Browsing by Subject "pancreatoduodenectomy"

Background: A high proportion of acinar cells in the pancreatic resection line (ranging from 40% to more than 60%) has been found to be a predictor for complications after pancreatic surgery. In this study, we aimed to analyze the proportions of acinar, fibrous, and fat cells in pancreatic resection line samples and to find out whether these proportions have predictive value for the development of postoperative pancreatic fistula. Methods: Data from 668 consecutive patients who underwent pancreatoduodenectomy in 2000–2017 in Helsinki University Hospital were collected retrospectively from a patient database. The histological analysis of pancreatic resection line samples was conducted by an artificial intelligence software (Aiforia) to obtain the most objective analysis of the pancreatic texture. Results: 476 patients were included in the study, of which 22 patients (4.7 %) developed a clinically relevant postoperative pancreatic fistula (grade B fistula n = 13, grade C fistula n = 9). Patients who developed a clinically relevant postoperative pancreatic fistula had a significantly higher acinar content in the pancreatic resection line compared with patients who did not develop a fistula (52.1% vs. 11.2%, p < 0.001). The amount of fibrosis in the resection line was significantly lower in patients who developed a clinically relevant postoperative pancreatic fistula compared to patients who did not develop a fistula (9.1% vs. 39.5%, p < 0.001). There was no statistically significant difference in the fat cell proportions in the resection line between the two groups (p = 0.9). Conclusions: A higher proportion of acinar cells and a lower proportion of fibrous cells in the pancreatic resection line is associated with the development of a clinically relevant pancreatic fistula. The major limitation about using artificial intelligence is that if the sample is not of good quality (if it is too thick or too darkly stained), artificial intelligence is unable to analyze the sample, even if it can be analyzed by human eyes. These limitations must be considered before using AI in larger research settings.