State-of-the-Art and Future Prospects of Ion Beam Therapy: Physical and Radiobiological Aspects

The number of facilities offering radiotherapy with protons or heavier ions is continuously increasing; worldwide, more than 160 000 patients have been treated with protons, and more than 25 000 with heavier ions. Despite this substantial clinical experience, there is still a need for further developments and improvements which are specific to the properties of particle beams. This contribution briefly summarizes the main physical and radiobiological properties of ion beams which make them favorable for application in tumor therapy. In addition, major challenges that are currently addressed in different research areas are reviewed. These comprise the fields of biophysical modeling, treatment planning, mitigation of target motion and novel approaches based on particular spatio-temporal beam delivery techniques.


I. INTRODUCTION
I ON BEAM therapy represents a rapidly developing branch of radiotherapy, mainly using proton beams and -to a lower extent -also heavier ions like, e.g., carbon ion beams. After the proposal by Wilson [1] to use ion beams for medical applications, first clinical applications were implemented with proton beams at Berkeley in 1954 [2] and in 1957 at Uppsala [3]; treatments with He beams were initiated in 1957 and with heavier ions in 1975 also at Berkeley [2].
Until the establishment of a dedicated, clinically based proton therapy in Loma Linda with first patient treatments in 1990, therapeutic applications were restricted to accelerator facilities generally built for fundamental physics research purposes. In the following decades, the number of dedicated proton and since 1994 also carbon ion therapy facilities is continuously rising; currently 78 proton facilities and 12 carbon ion facilities are in operation [4]. But in addition, the number of facilities under construction or planned is still substantially increasing.
The main motivation for application of ion beams in radiotherapy is their advantageous depth dose profile, allowing maximizing the dose to the tumor by simultaneously sparing the surrounding normal tissue as compared to conventional photon radiation. Whereas this property is shared by all ion species, in particular heavier ions like carbon ions show an additional advantage with respect to their biological effectiveness. They exhibit an increased biological effectiveness in particular toward lower energies, i.e., in the region where they come to rest when penetrating tissue (the so-called "Bragg peak"). These main characteristics are depicted in a schematic way in Fig. 1. In addition, lateral scattering of heavier ions is substantially reduced as compared to protons, thus allowing to spare normal tissue at the side of treatment fields.
Since the longitudinal extension of a Bragg peak is only in the order of a few mm and thus much smaller than the typical extension of a tumor, several Bragg peaks at different positions in depth have to be superimposed in order to achieve a homogenous coverage of the target volume in depth (see Fig. 2). This variation in depth can be either achieved by This  passive methods, e.g., using ridge filters, or by active energy variation.
Furthermore, also lateral shaping of the treatment field can be achieved by passive as well as active methods. Passive methods are based on collimators, whereas active methods exploit the possibility of magnetic deflection of a narrow pencil beam. Combining active energy variation with magnetic lateral scanning of a pencil beams allows the optimal conformation of the dose also to almost any irregular shaped target volume [5].
Although already in clinical use, the full potential of ion beams in therapy has not yet been exploited. This review will thus briefly describe the state-of-the-art and the fields of ongoing research, addressing major important challenges in ion beam therapy. As a quite broad range of aspects is covered in this review, it is not intended to be exhaustive, but rather pointing to a selection of topics that are considered to be relevant from a biophysicists point of view.

A. Definition of the Relative Biological Effectiveness
In general, ion beams exhibit an increased biological effectiveness as compared to conventional photon radiation, i.e., a lower dose is required to achieve a desired biological effect, e.g., a given level of cell killing. In terms of typical dose response curves depicting the fraction of surviving cells after radiation this is reflected in the steeper slope of survival curves as schematically shown in Fig. 3.
The relative biological effectiveness is defined by the doses required to achieve a given survival level with photons and ion beams, respectively However, the curves for ion irradiation are not only steeper, but at the same time also the shape of the dose response curves changes, as they typically get straighter as compared to the shouldered shape of survival curves observed after conventional photon radiation.
The dose response curves are characterized in general in terms of the linear-quadratic (LQ) model The changes primarily result from the increase of the linear term α, whereas the impact on the quadratic term is less pronounced and subject to larger uncertainties. As a consequence of the more linear dose response curves after ion irradiation, the RBE depends on the effect level for which it is determined. In general, RBE is maximal at low doses and decreases toward higher doses and correspondingly lower cell survival levels. Furthermore, the RBE varies with energy and thus with the stopping power of the ions as characterized by the linear energy transfer (LET).

