Introduction

Despite strides in cancer treatment, pancreatic ductal adenocarcinoma (PDAC) remains at a 5-year survival rate lower than 5% –a rate nearly unaffected for a century and which makes it the fourth major cause of cancer-related deaths.[1] This prognosis reflects PDAC’s high metastatic index, coupled to a paucity of early detection methods.[2] By contrast to what happens with other malignancies MR plays a relatively minor role in detecting PDACs, which generally lack contrast vis-à-vis surrounding tissues, and for which contrast agents do not extravasate efficiently. Evidences of differentiated pancreatic cancer metabolism have recently been demonstrated by Hyperpolarized MRSI.[3] Recent work has also shown potential correlations between diffusion MRI (DWI) and tissue fibrosis [4] –yet these measurements are uncertain when implemented using EPI in regions that, like the abdomen, are affected by motions and close to fat/air/water interfaces. The goal of this work is to explore the usefulness of emerging MRI forms, and in particular of robust DWI approaches based on SPatiotemporal ENcoding (SPEN,[5]) to develop new contrasts that can target this disease. DWI and DTI was complemented by relaxation (T1, T2) and magnetization-transfer (MT) measurements as well as with hyperpolarized ¹³C metabolic imaging, in the search for a combination that might serve as a toolkit for early-detection purposes. Results observed in a preclinical model are here reported.

Results

PDAC tumors could barely be identified by anatomical ¹H MRI until after ca. 14 days post-implantation –a period henceforth called the early-detection stage. Different ¹H MR contrasts were explored throughout this and later periods. Neither T2 nor MT weighting/mapping could identify the tumor until weeks following implantation, when it had already reached ≈1cm and was visible in the anatomic image (Figure 1). It has been reported that in humans, pancreatic tumors have lower ADCs than surrounding tissue;[6] due to the presence of motion, fat and air, however, EPI-based DWI measurements could barely observe tumors on the basis of this behavior. Robust SPEN-based DWI techniques, by contrast, could unambiguously observe in vivo PDACs in mice (Fig. 2), with isotropic diffusivities of ≈7x10-4 mm²s^-1 compared to surrounding values of 1.1x10^-3 mm²s^-1. In addition (not shown) diffusion tensor properties with specific alignment were detected by SPEN. This agreed with more sensitive results collected in vitro on surgically extracted tumors (Fig. 3), according to which both a reduced diffusivity and kurtosis could be related to a dense, locally ordered morphology revealed for the tumorous stroma by staining and microscopy. Interestingly, the robustness of in vivo SPEN DWI experiments successfully identified the tumors with their anomalously low ADCs, when these were only ≈2 mm in diameter (Fig. 4A). The sole other ¹H MRI contrast that highlighted the presence of the tumor was the T1, which when mapped lead to slightly longer values (≈1.45 s) than most of the tissues in the abdominal region (0.9 s, Figure 4B). In addition to ¹H measurements, MR’s competence to detect PDAC tumors was assayed by injection of hyperpolarized ¹³C₁-pyruvate. While ¹³C MRSI results were characterized by much poorer spatial resolutions than ¹H counterparts, they could clearly identify the presence of even incipient tumors by a large increase in the ¹³C₁-lactate production (Fig. 5): The most intense lactate signals tended to originate from the approximate center of the tumor, and a several-fold increase in the Lac/Pyr ratio were observed when compared against ratios arising from other abdominal regions or from controls.

Discussion & Conclusion

A comprehensive multimodal exploration of contrasts for early PDAC detection was conducted in a murine model. Three potentially translatable contrast sources were identified: SPEN-based DWI/DTI, T1 mapping via Inversion Recovery, and hyperpolarized ¹³C MRSI. Findings from these methods could jointly reveal the presence of PDAC even at an early stage, when the tumor is small and treatable yet undetectable by anatomical ¹H images. It remains to be seen whether these tools can also highlight malignancies in human cases, and whether they can distinguish malignancies from pancreatic inflammations. Research into these avenues is currently in progress.

Acknowledgements

This work was supported by the Kimmel Institute for Magnetic Resonance (Weizmann Institute), the Israel Science Foundation (grants 2508/17 and 965/18), a Thompson Family Foundation grant, and the EU Horizon 2020 programme (Marie Sklodowska-Curie Grant 642773).