Preoperative Evaluation And Risk Prediction Model Construction Of Difficult Laparoscopic Cholecystectomy

Objective:To establish a preoperative evaluation model for DLC that can effectively predict the degree of surgical difficulty,minimize the incidence of operation-related complications,and improve overall surgical safety.Methods:We conducted a retrospective analysis of the clinical data from 521 patients who underwent LC at our hospital between January 2019 and December 2021.Based on the standard operation time,the patients were divided into two groups:difficult group(DLC group)and non-difficult group(NDLC group).The age,sex,medical history,and auxiliary examination data of both groups were recorded.Quantitative data were transformed into classified data according to certain standards based on clinical experience.Chi-square test was used to compare groups for qualitative data,while Chi-square test was used to compare groups for quantitative data.Intraoperative and postoperative data were used to verify the rationality of the grouping.Univariate analysis was performed to screen statistically significant indicators,which were then entered into multivariate logistic regression analysis.Model fit was evaluated using the Hosmer-Lemeshow test,and model effectiveness was evaluated using receiver operating characteristic curve analysis and calculation of the area under the curve(AUC).Based on the above regression equation,a nomogram model for preoperative risk prediction of difficult laparoscopic cholecystectomy was constructed.The reliability of the nomogram was assessed by drawing a calibration curve.The clinical practical value of the nomogram was verified by quantifying the net benefit of LC patients using this model under different threshold probabilities,and by performing decision curve analysis and clinical impact curve analysis.Results:Among the 521 patients,156 were in the DLC group and 365 were in the NDLC group.Significant differences were observed between the two groups in terms of intraoperative blood loss(χ2=22.221,P<0.001),indwelling drainage tube(χ2=61.563,P<0.001),postoperative pain(χ2=9.848,P=0.002),postoperative bile leakage(χ2=4.100,P=0.043),and postoperative hospital stay(χ2=9.337,P=0.002).These results indicate that there were significant differences between the difficult and non-difficult groups,suggesting the grouping was reasonably sound.Gender(χ2=11.161,P=0.011),history of acute cholecystitis(χ2=11.161,P<0.001),neutrophil count(χ 2=17.346,P<0.001),gallbladder wall thickness)4mm(χ2=34.024,P<0.001),gallstone diameter)2.5cm(χ2=18.693,P<0.001),and gallbladder neck stone incarceration(χ2=27.103,P<0.001)were identified as independent risk factors for DLC(all P<0.05 and OR>1).Based on this model,a DLC risk prediction line chart was established using R language software.The receiver operating characteristic curve showed an area under the curve of 0.776,indicating good consistency of the model based on the calibration curve.Additionally,the clinical decision-making curve and clinical impact curve demonstrated the model’s clinical practicality.Conclusion:The preoperative evaluation model of deep learning-based classification(DLC)can effectively predict the difficulty level of surgical procedures,enabling clinicians to adopt corresponding clinical strategies.This can significantly reduce the incidence of postoperative complications and mitigate the overall risk associated with the procedure.Furthermore,the implementation of this predictive model can facilitate the practice of precision medicine and enhance clinical efficiency.Importantly,this model is particularly beneficial for young doctors who are in the process of overcoming the steep learning curve associated with laparoscopic cholecystectomy(LC),as well as aiding departments in the early stages of implementing new technology projects by assisting with case screening.