Establishment And Application Of Diagnostic Model For Abdominal Enhanced CT To Detect Left Gastric Artery Variation

Objective:1.Abdominal enhanced CT was used on the variation and classification of the left gastric artery.2.The variations of left gastric artery was detected with the practical value of artificial intelligence technology.3.Analyze the clinical value of anatomical variation of perigastric artery during abdominal surgery.Methods: 1.The imaging data of 305 patients who underwent abdominal enhanced CT scan in the radiology department of the first affiliated hospital of the university of south China from January 2019 to December 2019 were selected retrospectively.The patient's enhanced CT vascular image is read to classify the aberrant perigastric artery.2.Mark the variant arteries and divide all data into five groups randomly,four training groups,and one test group.At the same time,a classification-detection model is constructed to perform deep learning on the data,the obtained Area Under the Curve(AUC)is the detection performance value of the model.Results: According to the Michels classification,there are 247 cases of type I,58 patients have vascular variation,and the overall variation rate was 19.0%.type II: 8 patients,2.6%;type III and type IV: none;type V: 5 patients,1.6%;type VI: 6 patients,2.0%.In addition,there are also found 1 case of absence of the common hepatic artery,2 cases of absence of the left gastric artery,2 cases of the superior mesenteric artery gave rise to the right gastroomental artery,and another 34 cases not included in Michels classification.According to Hiatt classification,Hiatt type I: 247 cases,about 81%;Hiatt type II: 24 cases,about 7.9%;Hiatt type III: 10 cases,about 3.2%;Hiatt type IV: 3 cases,about 1%;Hiatt type V: 2 cases,about 0.6%;Hiatt type VI: 2 cases,about 0.6%.Classification of vascular variantion related to the left gastric artery in this study: Type 1: 4 cases of left gastric artery from splenic artery,about 1.3%;Type 2: 3 cases of left gastric artery from abdominal aorta,about 1.0%;Type 3: 1 cases of left gastric artery from common hepatic artery,about 0.3%;Type 4: 12 cases of accessory left hepatic artery,about 3.9%;type 5: 15 cases of replaced left hepatic artery,about 4.9%;type 6: 2 cases of absence of the left gastric artery,about 0.7%.The average AUC,recall rates,precision rates and normal rates of the five groups of the classification network model were 0.82,69.4%,82.5% and 79.5%,respectively.The average AUC,recall rate,precision rate and normal rate of the five groups of the detection network model were 0.87,65.6%,87.7% and 77.8%,respectively.Conclusion: 1.Among the new method of observing the variation of blood vessels,the replaced(accessory)left hepatic artery is the most common.2.The model has good detection efficiency for the left gastric artery and the replaced(accessory)left hepatic artery.