A Machine-learning Model For The Difficulty Of Retroperitoneal Laparoscopic Adrenalectomy Of Adrenal Benign Tumor Based On Radiomics

Objective:Based on the clinical and radiomics characteristics of patients with adrenal benign tumors,a machine-learning prediction model for the difficulty of retroperitoneal laparoscopic adrenalectomy was constructed and validated.Methods:A retrospective collection of patients with adrenal benign tumors who visited the Third Hospital of Shanxi Medical University from August 2014 to December 2020 was conducted.The patients were randomly divided into a training set and an internal validation set at a ratio of 7:3.The training set was used for model construction,and the internal validation set was used for model validation.The 3D Slicer Image Computing Platform and Python were used to draw three regions of interest(adrenal tumors,peri adrenal-tumor fat,and retrorenal fat)and extract radiomics features.Key features were screened through LASSO regression and radiomics-score was calculated.Combining radiomics-score with clinical characteristics,the optimal model is selected through a comprehensive evaluation of several machine learning models.The effect of the model was verified by ROC curves and AUC.Validate the model using the Bootstrap resampling method and draw a calibration curve to evaluate the consistency of the model.The net benefit of patients was analyzed through DCA.The difference is considered statistically significant when the P value is less than 0.05.Results:After comparing seven machine learning models,the LASSO model was finally selected.The independent influencing factors for screening were: radiomics-score,history of diabetes,history of the spinal disease(scoliosis),and previous surgery history.The AUC of the training and internal validation sets were 0.894 and 0.929,respectively.Clinical DCA shows that the model can improve patient benefits by approximately 20%.In addition,108 patients were collected for prospective validation,with an AUC of 0.810.Conclusion:This study constructs a machine learning model that combines clinical and radiomics characteristics to predict the difficulty of retroperitoneal laparoscopic adrenalectomy of adrenal benign tumors.The model has good predictive performance,which helps the surgeon evaluate the difficulty of surgery,thereby personalized and improving preoperative preparation,reducing surgical risks,and improving patient benefits.