B. In-Vitro Studies of RBE
In order to study these systematic dependencies, numerous in-vitro studies have been performed to determine the RBE for a wide panel of different cell lines, ion beam species and irradiation conditions. Friedrich et al. [6] have compiled these data in the particle irradiation data ensemble (PIDE), which contains more than 1000 published cell survival curves obtained with a wide variety of different ion beams and cell types and is continuously updated with new data. 1 Based on these in-vitro studies, the fundamental characteristics of ion beam radiation as compared to photon radiation can be summarized as follows (see Fig. 4).
1) RBE rises with LET up to a certain maximum and drops toward higher LET values. 2) The RBE (LET) curves are shifted toward higher LET values with increasing particle mass. 3) RBE decreases with increasing dose and thus decreasing survival level. 4) RBE in general is higher for cells that are radioresistant against conventional photon radiation as compared to cells that are sensitive against photon radiation. The LET dependencies can be interpreted based on the microscopic features of the energy deposition pattern of individual ions. With increasing LET, the local energy deposition within the particle tracks increases, leading to more complex damages and thus an increasing biological effectiveness [7], [8].
At very high LET values, however, saturation effects occur, leading to a corresponding decrease of the biological effectiveness. The cell line dependence can be explained by the corresponding differences in the photon dose response curves: whereas radioresistant cell lines are typically characterized by a pronounced shoulder of the photon dose response curve, very sensitive cell lines exhibit a steeper and straighter photon dose response curve. Therefore, for resistant cell lines a higher gain in effectiveness is expected when comparing high doses to low doses; combining this information with the more pronounced contribution of high local energy deposition for ion beam radiation, a higher RBE can be expected in the case of radioresistant cell lines [9]. The dose dependence, as mentioned above, can be traced back to the differences in the general shape of the dose response curves.
For the dose prescription in ion beam therapy, the increased effectiveness needs to be adequately taken into account. As a consequence of the above mentioned complex dependencies of RBE, this obviously is not a trivial task, and the strategies developed in that direction will be addressed in more detail in Sections III and IV.
As a result of the inverted depth dose profile and the resulting sparing of the normal tissue surrounding the tumor, higher doses per fraction to the tumor are feasible in ion beam therapy as compared to conventional treatment modalities without enhanced side effects. The accurate understanding of the highdose portion of the dose response curves is thus of high importance. Concerning the photon dose response curve there is clear evidence that the pure LQ model that is generally used to characterize the dose response curves is valid only in the low and intermediate dose range, and a transition to a more straight shape is observed toward higher doses (approx. 5-10 Gy) [10]- [13]. This deviation from the LQ-shape might also affect RBE values at high doses; the corresponding analysis of the high-dose part of high-LET dose response curves is therefore of utmost importance for applications of hypofractionation in ion beam therapy [14]. Limitations of the LQ-model might also occur at very low doses (<1 Gy), if cell lines exhibit the so-called "low dose hypersensitivity," which is characterized by a steep initial slope, followed by a transition to the normal LQ-shape. However, this effect is not observed in all cell lines, and the deviation is less critical with respect to RBE determination as the contribution to the overall effect of such low doses is still small.

