[Factors of delayed gastric emptying after pancreaticoduodenectomy].

OBJECTIVE

To study the factors of delayed gastric emptying after pancreaticoduodenectomy using artificial intelligence.

MATERIAL AND METHODS. THE: Study included 142 patients after pancreaticoduodenectomy for malignant pancreaticoduodenal tumors. The probability of delayed gastric emptying was determined using neural network analysis (SPSS Statistics v.26). Quality of the model was assessed considering sensitivity, specificity, diagnostic efficacy and AUROC.

RESULTS

Delayed gastric emptying after pancreaticoduodenectomy occurred in 33 (23.2%) patients. In multivariate analysis, delayed gastric emptying was associated with BMI > 30 kg/m2, tumor diameter ≤2 cm and postoperative pancreatitis. Model sensitivity was 93.9%, specificity 97.2%, diagnostic efficacy 96.5%, AUROC 0.924±0.011 (95% CI: 0.834-0.967).

CONCLUSION. A: Rtificial intelligence makes it possible to determine the factors of delayed gastric emptying after pancreaticoduodenectomy.