3D modeling to predict vascular involvement in resectable pancreatic adenocarcinoma

Abstract

Background: Current management of patients with borderline resectable pancreatic adenocarcinoma (BR-PDAC) depends on the degree of involvement of the major arterial and venous structures. The aim of this study was to evaluate 3D segmentation and printing to predict tumor size and vascular involvement of BR-PDAC to improve pre-operative planning of vascular resection and better select patients for neoadjuvant therapy.

Methods: We retrospectively evaluated 16 patients with BR-PDAC near vascular structures who underwent pancreatoduodenectomy (PD) with or without vascular resection between 2015 and 2021. The pre-operative computed tomography (CT) images were processed by segmentation with 3D reconstruction and printed as 3D models. Two radiologists specialized in pancreatic imaging and two pancreatic surgeons blindly and independently analyzed the pre-operative CT scans and 3D models using a defined checklist. Their evaluations were compared to the pre-operative 2D-CT reports utilized for patient management. A positive delta was defined by the 3D analysis resulting in greater accuracy in predicting vascular involvement as proven intraoperatively or histopathologically.

Results: Fourteen PD, one total pancreatectomy, and one exploratory laparotomy were performed. Ten patients had a positive delta concerning vascular involvement of the superior mesenteric or portal vein. Tumor extension was also more accurately evaluated by 3D modeling than by 2D-CT (p < 0.05).

Conclusions: Our pilot study demonstrates that 3D segmentation can provide additional information for choosing the best treatment strategy and surgical plain in patients with BR-PDAC. Especially for upcoming mini-invasive techniques like laparoscopic and robotic resections, better pre-operative planning is essential to allow safety and prevent vascular injury.

Keywords: 3D printing; 3D rendering; Borderline resectable pancreatic cancer; R1 resection; Vascular involvement of pancreatic cancer.

Fig. 1

Images from an 80-year-old patient with isolated painless jaundice. A. CT showing dilation of the pancreatic duct at 12 mm with stenosis in the pancreatic head, which was poorly defined with no vascular invasion. Intra-operative findings: portal vein (PV) involvement requiring resection and anastomosis. Pathology showed that ductal type pancreatic adenocarcinoma developed as a PanIN 3 lesion (pancreatic intraepithelial neoplasia) of the pancreatic ducts without infiltration of the PV: pT2 (3.5 cm) N1 (2/30) G2 L1 V0 Pn1 R0. B, C. 3D reconstructions. D. 3D printing showing no invasion of the hepatic artery but superior mesenteric vein (SMV) involvement. The SMV involvement was detected retrospectively. For the surgeon, the initial interpretation was ‘no vascular invasion’ and intra-operatively found involvement of the PV, which was not confirmed by the pathology report.

Fig. 2

Radiology check-list for the pancreatic adenocarcinoma reporting template. This "Pancreatic Adenocarcinoma Reporting Template" is a structured checklist for documenting findings related to pancreatic cancer. It covers: 1. Tumor Details: Records the size and specific location within the pancreas. 2. Duct and Biliary Involvement: Notes if there's blockage or dilation in the pancreatic duct or bile duct. 3. Major Artery and Vein Involvement: Assesses contact, degree of involvement, and any narrowing or irregularity for key blood vessels: Superior Mesenteric Artery (SMA); Celiac Artery (CA); Common Hepatic Artery (CHA); Main Portal Vein (MPV); Superior Mesenteric Vein (SMV); 4. Arterial Variants: Identifies any arterial variations based on a classification system.Each section checks for tumor contact, involvement extent, vessel narrowing, and other abnormalities, helping clinicians evaluate the spread and surgical options for pancreatic cancer.

Fig. 3

Wilcoxon Mann-Whitney comparison of the mean diameters shows that the tumor size evaluated by 3D was closer to the pathology measurement than the pre-operative 2D imaging. This figure presents four box plots comparing tumor size measurements using different methods in patients with pancreatic cancer:Panel A: Tumor Size Distribution by Method: this plot compares tumor size distribution across three methods: m_preop_ct (blue): tumor size measured with pre-operative 2D-CT; m_3d (yellow): tumor size measured with 3D-CT; m_patho (orange): tumor size determined by pathology (considered the most accurate post-surgical measurement). The plot shows that 3D-CT measurements are closer to pathology measurements than 2D-CT measurements. Panel B: Comparison of Tumor Size on 2D-CT and 3D-CT. This plot connects tumor sizes for each patient measured on 2D-CT and 3D-CT. The mean difference is 7.3 mm with a p-value of 0.004, suggesting a significant difference between 2D-CT and 3D-CT measurements, with 3D-CT measurements being closer to the actual size. Panel C: Comparison of Tumor Size on 3D-CT and Pathology. This plot compares 3D-CT and pathology measurements for each patient. The mean difference is 7.3 mm with a p-value of 0.005, indicating that 3D-CT provides a more accurate estimate of tumor size compared to pathology than 2D-CT. Panel D: Comparison of Tumor Size on 2D-CT and Pathology. This plot compares tumor sizes measured on 2D-CT with those measured by pathology. The mean difference is 13.5 mm with a p-value <0.001, showing a significant discrepancy between 2D-CT measurements and actual tumor size.Conclusion: 3D-CT provides a more accurate approximation of tumor size as confirmed by pathology compared to traditional 2D-CT, with statistically significant differences in all comparisons.