C. In-Vivo Studies of RBE
Most of the systematic studies that have been performed to characterize the radiobiological properties of ion beams are based on in-vitro cell culture experiments, in particular on measurements of cell survival using the so called "clonogenic assay." In this assay, the continuous proliferative capacity of individual cells and their ability to build colonies with more than 50 daughter cells is used to define the survival of cells. This continuous cell division, however, more resembles the situation of tumor like tissues, but is considered to be less relevant to mimic normal tissue response. In that respect, to distinguish between, e.g., rodent and human cell lines might be less important, as in general the systematic dependencies on LET, dose, etc. do not qualitatively differ between these cell lines.
A more relevant aspect to be considered is that in general in-vitro assays based on single cells do not reflect the high complexity of biological processes, e.g., cell-cell interactions, underlying the radiation response of tissues and are thus of limited value with respect to a direct quantitative translation to the clinical situation. In-vivo animal experiments thus represent the next important step to validate the systematic dependencies described above for other biological endpoints more closely related to the clinical situation.
One important aspect that can be better addressed with invivo experiments is related to therapeutic benefit of different ion species, where the therapeutic benefit is defined by the balance between tumor RBE and surrounding tissue RBE. If the RBE in the tumor is low, but high in the surrounding tissue little advantage is expected for heavier ions compared to protons, whereas in the opposite case of a high tumor RBE and low normal tissue RBE the highest benefit is expected. Strategies for application of ion beams thus should always be based on consideration these differential RBE effects.
The Japanese facilities have chosen, e.g., skin response or intestinal crypt cell survival as normal tissue response to determine the therapeutic ratio/gain of carbon ion irradiation and conclude that larger doses per fraction might be beneficial in carbon ion treatments to maximize the therapeutic benefit [15]- [17]. For the treatment of tumors in the head and neck region, the central nervous system represents another particularly relevant normal tissue system that needs to be spared in radiotherapy treatments. Therefore, in the framework of the pilot project with carbon ions performed at GSI, an extended series of measurements of tolerance doses for the rat spinal cord has been performed, comprising 1 Fx, 2 Fx, 6 Fx, and 18 Fx irradiations [18], [19]. These studies thus cover a broad range of fractionation schemes that come very close to the 20 Fx scheme that was applied in the framework of the pilot project for carbon ion treatments at GSI. Apart from the clinical relevance of this endpoint, the system is particularly suitable for accurate RBE determinations due to the small cross section of the animal's spinal cord. The target can thus be precisely positioned in the radiation field, and differential measurements with varying radiation quality along the penetration depth of a treatment field can be performed with high resolution. More recently, these studies have been substantially extended to a wider range of radiation qualities [20]- [22], and studies are ongoing also with other ion species like, e.g., protons [23], He ions and O ions. Despite the much more complex biological endpoint, in general these studies qualitatively confirm the systematic dependencies with respect to dose-and LET-dependencies as described above for in-vitro cell culture studies.
Comparably few studies have been performed concerning the tumor response in the framework of carbon ion therapy. Ando et al. [16] reported a combined study to determine the biological gain of carbon-ion radiotherapy for the early response of tumor growth delay and against early response of skin reaction in mice. Sørensen et al. [24] compared tumor control, acute skin reactions, and late radiation induced fibrosis in a mouse model. Brownstein et al. [25] analyzed the response of primary sarcomas in a mouse model after carbon ion irradiation. Systematic studies using different sublines of a syngeneic prostate carcinoma model have been reported in [26]- [29]. Three different sublines of this tumor model are available, characterized by different radiosensitivities. As observed also in-vitro, the most resistant tumor showed the highest RBE value, whereas the most sensitive tumor is characterized by the lowest RBE value. It is, however, important to note that this variation in RBE is a consequence of the large variation of sensitivity against photon radiation, whereas after ion beam irradiation the variations were much less pronounced [28], [29] (see Fig. 5). This again is compatible with results from in-vitro studies, showing that in general the range of variation in radiosensitivity observed after low-LET radiation is substantially compressed after high LET radiation [9]. This aspect is also relevant with respect to the heterogeneity of tumor cells, which is expected to affect the overall survival of tumor cells after photon radiation more pronounced as after ion beam irradiation. This results in a steeper TCP curve for ion irradiation, i.e., a larger incremental control probability per percent increase in dose [30].
Another important aspect of high-LET radiation is the reduced dependence of the effectiveness on the oxygen concentration in the surrounding medium. In general, about 3-times higher doses are required to kill fully anoxic cells as for killing of well oxygenated cells. This is described by the oxygen enhancement ratio (OER), defined by the ratio of doses required to kill the same fraction of cells under hypoxic and oxic conditions, respectively. The OER drops with increasing LET, and approaches values close to 1 at very high LET (>100-200 keV/μm) [31]- [34], reflecting the independence of the biological response to the oxygen concentration in this case and indicating an advantage of high-LET radiation in particular in the case of hypoxic tumors.

D. Specific Aspects for Protons
For protons, in general the differences in biological effectiveness as compared to photon radiation are less pronounced due to their lower LET [35]. This motivates simplified approaches to take RBE into account in treatment planning; currently it is thus recommended to simply multiply the absorbed dose by a constant RBE of 1.1 [36]. Nevertheless, in-vitro experiments clearly show a significant rise of RBE toward the end of a spread-out Bragg-peaks (SOBPs), and at the distal edge typically RBE values substantially higher than the clinically applied value of 1.1 are observed [35], [37]. However, reports about unexpected effects in patient treatment that can uniquely be attributed to such an increased RBE are scarce. For example, Peeler et al. [38] reported about the first clinical evidence for a correlation between normal tissue damage and LET in proton treatments of ependynoma. Gensheimer et al. [39] analyzed an overshoot of the proton beam visible in MR images; this overshoot effect might be attributable to an increased RBE at the distal edge [37]. But it remains to be elucidated how the apparent discrepancy between in-vitro and clinical results can be resolved, and the debate about whether variable RBE values larger than 1.1 need to be used in treatment planning is still ongoing [40]- [43]. Additional in-vivo experiments to determine tolerance doses of late responding tissues, e.g., spinal cord, after proton irradiation could help to bridge the gap between in-vitro systems and clinical data. Saager et al. [23] have reported first measurements; however, these experiments were restricted to 1F and 2F irradiations involving very high doses per fraction, and therefore further experiments with fraction doses in the therapeutically relevant regime are highly desirable here.
One potential explanation of this apparent discrepancy might be due to the small volumes in which typically enhanced RBE values are expected in proton therapy. Although partial volume dependent tolerance doses have been analyzed [44], [45], there is still substantial lack of data to estimate tolerance doses particularly for the case that only very small subvolumes of an organ are irradiated. This urgently calls for experiments to better understand the mechanisms relevant for recovery from tissue damage and their role in partial volume irradiation, similar to those reported (see [46]- [48]).

