The physics of radioembolization

Abstract

Radioembolization is an established treatment for chemoresistant and unresectable liver cancers. Currently, treatment planning is often based on semi-empirical methods, which yield acceptable toxicity profiles and have enabled the large-scale application in a palliative setting. However, recently, five large randomized controlled trials using resin microspheres failed to demonstrate a significant improvement in either progression-free survival or overall survival in both hepatocellular carcinoma and metastatic colorectal cancer. One reason for this might be that the activity prescription methods used in these studies are suboptimal for many patients.In this review, the current dosimetric methods and their caveats are evaluated. Furthermore, the current state-of-the-art of image-guided dosimetry and advanced radiobiological modeling is reviewed from a physics' perspective. The current literature is explored for the observation of robust dose-response relationships followed by an overview of recent advancements in quantitative image reconstruction in relation to image-guided dosimetry.This review is concluded with a discussion on areas where further research is necessary in order to arrive at a personalized treatment method that provides optimal tumor control and is clinically feasible.

Keywords: Dose-effect relationship; Dosimetry; Personalized medicine; Radiobiological model; Radioembolization; Theranostics.

Fig. 1

Adapted from [27]. Absorbed dose to the whole liver was not correlated to the administered activity (a). However, liver weight was negatively correlated with whole liver absorbed dose (r = − 0.723, P < 0.001), leading to patients with small liver being relatively over-dosed and patients with larger liver under-dosed (b)

Fig. 2

Exemplar case where a VOI delineation based on SPECT thresholding only (blue contour) does not match the CT-based anatomical tumor definition (teal contour). The mismatch results in a difference in tumor volume and mean tumor uptake

Fig. 3

Example of a large neuroendocrine tumor, which was treated with glass microspheres. Activity was prescribed according to the MIRD mono-compartment method to reach 120 Gy. According to the PM model, the average absorbed dose to the tumor was 150 Gy. The patient has shown no response after treatment (RECIST, mRECIST, and EASL). The contrast-enhanced CT shows the tumor as a large enhanced area (orange solid line) and necrosis (yellow dotted line) (a). A strong absorbed dose inhomogeneity can be observed (b). Voxel-based dosimetry and radiobiological models may account for such absorbed dose inhomogeneities

Fig. 4

Hypothetical cDVHs illustrating key concepts in voxel-based dosimetry which may be used for outcome prediction. In panel (a) the situation of the red absorbed dose distribution may be expected to have a smaller impact on the tissue under consideration (less toxic or less tumor kill). This is also reflected in the D₇₀ and V₁₀₀ being lower for the red than that for the blue curve. Due to highly heterogeneous absorbed dose distributions, which is typical for radioembolization, two different cases with cDVHs as depicted in panel b might occur. Which of these cDVHs may be expected to have a larger effect on the tissue, is ambiguous (same D₇₀ and V₁₀₀) and might depend on the tissue type. c Depicts the hypothetical differences in equivalent uniform doses (EUD), derived from the situation in panel b, potentially resolving the ambiguity

Fig. 5

Simulated arterial tree (a) and subsequently simulated microsphere distribution after flow through the arterial tree (b, c), which explains PET ‘mottled’ look often found in patients (d) but not in phantom scans (e). This research was originally published in JNM [72]. Copyright by the Society of Nuclear Medicine and Molecular Imaging, Inc.

Fig. 6

a Small clusters (white arrow) and large clusters (black arrow) are apparent in the Monte Carlo simulations by Pasciak. These simulated distributions seem to be consistent with the histological findings of (amongst others) Högberg (b, c, d). Panel a was originally published in JNM [75]. Copyright by the Society of Nuclear Medicine and Molecular Imaging, Inc. Other panels are adapted from [74]