Synopsis

Keywords: Pancreas, Surgery, Magnetic resonance elastographies, Pancreatectomy, Pancreatic fistula

This prospective study enrolled the patients who underwent both preoperative tomoelastography and pancreaticoenteric anastomosis. Eighty-two patients were included (median age: 59.5 years, 40 men, 18 patients with POPF). Main pancreatic duct diameter (MPDD) (P=0.012), c (P<0.001) and φ (P=0.001) were relevant factors for POPF. The area under the curve (AUC) of c, φ and MPDD for predicting POPF was 0.880, 0.816 and 0.747. Fibrosis was the relevant factor of POPF (P=0.001). There was positive correlation between fibrosis and stiffness (r=0.681, P<0.001). Tomoelastography is a novel and robust multi-frequency MRE technique that can facilitate the prediction of POPF.

Manuscript

Introduction
Postoperative pancreatic fistula (POPF) is the most common and serious complication of pancreatectomy (1). Preoperative risk stratification of POPF in patients undergoing pancreatectomy facilitates perioperative precision medicine (2, 3). The fistula risk score, obtained perioperatively, from a more practical point of view, tend to be fraught with potential dangers.
Previous studies had been conducted to evaluate the relationship between parameters derived from imaging and pancreatic texture for predicting POPF, such as the signal intensity ratio of T1-weighted MRI (4), magnetization transfer imaging (5) and dual-energy CT (6). Whereas, those quantitative tools cannot directly quantify pancreatic texture but also influenced by other factors. Pancreatic US elastography could be used to assess pancreatic stiffness but suffered from limited reproducibility (6).
Tomoelastography, a novel multi-frequency MRE technique with noise-robust data postprocessing, can generate quantitative maps for biomechanical properties with high-resolution anatomical details (7, 8). To the best of our knowledge, the possibility of using tomoelastography to predict the occurrence of POPF has not been explored.
Therefore, the present study was to prospectively investigate the utility of tomoelastography for preoperative risk prediction for occurrence of POPF in patients undergoing pancreaticoenteric anastomosis.
Materials and Methods
This prospective single-center study was approved by the institutional review board of our hospital and the written informed consent were received from patients. The flow chart was shown in Figure 1.
The grading criteria of pancreatic fibrosis at the resection margin were performed as previously described as 4 grades (9). The exocrine gland atrophy was classified according to the percentage of viable exocrine gland as previously described as 3 grades (10). The grading criteria of lipomatosis were adapted as previously described as 4 grades (10, 11).
The entire pancreatic tomoelastography was scanned in 7 minutes and 22 seconds covered by 35 contiguous axial sections. The multifrequency wave field data was processed at https://bioqic-apps.com.
The region of interests (ROIs) (larger than 100 mm3) were placed at the resection margin of pancreas, avoiding visible pancreatic duct dilation, boundary and artifacts, to measure fat fraction (on multi-echo Dixon image), width (on axial fs-T2WI), thickness (on coronal T2WI), stump area (width multiple thickness), stiffness (on c map) and fluidity (on φ map). The main pancreatic duct diameter (MPDD) was also measured (Figure 2).
Results
A total of 82 patients were enrolled in this study (median age, 59.5 years, range from 20 to 83 years) and 40 patients were men. There were 18 patients with POPF (ISGPS B and C) among all. The Bland-Altman analysis showed agreements in different readers, different measurements with the same reader (Figure 3). The ICC showed near perfect agreements in c (0.987, P<0.001) and φ (0.993, P<0.001).
There was significant difference between patients with and without POPF in fibrosis (P=0.002), acinar atrophy (P=0.024), MPDD (P=0.001), c (P<0.001) and φ (P<0.001). According to the univariate analysis, fibrosis (P=0.001; OR, 0.29), acinar atrophy (P=0.011; OR, 0.19), MPDD (P=0.012; OR, 0.59), c (P<0.001; OR<0.001) and φ (P=0.001; OR<0.001) were relevant factors of POPF. The only independent relevant factor for preoperatively predicting POPF was c (P<0.001) at the multivariate regression.
The use of c enabled prediction of POPF with an AUC of 0.880 (cutoff, 1.413 m/sec), which was comparable to the predictive performance of φ (AUC, 0.816; cutoff, 0.796 radian) (NRI: 0.029, P=0.843; IDI: 0.066, P=0.342) and higher than the predictive performance of MPDD (AUC, 0.747; cutoff, 2.67 mm) (NRI: 0.125, P=0.344; IDI: 0.132, P=0.021), as the Figure 4 shown.
There were 6 patients in F0 fibrosis (3 were with POPF), 26 in F1 (11 were with POPF), 30 in F2 (3 were with POPF), 20 in F3 (1 were with POPF), as the Figure 5 shown. The mean c was 1.321 m/sec in F0 fibrosis, while 1.396, 1.538, 1.612 in F1-3 fibrosis respectively (P<0.001). There was positive correlation between c and fibrosis grades (r=0.681, P<0.001).
Discussion
From the perspective of precision medicine, there is still a clinical need for straightforward and non-invasive modality to quantify pancreatic texture thus preoperatively identify patients at high risk of POPF. In the present study, both mechanical (pancreatic stiffness and fluidity quantified with tomoelastography) and morphologic (MPDD) features were relevant factors for POPF. The predictive performance of pancreatic stiffness was comparable to the fluidity and better than MPDD. Pancreatic fibrosis was the relevant factors for POPF and there was positive correlation between fibrosis and pancreatic stiffness. To the best of our knowledge, this is the first attempt to explore the potential of tomoelastography for preoperative risk prediction for POPF in patients undergoing pancreaticoenteric anastomosis.
In conclusion, tomoelastography is a novel and robust multi-frequency MRE technique that can facilitate the assessment of pancreatic texture. The elevated pancreatic stiffness and fluidity at the resection margin, as quantified by tomoelastography, are mechanical signatures of POPF. Preoperative identification of patients undergoing pancreaticoenteric anastomosis with high-risk POPF using tomoelastography will help improve perioperative precision treatment.