E. Combined Treatments
Finally, with respect to clinical applications the understanding of combined treatments, e.g., chemotherapy + Rx and immunotherapy + Rx are of great interest, and several studies have been reported in this direction. For example, Melzig et al. [49] have combined external carbon ion beam therapy in combination with radioactive labeled antibodies against a cell surface receptor involved in cell growth; they report an activation of the immune-response by carbon ions that markedly enhanced the antibody-based therapy. Ohkubo et al. [50] analyzed the antimetastatic efficacy of a combination of carbon ion radiotherapy and immunotherapy by use of an in vivo murine model; they report the effective inhibition of distant metastasis after this combination therapy. Durante et al. [51], [52] discussed potential mechanisms that are involved in the radiation induced activation of immune pathways.

III. BIOPHYSICAL MODELING
As a consequence of the complex RBE dependencies discussed above, RBE values, e.g., will vary across the tumor and depend on the fractionation scheme used, and in general they can therefore not adequately be represented by a single number for conversion of absorbed dose to RBE-weighted dose. To fully exploit the advantageous properties of ion beams, the systematic dependencies of the RBE have to be considered adequately in treatment planning in ion beam therapy, allowing taking advantage of the vast experience made with conventional photon therapy.
Biophysical models can represent an important tool for treatment planning in that respect. However, the transition from the initial energy deposition to the final observable biological effect after a radiation insult includes numerous complex biological processes and pathways, from which many are still unknown or at least not yet accurately quantified, and any model thus can represent an approximation to reality only. One of the major challenges of modeling in the framework of treatment planning therefore is to find the right balance between accuracy and model complexity, i.e., number of different processes and mechanisms to be taken into account. A higher level of detail corresponds to an increasing number of degrees of freedom by introduction of additional parameters. If, however, the number of degrees of freedom is too high, no significant parameter values can be expected any more when fitting the model to experimental data. In contrast, if a model has too few degrees of freedom, the model is incomplete, corresponding to a reduced predictive power of the model. One has to keep in mind that as a consequence of the required approximations different ways of approximation might be feasible, which may lead to similar predictions, even if assumptions about the underlying mechanisms are different.
At present, two different models are actually used in treatment planning for carbon ion beam therapy: the microdosimetric-kinetic model (MKM) is used in the Japanese facilities, whereas the local effect model (LEM) is used in the European facilities. In both approaches, the characterization of the microscopic energy deposition pattern represents a major ingredient, although the details how this energy deposition pattern is translated into a biological response substantially differs. The characterization of the dose response curve after low-LET radiation represents the second pillar of these models.
The MKM makes use of characterization of energy deposition in micrometer sized volumes. Its original version has been developed by Hawkins [53]- [55], and subsequent further developments have been implemented in the framework of the Japanese heavy ion therapy projects. For example, a correction related to the overkill effect at very high LET has been introduced [56] and implemented in a treatment planning environment [57]. In the meantime, it serves as a replacement for the former experimentally based approach used to optimize the shape of SOBPs in the Japanese treatment planning approach (for details see Section IV).
The LEM is in its original version (LEM I; [58]) is used for treatment planning in the European carbon ion facilities. It resembles the MKM approach in that it directly translates the microscopic energy deposition pattern into an observable biological effect, i.e., cell killing. The model had been further developed and improved to achieve better agreement with experimental data [59], [60]. In the most recent version (LEM IV; [61], [62]), the major change as compared to the LEM I is the introduction of an intermediate step, where first the microscopic distribution of DNA damages is calculated, and based on the evaluation of the clustering properties of these DNA damages the final observable effect is determined. LEM IV has been demonstrated to accurately represent experimental data in-vitro over a larger range of different ion species from protons to oxygen ions [63]- [65]. Furthermore, besides in-vitro experiments also RBE for in-vivo experiments, e.g., the tolerance of the rat spinal cord, can be modeled with LEM IV [21], [22]. Interestingly, the concept of damage classification that had been developed for the LEM IV has been demonstrated to be applicable also to other radiation qualities, and several key aspects like LQ-shape of survival curves, rejoining kinetics, dose rate effects, and cell cycle effects can be correctly modeled by this approach [66]- [69].
Apart from the LEM and the MKM, other approaches are discussed for potential applications in ion beam therapy, as, e.g., reported in [70]- [72]. However, a conceptual comparison of the models has revealed substantial differences, e.g., with respect to the impact of overkill at very high LET and the change of the quadratic component with LET (Fig. 6, see also [73]), suggesting the need for more detailed validation of the models by means of experimental data in order to assess the impact of these model differences on the accuracy of the model prediction.
The demand for RBE modeling is clearly more relevant in the case of heavier ions as for protons, since RBE values are substantially greater for heavy ions, and it is still under discussion in how far a variable RBE instead of using a constant RBE = 1.1 could be beneficial in proton therapy. The MKM and LEM are models applicable over a broad range of particles, covering protons as well. However, numerous simplified, empirical models have been also proposed which are applicable solely to the case of protons, as they are based on certain simplified assumptions that are not valid for heavier ions. A common concept behind many of these models is the characterization of RBE in terms of LET d . Despite this similarity, depending on the datasets that have been used for calibration of the model parameters, the models lead to quite different predictions. For example, Rørvik et al. [74] have compared 13 models that are based on a parametrization of RBE as a function of LET d and other biological parameters like, e.g., the photon α/β-ratio; they report differences for RBE-weighted dose in the Bragg peak region of more than a factor of 1.5, thus highlighting substantial uncertainties in RBE predictions of these models. However, no direct comparisons to experimental data have been performed in this article.
With increasing number of models, comparisons, and systematic tests/validations against experimental data are of increasing importance in order to allow the choice of an appropriate model for treatment planning. It would be extremely helpful here to agree on a common set of experimental data which are most relevant for model testing. As a minimum requirement for a general purpose model, simultaneous prediction of RBE (LET) dependencies in-vitro for p, He, and C ions and for RBE α and RBE 10 for different cell lines covering a broad range of sensitivities might be a starting point. Furthermore, applicability to predict RBE values in-vivo can be considered as prerequisite for clinical application. However, the choice of the relevant endpoints will largely depend on the specific clinical application, and therefore guidance from a clinical perspective would be highly desirable here. This also includes the definition of potential gaps and limitations in experimental data as well as in the modeling approaches, which are needed in order to determine the strategy for potential model improvements.

