Research

Particle therapy with light ions, i.e. primary beam delivered ranging from proton up to carbon and oxygen, exploits the advantageous depth-dose distribution. The latter, referred as Bragg peak, is superior to the intrinsic depth-dose distributions of conventional radiotherapy particles due to finite range of the primary particles and to the characteristic increase of the dose deposition towards the end of the range followed by a steep fall-off to the no-dose region. Such advantageous distribution requires superior control of the positioning of the Bragg peak. 

Our research investigates the physics interaction of the ion beams with matter for monitoring of the delivered radiation. Ion beams traveling through matter mostly lose energy in collisions with the electrons of the target atoms leading to the Bragg peak dose deposition and determining the particle range. Nuclear collisions play a minor role in the energy loss, nonetheless, for purposes of ion-beam therapy, it is highly relevant to correctly characterize these in order to achieve a correct description of the primaries’ fragments dose and to describe and monitor the products of the nucleus-nucleus interactions. The latter are the focus of our group research. 

Proton, Helium and Carbon beams interacting with the elements of human tissues produce excited nuclear states. Depending on the produced state, the excited nuclei can decay to a stable state via multiple channels. For means of beam range verification, are relevant the states emitting gamma radiation leaving the target volume, which could be measured with a detector system. Possible gamma sources are: beta-plus decaying nuclei (PET Imaging) and excited nuclei relaxing with the emission of electromagnetic quanta (Prompt Gamma). The latter are favorable due to the prompt emission of the radiation and due to the presence of discrete lines in the emitted spectrum. The prompt emission allows for an online monitoring of the delivered radiation, while resolving the spectrum (Prompt Gamma Spectroscopy) allows to distinguish the different excited nuclear states and retrieve information on the beam residual energy and elemental composition of the target tissue. 

We are currently developing an high energy resolution detector system to perform spectroscopy of the prompt gamma radiation emitted during proton, Helium and Carbon ion therapy. We perform Monte Carlo radiation transport simulations to investigate the nuclear excited states produced by the different ion beams and to optimize the detection of the Prompt Gamma. Our aim is to perform online monitoring of the delivered Bragg peak during therapy.

Recently, this research emerged again with the advent of proton therapy. The proton therapy planning system requires knowledge of the proton stopping power within the patient, which can be measured by particle tomography. Currently this quantity is clinically obtained through a conversion from x-ray tomography Hounsfield units (Schneider et al., 1996). Such a process introduces large uncertainties in planning and reduces the flexibility and advantages of proton treatment (Matsufuji et al., 1998; Schaffner and Pedroni, 1998; Chvetsov and Paige, 2010; Yang et al., 2012). It has been proven that single-event pCT could help reduce the uncertainty by directly measuring the stopping power in the patient (Zygmanski et al., 2000). Moreover, particle imaging possesses several clinical and diagnostic qualities. It has a higher density resolution, a significantly lower noise level, and lower dose to the patient (Schulte et al., 2005; Depauw and Seco, 2011) than conventional x-ray CT imaging. Finally, pCT suffers from different artifacts than x-ray CT (Depauw and Seco, 2011). However, one of the major problems encountered in pCT is the lower spatial resolution compared to x-ray CT.

The multiple deflections particles suffer throughout their path, known as multiple Coulomb scattering (MCS), reduce the spatial resolution of the images acquired. Consequently, the conventional x-ray tomographic reconstruction algorithm struggles when using unaltered proton radiographies to reconstruct the proton CT. To solve the problem of MCS, accurate proton path estimate methods were developed and are constantly improved. However, heavier ions suffer less from MCS compared to protons due to smaller average angular deviations and are viable candidates to acquire tomographic images of higher quality compared to proton CT images even with the currently available path reconstruction algorithms. In particular, helium is hypothesized to be the ion producing the highest image resolution, while minimizing the physical dose to the patient when compared to heavier ions, since the path estimates maximum RMS deviation was found to be smallest for helium ions (among ions from hydrogen to carbon), when compared for equal initial energy. A depiction of the maximum RMS deviation of three different path estimates for all ion species up to carbon can be found in Figure 1 (Collins-Fekete et al., 2017).

For this reason, our research on tomographic and radiographic imaging with helium ions. Simulations were implemented using the Geant4 simulation toolkit (Agostinelli et al., 2003; Allison et al., 2006) and the Geant4 wrapping tool TOPAS (TOol for PArticle Simulation, Perl et al., 2012) to investigate the potential of helium CT scans. As a preliminary result of the TOPAS study, a simulated CT image of the Catphan® (Catphan® 600 series The Phantom Laboraty, Salem, NY) high resolution phantom (CTP528) is shown in Figure 2.

