’s diagnostic efficacy in staging of gastric cancer.Methods:The baseline characteristics,and clinical and pathologic TNM stage were collected from the patients with gastric cancer undergoing laparoscopic surgery in Nanfang Hospital.And then Cohen’s kappa index was calculated in groups.Results:There’s no significant difference in clinical T stage and M stage at different periods while difference in pathological T stage was observed.But the comparison of clinical and pathological N stage is converse.The Cohen’s kappa index of laparoscopic evaluation for T stage and N stage at different periods was 0.463～0.574 and 0.014～0.380,respectively.Conclusion:Our finding suggested laparoscopy can evaluate T stage in GC availably rather than N stage and M stage,and implied there’s still a need to build a novel tool for staging of gastric cancer.Part 3.Bi-regional and Bi-phasic Automated Machine Learning Radiomics for Defining Metastasis to Lesser Curvature and Supra-pancreatic Lymph Node Stations in Gastric CancerObjective:Given the complicated lymphatic network system of stomach,a numeric-based classification can’t reflect on the anatomic extent of gastric cancer(GC).And lymph node metastasis(LNM)is observed in lesser curvature and supra-pancreatic areas more frequently.We aimed to build a radiomic model to predict the status of lesser curvature and supra-pancreatic LN stations.Methods:A combined size of 939 gastric cancer patients were retrospectively collected from two centers.A hybrid radiomic model was built by the combination of bi-regional and bi-phasic model and validated internally and externally.Further,the potential generalization ability of the hybrid model was investigated in the prediction of metastasis status of supra-pancreatic LN stations(Nos.7,8,9,and 11).Results:The hybrid model yielded substantially higher performance with AUCs of 0.847(95%CI,0.770-0.924)and 0.833(95%CI,0.800-0.867)in the two test cohorts.Moreover,the hybrid model showed various performances in the prediction of LNM in Nos.7,8,9,11 with AUCs ranging from 0.678 to 0.761 in two test cohorts and showed a slightly worse performance in Nos.8.Conclusion:The features of primary tumor and nearby tissues in CT imaging are related to LN status.And our as-developed model showed great predictive performance and might be of great application in the individual treatment of GC. |