A. Physical Aspects
Both physical and biological aspects represent an important challenge for treatment planning in ion beam therapy. Accurate stopping power determination is an extremely relevant issue to determine the range of ions in tissue, as success of ion beam therapy will largely depend on the accuracy with which the potential precision of ion beams can be actually exploited in patient treatments. Ion beam therapy is thus much more sensitive to all aspects of range uncertainty and positioning errors [75]- [77]. Dual-energy CT, ion computed tomography and ion beam radiography are discussed as potential solutions to substantially reduce uncertainties in range estimation [78]- [84]. With respect to range verification, prompt gamma imaging is one of the promising new approaches discussed as alternative to, e.g., PET-based methods [85]- [88]. Due to the timing, prompt gammas are particularly suitable for online imaging, and a major advantage compared to PET is that distortions resulting from wash-out effects can be avoided. First clinical installations have demonstrated the feasibility and the particular benefits of prompt gamma imaging and its potential for improving the accuracy of particle therapy [89], [90].
Accurate planning is specifically challenging for complex treatment situations, involving large heterogeneities, or metal implants [91]. With respect to the physical characterization of the radiation field, Monte-Carlo algorithms have proven to represent an accurate tool in treatment planning also in ion beam therapy [92]- [94]. In combination with the appropriate nuclear physics approaches, they are particularly suitable to represent the complex nuclear fragmentation and scattering processes that are indispensable for an accurate description of the radiation field [95]. Monte-Carlo calculations also represent an important tool to characterize microscopic energy deposition features, e.g., in novel treatment approaches aiming at the enhancement of effectiveness by combining ion beam irradiation with nanoparticles [96]- [98].
More general aspects of treatment planning developments refer, e.g., to advanced automization techniques, multicriteria optimization and/or aspects of plan robustness [99]. For example, the optimal choice of beam angle directions at present typically requires manual interventions and thus offers high potential for further improvements based on automization. Multicriteria optimization -aiming at the right balance between eradicating the cancer and avoiding unacceptable injury to normal tissues -inherently includes tradeoffs: more conformal dose to the target comes at a cost of higher normal tissue complications, and sparing of one organ at risk gives increased dose to another. The aim of multicriteria optimization is to manage the inherent multidimensional tradeoffs better, and to present these tradeoffs in a transparent way, thus enabling a more efficient planning process [100]- [102]. Furthermore, comparably small uncertainties in patient positioning, stopping power calculations and anatomical changes might have significant consequences like hot or cold spots in the resulting dose distribution in the patient. A major topic in treatment planning therefore is the plan robustness against these uncertainties [103]- [106]. This also includes the impact of a variable RBE in proton treatment planning, which could be related to hot spots in the RBE-weighted dose distribution at the distal edge of the field or in general in regions characterized by an increased LET [107]- [109].

