Porto-Venous Radiomics in Pancreatic Cancer

Pancreatic cancers are often inoperable at diagnosis, with only 13-22% of patients being suitable for surgical resection.1 Surgery is the only curative therapeutic option for pancreatic cancer and higher rates of long-term patient survival following surgery are seen in patients where the tumour is completely removed (termed R0 resection). However a majority of those patients who undergo surgery will develop tumour recurrence at the primary surgical site or a distant location. The likelihood of cancer recurrence following surgery increases when there is involvement of major blood vessels that lie adjacent to the pancreas. This involvement is often microscopic in nature and only detected after surgery when the surgical specimen is examined microscopically (termed R1 resection). Pre-operative cross-sectional imaging (usually Computed Tomography (CT) is the standard tool used to assess the risk of R1 resection prior to pancreatic cancer. However this approach significantly overestimates the number of patients with a resectable tumour as a significant proportion ultimately have an R1 resection after surgery. If there is felt to be an increased risk of R1 resection prior to surgery the patient can be offered oncological treatment prior to surgery to improve the rates of R0 resection that in turn improve long-term patient survival. The emerging field of Radiomics offers the opportunity to stratify patients undergoing pancreatic cancer surgery into those at risk of R1 resection prior to surgery. Radiomics has emerged as a valuable adjunct to medical decision making over the past 10 years. The basic premise is to extract and analyse large amounts of information not usually visible to the human eye from medical images, and it is increasingly being used in the field of cancer diagnostics. The aim of this study is to employ 3D radiomic analyses to pre-operative CT scan images for cases of pancreatic head cancer that underwent surgical resection, to determine whether this information could be used to predict R0 and R1 resection margins pre-operatively.