First experiments on helium CT imaging were performed in December 2016 in collaboration with Loma Linda University (LLU) and the University of California Santa Cruz (UCSC) using their latest proton CT scanning system and testing it for helium ion imaging at the Heidelberg Ion Beam Therapy Center (HIT). The scanner was developed in a collaboration among Loma Linda University (LLU), University of California SantaCruz (UCSC), and California State University, San Bernardino (CSUSB) in 2011 and is capable of single event trakcking making the usage of the recently established most likely path algorithm (Schulte et al., 2008) possible. The spatial resolution of such a scanner was shown to be of about 5 lp/cm for protons (Plautz et al., 2016) and the TOPAS simulation showed that the RSP values can be obtained with an average error of 0.34% for inserts of different materials in a water phantom (Piersimoni et al., 2017). This imaging technique proved to be a precise and accurate imaging tool, rivaling the current x-ray based techniques, related to the advantage of being directly sensitive to proton stopping power rather than photon attenuation coefficients. Preliminary results from the experiments at HIT have shown that helium ions further increase the achievable image quality and therefore further enhance the potential of the particle CT scanning system. First results of a helium CT scan of the CIRS pediatric head phantom model HN715 (CIRS, Norfolk, Virginia, USA) are shown in Figure 3.

In the future, based on the simulations and the experimental results acquired at HIT, the CT scanning system is to be improved and optimized for helium ion imaging.

Radiation therapy planning and dose calculation requires the knowledge of the density of the irradiated tissues. Today, computer tomography images (CTs) are the gold standard for determining the density distribution inside a patient's body. These so called planning CTs are a reliable method but imply a statistical risk due to the ionizing radiation and require additional efforts for the CT acquisition. In many cases planning CTs do not reveal additional diagnostic value due to the lack of contrast agent and the fact that often, MRIs are acquired earier in the diagnostic process and exhibit a higher soft tissue contrast and resolution. In the recent past efforts have been undertaken to derive so called pseudo-CTs from MRIs to allow for radiation therapy planning based on MRI only.  [Andreasen:2016, Su:2015, Arabi:2016, Merida:2015, vanderMeer2016, Torradocarvajal:2016, Rank:2013].

Since the MR-intensitites are not directly correlated to the physical density of the tissue, the generation of pseudo-CTs is non-trivial, but MRI based radiotherapy implies several advantages over the classical CT based approach. The benefits include: 

1. no spatial fusion error when transfering contours generated on the diagnostic MR, 

2. reduction of cost for the additional treatment, 

3. no extra ionizing radiation. 

These benefits especially apply to children, since many of the young patients need narcosis for the planning CT acquisition, in order to keep still. The new technique would allow to spare the planning CT and thus the narcosis. Additionally, pediatric patients are more prone to ionizing radiation. Leukemia and cranial tumors are the two most common malignities in pediatric oncology. Many cranial tumors in pediatric patients are treated by radiation therapy cite [Conter:1997], in some cases the treatment protocols of leukemia patients include radiation therapy of the head to minimize the probability of relapse. For this reason, this work aims at providing sufficient methods to derive pseudo-CTs from cranial MRIs for radiation therapy, with the prospect of helping a variety of young and grown-up patients. Three main approaches for pseudo-CT generation from MRI generation were published in the recent years:voxel-based [Berker:2012, Rank:2013b], atlas-based [Downling:2012, Uh:2014, Izquierdo-Garcia:2014, Sjolund:2014] and patch-based [Andreasen:2015, Torradocarvajal:2016]. Voxel-based algorithms convert MRI-intesities to pseudo-HU values voxel by voxel deriving translation rules from an available patient dataset containing MRIs and clinical CT. Atlas-based approaches deformably register preacquired MRI/CT pairs to the input MRI and combine multiple of the co-registered atlas CTs to calculate the output pseudo-CT. Patch-based algorithms extract subvolumes (patches) from the MRI input image and assign pseudo-HU by searching for similar patches in atlases containing coregistered MRIs and CTs. The CT-values corresponding to the found similar patches are then used to calculate the pseudo-CTs.Especially the so far published realisations of voxel-based algorithms required UTE (ultra short echo MRI sequences) MRIs to produce pseudo-CTs of sufficient quality. 

We are developing two new algorithms to be used in standard protocol MRIs (MP-RAGE): a voxel-based approach with regioning based on localized correlations and a new patch-based approach with multi-modal input MRIs and an additonal recursive step. The developed algorithms were tested on a dataset of 15 patients, who were treated for brain tumors.

New Algorithm developed for Brain in Following Paper

Deep Convolution Neural Network (DCNN) multi-plane approach to synthetic CT generation from MR images - application in brain proton therapy (link)