B. Biological Aspects
With respect to biological aspects, the major goal of biologically optimized treatment planning is to achieve a homogenous RBE-weighted dose coverage of the target volume. This requires the reduction of the absorbed dose toward the distal end of an SOBP in order to compensate for the increased biological effectiveness as a consequence of the increase of LET with depth (see Fig. 2). The models described in Section III represent important tools to determine RBE in treatment planning for ion beam therapy, although different strategies are followed in the Japanese and the European approach in that respect. The basic idea of the Japanese approach is based on a hybrid approach, combining aspects of the model-based characterization of RBE for cell killing with clinical experience in neutron therapy [110]. This approach exploits the fact that under certain conditions carbon ion beams and neutron beams exhibit similar radiobiological properties. The European approach based on the LEM, in contrast, is solely based on biophysical modeling directly linking photon dose response curves to ion beam response curves [111], [112].
RBE modeling is frequently considered to be subject to large uncertainties, depending on the accuracy with which the model input parameters are known [113]. However, one has to keep in mind that similar uncertainties need to be considered in conventional photon therapy when, e.g., α/βvalues are used to determine dose corrections in the case of changing fractionation schemes. Actually, in particular at lower doses as they are clinically applied the impact of uncertainties can expected to be less pronounced after ion beam irradiation as compared to photon irradiation [114]. This is a consequence of the less pronounced differential, e.g., cell line specific, effects after high-LET radiation [9], [28]. Nonetheless, attempts have been reported to circumvent biological modeling and refer to LET distributions as a surrogate for RBE distributions. This approach is particularly promoted for proton therapy [115]- [117], as in this case actually due to the low RBE values uncertainties of RBE might be higher than the value itself. It has, however, been demonstrated that LET is not an accurate estimator of RBE [118], because for a given dose-mean LET value both the width of the LET distribution as well as the particle species will affect the RBE [118], [119].
In order to get less dependent on uncertainties in RBE and thus on potential hot spots in the RBE-weighted dose distribution, strategies have been proposed to more homogenously distribute the high-LET components throughout the target region [115], [116], [120]- [122]. This is of interest also for avoiding high-LET components in critical structures of the surrounding normal tissue, since taking into account safety margins, the distal edge of a treatment field and thus the region with highest LET and RBE values is typically located in the normal tissue.
Another approach aims at specifically distributing the high-LET component to the most resistant regions of a tumor, e.g., to the hypoxic region (so called LET-painting) [123]- [125]. As LET alone is not an accurate descriptor of RBE and thus of the expected biological effectiveness, more advanced approaches aim at directly optimizing the biological effect, e.g., the so called "kill-painting" approach [126], [127] (see Fig. 7). In order to optimally apply these approaches, the accurate characterization of the spatial distribution of sensitivities is required. This includes, e.g., the delineation of hypoxic regions and the quantification of the level of hypoxia; advanced imaging techniques will therefore play a crucial role in this respect. In order to further optimize ion beam treatment, also mixed irradiation with multiple ion species has been proposed, where heavier ions are used to specifically treat the most resistant parts of the tumor, and lighter ions to treat the more sensitive regions. This approach requires highest flexibility in treatment planning as well [127].
A specific aspect of ion beam treatment planning that needs to be considered in particular for intercomparison of treatment plans of different institutions are the different strategies to take into account the increased biological effectiveness of ion beams. Steinsträter et al. [128] and Molinelli et al. [129] have reported, e.g., intercomparison of plans based on the approach developed at HIMAC and at GSI/HIT, respectively, to determine RBE-weighted doses. Although at RBE-weighted doses of 3 · · · 5 Gy/fraction both approaches are similar, at lower as well as at higher doses per fraction substantial differences are expected; at doses relevant, e.g., for single fraction treatments, estimated RBE values can differ by approximately a factor of 2. Therefore, if the same values of prescribed RBE-weighted doses are reported by the Japanese and European centers, the underlying absorbed doses might considerably differ as a consequence of the different RBE values. This finally will lead also to different clinical outcome for the same prescribed dose.

V. MOVING TARGETS AND MOTION MITIGATION STRATEGIES
The treatment of moving targets represents a challenge in ion beam therapy in particular for active beam delivery methods, as the interplay between beam scanning and target movement might lead to undesired interference patterns, finally resulting in inhomogeneous dose distributions. In addition to intrafractional movements, interfractional movements are of interest here, justifying the strong interest in in-room imaging and range verification. Different strategies like, e.g., rescanning, gating and tracking are discussed as potential motion mitigation techniques, each having their specific advantages and disadvantages [130]- [133]. Rescanning is the option easiest to implement, since the dose per fraction only needs to be split into the appropriate number of smaller doses which are then applied multiple times at a given treatment day. This method is thus based on the statistical averaging of positional errors, since dose errors are smoothed out by irradiating the target volume multiple times. Gating is expected to be more accurate in that respect, as the target is irradiated only in defined motion phases characterized by a much smaller residual motion. Tracking theoretically is the most elegant and accurate method, as magnetic scanning can in principle be corrected for the movement of the tumor online [134]. However, in this case the determination of the actual tumor position with high accuracy represents the major challenge, and at present no method seems accurate enough to follow that strategy, since in general the tumor cannot be tracked directly, but usually sampling is based on an external motion surrogates. In order to partially circumvent the interplay between active scanning and target motion, also hybrid techniques combining passive elements with active scanning have been proposed [135].
Precise motion mitigation techniques require quantification of the motion, for example by time resolved computed tomography (4DCT), sampling periodical motion in several motion phases, and the corresponding adaptations of the treatment planning procedures. Individual motion phases can be identified with quasi-static 3-D volumes, e.g., standard CT volumes, for which individual plans are optimized. Treatment then requires the fast switching between the different plans, depending on the particular motion phase [106], [136], [137].

