Three-dimensional simulation of the pancreatic parenchyma, pancreatic duct and vascular arrangement in pancreatic surgery using a deep learning algorithm.

Abstract

Three-dimensional surgical simulation, already in use for hepatic surgery, can be used in pancreatic surgery. However, some problems still need to be overcome to achieve more precise pancreatic surgical simulation. The present study evaluates the performance of SYNAPSE VINCENT® (version 6.6, Fujifilm Medical Co., Ltd., Tokyo, Japan) in the semiautomated surgical simulation of the pancreatic parenchyma, pancreatic ducts, and peripancreatic vessels using an artificial intelligence (AI) engine designed with deep learning algorithms. One-hundred pancreatic cancer patients and a control group of 100 nonpancreatic cancer patients were enrolled. The evaluation methods for visualizing the extraction were compared using the Dice coefficient (DC). In the pancreatic cancer patients, tumor size, position, and stagewise correlations with the pancreatic parenchymal DC were analyzed. The relationship between the pancreatic duct diameter and the DC, and between the manually and AI-measured diameters of the pancreatic duct were analyzed. In the pancreatic cancer/control groups, the pancreatic parenchymal DC and pancreatic duct extraction were 0.83/0.86 and 0.84/0.77. The DC of the arteries (portal veins/veins) and associated sensitivity and specificity were 0.89/0.88 (0.89/0.88), 0.85/0.83 (0.85/0.82), and 0.82/0.81 (0.84/0.81), respectively. No correlations were observed between pancreatic parenchymal DC and tumor size, position, or stage. No correlation was observed between the pancreatic duct diameter and the DC. A positive correlation (r = 0.61, p<0.001) was observed between the manually and AI-measured diameters of the pancreatic duct. Extraction of the pancreatic parenchyma, pancreatic duct, and surrounding vessels with the SYNAPSE VINCENT® AI engine assumed to be useful as surgical simulation.