Abstract PO-010: Detection of early tissue changes on historical CT scans in the regions of the pancreas gland that subsequently develop adenocarcinoma using quantitative textural analysis and fat fraction analysis

Abstract

Abstract Early detection of Adenocarcinoma of the Pancreas (ACP) is critical to improving outcomes. Since the development of ACP is thought to begin a couple decades prior to clinical presentation, the possibility exists that evolving changes in the pancreas gland (PG) may already be present on historical standard of care (h-SOC) CT scans obtained years earlier in patients who present for other indications. Advanced image analysis using quantitative texture analysis (QTA) techniques can detect subtle changes in tissue/tumors composition (including fat) and may be useful in tumor detection, diagnosis and response assessments. We hypothesized that changes in tissue texture are detectable on h-SOC in patients who subsequently develop ACP and, if identified, would contribute significantly towards the development of tools to aid in identifying tissue at risk for ACP An IRB-exempt, retrospective, single institution study of 27 matched h-SOC and ACP diagnostic CTs from a single institution was performed. Subjects who had ACP and h-SOC CTs between 3-15 years prior were included. Deidentified scans were transferred to the imaging core lab for QTA and fat fraction (FF) analysis. The pancreas gland (PG) was divided into 7 regions (uncinate, head, neck-genu, body [prox,mid,dis] and tail) for volumes-of interest (VOIs) placement on the h-SOC portal venous phase axial images. A single radiologist verified VOI placement, confirmed the lack of ACP on h-SOC images and was blinded to the location of subsequent ACP development. First order QTA histogram frequency curves derivatives (mean, SD, skewness, kurtosis, mean positive Pixel (MPP), at each cluster [SSF] setting from 0-6mm) were recorded along with FF. Inferential statistical analysis was performed to identify QTA/FF differences between PG regions that subsequently developed ACP from those that did not (p <0.05 significance). A total of 22/27 subjects (81%) had suitable portal venous phase CTs for QTA while evaluable FF data was available for all subjects. ACP developed in PG head in 45% of subjects. The average time from h-SOC to diagnostic ACP scan was 7.6 years (range, 3.9 years-13.9 years). QTA results showed a difference in tissue texture (QTAskewness > -0.480) in PG regions that subsequently developed ACP compared to regions that did not (T-statistics= -2.148; AUCROC= 0.625 p=.038). There was also a > 3-fold increase in PG tail fat in subjects that developed tail ACP (t-statistics= -3.048; p = 0.002; AUCROC= 0.819 p=.023). LOGR models showed that PG regions that subsequently developed ACP had differences in QTA mean, skewness, kurtosis and total FF on h-SOC scans compared to normal PG tissue (LL ratio-p=.004, pseudo R2=.104). Tissue texture and fat composition differences in regions of the PG may be present on h-SOC CT scans several years prior to clinical manifestation of ACP. If validated, tissue at risk could be identified well in advance of actual ACP development allowing for new opportunities for PAC interception. The investigators acknowledge the Marley Foundation for their generous support. Citation Format: Ronald L. Korn, Daniel D. Von Hoff, Andre Burkett, Dominic Zygadlo, Taylor Brodie, Kathleen Panak, Sweta Rajan, Derek Cridebring, Michael J. Demeure. Detection of early tissue changes on historical CT scans in the regions of the pancreas gland that subsequently develop adenocarcinoma using quantitative textural analysis and fat fraction analysis [abstract]. In: Proceedings of the AACR Virtual Special Conference on Pancreatic Cancer; 2021 Sep 29-30. Philadelphia (PA): AACR; Cancer Res 2021;81(22 Suppl):Abstract nr PO-010.