VI. ION SPECIES OTHER THAN PROTONS
AND CARBON IONS Both the Japanese as well as the European facilities have chosen carbon ions as the first high-LET modality for comparison with proton and photon treatments. Having in mind the impact of the ion species on almost all RBE dependencies, it is frequently discussed which ion species would be optimal for therapy. However, as indicated above, the judgement concerning the potential benefit will critically depend on the differential RBE effect in the tumor and the surrounding normal tissue. It will thus primarily depend on the specific radiobiological characteristics of the tissues involved in the treatment field, and thus no general answer can be given to this question, but rather treatment plan comparisons need to be performed for the specific cases in order to allow for a thorough assessment. Nonetheless, some general remarks will be compiled here in order to indicate potential areas of interest for future improvements of ion beam therapy Heavier ions than carbon will particularly be interesting in the case of hypoxic tumors, where reduction of the OER is of importance. Several approaches have been reported to characterize the decrease of OER values with increasing LET [31], [33], [138], [139], indicating a drastic reduction of OER values with LET higher than 100 keV/μ. Recent studies [65], [123], [124], [126] indicate the possible role for 16O beams in this respect. An important aspect here is the accurate characterization of the hypoxic region, which typically is located in the central region of the tumor. The development of appropriate imaging methods will be crucial in that respect [140], [141]. As for heavier ions also the RBE in the entrance channel/healthy tissue in front of tumor might be undesirably high, also combined treatments with light and heavier ions might be feasible. In this case, the heavier ions are solely used to treat the central/hypoxic part of the tumor, and the lighter ions for the treatment of the remaining part of the tumor, so that the dose delivered to the proximal normal tissue by high-LET ions can be kept to a minimum [65]. Mixed radiation are also discussed by Brahme [142] and Inaniwa et al. [143].
Lighter ions like, e.g., He are also discussed for therapeutic application [63], [144]. From the physical point of view, they might be advantageous since projectile fragmentation is less pronounced for He ions as compared to carbon ions, and lateral scattering is less pronounced as compared to protons. This is further confirmed by the results reported by Tessonnier et al. [145], [146], indicating that helium ions could afford a more conformal treatment and in turn, increased tumor control.
From the radiobiological point of view, the RBE of He is closer to protons, which is considered to be beneficial, e.g., in the treatment, e.g., of pediatric patients, where the potential risk of increased secondary cancer induction resulting from elevated RBE needs to be kept minimal.
As the interplay between the physical and biological properties depends on different parameters like the ion species, dose level, tissue characteristics and beam configuration, idealized treatment planning studies are expected to represent an appropriate guide for preselection of cases and choice of the optimal ion species. For example, Grün et al. [64] reported about such treatment plan comparisons using simple target volume and beam angle configurations. The comparison of p, He, and C ions suggests a strong dependence of the advantages of the three ions on the dose-level, tissue type, and field configuration. In terms of conformity, i.e., dose to the normal tissue, a clear gain is expected using carbon or helium ions compared to protons.
These results are in line with the conclusions drawn by Remmes et al. [147] that no single ion is optimal for all treatment scenarios. The heavier ions are superior in cases in which the α/β ratio of the target tissue is low and the α/β ratio of normal tissue is high, and protons are superior in the opposite circumstances. Lithium and beryllium appear to offer dose advantages similar to carbon, with a considerably lower normal tissue dose when the α/β ratio in the target tissue is high and the α/β ratio in the normal tissue is low.
Simultaneous acceleration of He + C has been proposed by Mazzucconi et al. [148] for combined imaging and treatment: fully stripped nuclei of 12 C and 4 He have the same magnetic rigidity; therefore, the two species can be accelerated together and a mixed particle beam delivered to the patient. With the Fig. 8. Principle of grid therapy: A (regular) pattern of narrow beams enters the body; with increasing depth, scattering processes lead to widening of the beams and thus to superposition and consequently a more homogeneous irradiation in depth, i.e., in the target region. same energy per nucleon, the range of 4 He is about three times the 12 C one. Therefore, carbon ions stopping in the tumor can be used for tumor cure, while helium, emerging from the patient, can be used for imaging by detecting and measuring the residual range and position of He.

VII. NOVEL TREATMENT APPROACHES
In recent years, two treatment modalities that are related to a specific spatio-temporal characteristics of the dose delivery-although already discussed for conventional photon radiation-have attracted substantial attention and triggered a series of investigations to better characterize their potential advantages in tumor therapy. They are based either on ultrahigh dose-rates (so called FLASH irradiation), or on radiation fields focused to narrow regions (so called grid-, microbeam-, or minibeam therapy) in the entrance region of the treatment field.

A. Minibeam Therapy
The basic idea of the latter has been exploited already decades ago [149], [150]. By focusing or collimating the beam to regular grids or spots at the entrance of the beam to the body (see Fig. 8) certain regions, e.g., of the skin remain undamaged, thus resulting in an enhanced recovery in the superficial layers of the normal tissue in front of the tumor. With increasing depth, scattering of the beam leads to increased filling of the dose valleys in between the grid structure, leading to a more homogenous irradiation of the deep-seated locations, i.e., the tumor. The particular role of small, unirradiated regions interlacing with high dose regions has been demonstrated to be beneficial also in the above mentioned studies of the rat spinal cord [46], [47], but is considered to be a generally valid principle also for other tissues. The corresponding increase of the normal tissue tolerance has been demonstrated for a variety of different dose delivery patterns, using different geometries as, e.g., parallel lines, grids, and point patterns with different beam width.
Although grid/sieve therapy have been originally proposed for conventional radiation like X-rays and high energy photon radiation, it has been recognized that the use of ion beam radiation in combination with grid therapy might show additional advantages [151]- [159] (see Fig. 8). In particular the less pronounced scattering of heavier ions like, e.g., carbon ions might have advantages here since it allows to preserve the grid pattern and with that the sparing effect also in regions close to the target. On the other hand, it requires additional techniques like interlaced irradiation with multiple fields in order to assure the homogenous irradiation of the target region [160], [161]. Since these interlacing techniques require extremely accurate positioning of the target, it remains to be clarified in how far this accuracy can be achieved in the clinical practice.
Also in general, a need is seen to specify ideal dosimetry parameters and to formulate robust clinical indications and guidelines for optimal standardized care using grid and minibeam therapy [162].

B. Ultrahigh Dose Rates
The second approach makes use of extremely high dose rates, which have been reported to also lead to an increased tolerance of the normal tissues surrounding the tumor without compromising the reduction of tumor growth [163]- [167]. Also here, the detailed mechanisms need to be clarified, as, e.g., in general in-vitro no significant difference between pulsed irradiation with ultrahigh dose rates and continuous irradiation with lower, conventional dose rates (∼Gy/min) have been found [168], [169]. Combination of ion beam therapy with FLASH therapy is also discussed, and studies are ongoing to demonstrate the potential synergy between the two radiation modalities [170].
Ultrahigh dose rates are also discussed in the framework of laser generated particle beams, where ions are accelerated by shooting thin foils with extremely short (fs) laser pulses [171]. In line with the results reported above, Zlobinskaya et al. [172] found no difference in the tumor growth response between conventional irradiation and pulsed proton irradiation of human tumor xenografts in nude mice. However, these authors do not analyze the corresponding normal tissue effects.
In general, more detailed studies related not only to tumor growth delay, but also to tumor cure would be highly desirable here to complement the existing studies, as the endpoint "growth delay" may not be sensitive enough to detect differences in the response that finally determine tumor cure, being the most relevant endpoint for the determination of the therapeutic window.

C. Treatment of Noncancer Diseases
Another research area of growing interest is related to the application of ion beams for the treatment of noncancer diseases. Also the early clinical trials included already, e.g., the treatment of noncancer diseases like arteriovenous malformations (AVM). More recently, the treatment of heart arrhythmia with carbon ion beams has been discussed, taking advantage of the extremely localized energy deposition of carbon ions in a small, mm-sized Bragg peak region. This makes ion beams particularly suitable to damage small subareas of the heart with very high doses, allowing to create chronic lesions that locally interrupt cardiac conduction and therefore enable treatment of heart rhythm disorders completely noninvasively while sparing surrounding tissues [173]- [176].
The principle of using external radiation for the treatment of heart arrhythmia has been also reported for highly conformal photon therapy (SBRT) [177]- [179]. However, ion beams allow a substantially better focusing of the high dose area and thus a potentially higher effectiveness in achieving the desired effect without exceeding the tolerance doses in the surrounding normal tissue.
A particular challenge for this type of application represents the movement of the target region as a consequence of the heart beat as well as of the breathing and corresponding movement of the whole heart. Appropriate measures have thus to be taken to compensate/mitigate these motion effects.

VIII. SUMMARY AND CONCLUSION
Despite considerable experience with both protons and ion beams in radiotherapy applications, still the potential of these radiation modalities is not yet fully exploited. Therefore, both on the physical/technical as well as radiobiological side a lot of research activities are ongoing, which have been briefly addressed in this contribution. Major aims of these research activities should be considered as follows.
1) Technical improvements to enhance conformity of the treatment even in challenging situations, e.g., for moving targets. 2) A better understanding of the radiobiological properties of ion beams in particular for the case of partial volume irradiations. 3) Investigations of the role of the immune response and the potential impact of high-LET radiation modalities to optimally exploit the immune response. 4) Preclinical experiments to elucidate the potential advantage of FLASH and minibeam irradiation techniques using ion beams. 5) Combination of imaging techniques to accurately characterize resistant subvolumes (e.g., hypoxic areas) with advanced treatment planning strategies to appropriately adapt the prescribed dose in these regions accordingly. 6) Extension of application to noncancer diseases. These research activities then will help to accurately define and identify patient groups that will mostly benefit from the different treatment options.