Image guided therapy – Do we get the picture?

Cancer therapy has made great progress through recent decades of clinical development and research, with major achievements within medicine, biology, chemistry, physics, and computer sciences. In the field of radiation oncology, major steps forward are expected from the wider introduction and development of proton and particle therapy, making this technology now available to many patients in the industrialized world [1–6]. However, it is of great importance that such high-cost radiotherapy (RT) facilities are accompanied by state-of-the-art diagnostic equipment, including positron emission tomography (PET)/computed tomography (CT) and magnetic resonance imaging (MRI). Furthermore, we have most likely just seen the beginning with respect to protein and gene expression screens and novel applications in cancer therapy [7–9]. Although such screens enable the use of targeted drugs tailored to the individual's genetic profile, the tests do not necessarily reduce the number of patients eligible for RT. Most likely, we may see more individualized RT dose regimens and tailored drugs based on the risk profiles. Also, biopsy-based assays are less suitable for assessing the response of tumors and normal tissues during therapy, while non-invasive medical imaging is ideal for this purpose [10]. Although we currently have an extremely high geometric precision in radiation dose delivery (typically in the order of a few mm), target delineation uncertainties are often in the range of 5–10 mm, depending on tumor site. Therefore, does it make sense to pursue an even higher accuracy in RT dose delivery, despite that the high-dose volume in the patient is limited by other factors? Or can more progress be made through tighter integration between diagnostics and therapy? For the latter, in order to both select patients of given risk profile to a tailored (most likely multi-modal) treatment and to guide RT, we need refined non-invasive diagnostics. A series of publications in this issue of Acta Oncologica nicely feature recent progress in cancer therapy where advanced medical imaging is used, either alone or in conjunction with targeted chemotherapy or RT.

Selecting the best imaging modality for identification of different tumor entities is a challenging task. Koolen et al. [11] presents a study on the accuracy of ¹⁸F-fluorodeoxyglucose (¹⁸F-FDG) PET/CT for visualization and staging of early-stage breast cancers. Of the 62 women included, all but one of the tumors larger than 10 mm in diameter were visible on PET/CT. Although only 60% of the small tumors (< 10 mm) could be detected by this method, all triple negative and HER2-positive tumors could be visualized and showed high ¹⁸F-FDG uptake. As such tumors may be treatment resistant, the authors argue that patients with these tumor profiles may benefit from treatment monitoring using ¹⁸F-FDG PET/CT in order to assess response to treatment. This was indeed the topic of Revheim et al.'s current study [12] on patients with gastrointestinal stromal tumors (GISTs), given targeted treatment with Glivec/Imatinib. The patients were investigated at two instances during treatment using diffusion weighted imaging (DWI), ¹⁸F-FDG PET and conventional CT. It was found that the apparent diffusion coefficient (ADC), derived from the DWI series, was the best parameter to predict the long-term response after three months of treatment. ADC is an appealing parameter for monitoring purposes, as it is believed to reflect tumor cell density. However, conventional CT taken early in treatment was unable to detect these response patterns, although changes in tissue density also was evaluated here. We eagerly await more mature results from the authors’ ongoing clinical trial. DWI, and the derived ADC parameter, was also used in the work by Hompland et al. [13]. Together with dynamic contrast enhanced MRI (DCE-MRI) and immunohistochemistry, they studied the supporting connective tissue (stroma) in cervical carcinoma xenografts. Tumor stroma may play a role in the progression of malignant tumors, and non-invasive assessment of this tumor component is thus relevant. It was found that ADC was significantly negatively associated with the fraction of connective tissue in the tumors, suggesting that water diffusion in the tumors was greatly restricted by the stroma. Conversely, image parameters derived from DCE-MRI did not show these associations.

In the study by Borren et al. presented in this issue [14], DWI and DCE-MRI was performed on prostate cancer patients before undergoing prostatectomy. From histological sections of the surgical specimens, cell density (CD) and microvessel density (MVD) was assessed, in addition to Gleason score. Comparing MR images with histology using quantitative methods, it was shown that tumor voxels with both low CD and MVD could not be identified with the combined imaging methods. Still, the authors argue that since high CD and high MVD represents aggressive disease (high Gleason score), targeting such MRI-positive intraprostatic regions with high radiation doses may increase local control for the group of prostate cancer patients referred to primary RT. MR-targeted radiation dose escalation was also the topic of Zilli et al.'s paper on recurrent prostate cancer [15]. Patients underwent multiparametric MRI (including DWI, DCE-MRI and spectroscopic imaging) with an endorectal coil (erMRI). In contrast to the study by Borren et al. [14], where quantitative criteria were used to identify intraprostatic lesions, this study used radiological criteria to define relapsing lesions. Of a total of 171 patients, 131 had local relapse identified by erMRI and received 64 Gy to the prostatic bed and additionally 10 Gy boost to the recurrent tumor. For the remaining 40 patients, 64 Gy was given to the prostatic bed only. Comparing the treatment groups, the frequency of genitourinary toxicity was positively associated with tumor boost, albeit not with a high significance (p = 0.06 and 0.048 in univariate and multivariate analysis, respectively). This shows that MR-guided boosting may increase normal tissue toxicity, but this must be weighed against the likelihood of salvage. The authors argue that such a dose-adapted protocol (boost vs. no boost) may be useful for patient groups with large variations in risk profiles.

Dose escalation was also the topic of Nielsen et al.'s RT planning study on non-small cell lung cancer (NSCLC) [16]. Here, the boost region was basically defined as the GTV, where the boost dose (allowed to be inhomogeneous) was individually determined for each patient. This was done by maximizing a mathematical function, describing the tumor control probability (TCP), during intensity-modulated radiation therapy (IMRT) optimization. The boost plans were compared to standard IMRT plans where 66 Gy was delivered to the planning target volume in 2 Gy fractions. The mean tumor dose could typically be increased by 8–9 Gy in total comparing the boost with the no-boost treatment plans, while the dose differences for normal tissues were much smaller. Thus, the study shows that individualized dose escalation of patients with NSCLC is feasible and may increase the local TCP, without significantly elevating the risk of normal tissue toxicity. However, if heavy ion beams of, e.g. carbon are available, there is another possibility for increasing the likelihood of local control: exploit that heavy ions have a high energy deposition density (quantified by the linear energy transfer; LET) at the end of their tracks in tissue. Just as radiation dose, LET may be estimated in each voxel of the patient images and spatial LET distributions may thus be visualized and potentially modulated. As a high LET yields a high biological effectiveness and reduced oxygen effect, it may be beneficial to distribute the LET in the tumor in such a way that the hypoxic (and thus radioresistant) tissue is traversed by ions of maximal LET; “LET-painting” [17]. This concept has received further attention recently [18–20], and in the current study by Bassler et al. it was further tested in silico [21]. Hypoxia images using 18F-fluoroazomycin arabinoside (FAZA) of a patient with head and neck cancer was employed in carbon and oxygen ion therapy planning. Forward treatment planning with ramp-like dose gradients were used to achieve high LET values in the hypoxic region, and the feasibility of LET-painting was well demonstrated. Supplemented with TCP modeling for various scenarios, the present work shows that LET-painting may become useful, in particular for patients with small hypoxic foci treated with oxygen ions. However, it is important to recognize limitations of current mathematical models used to estimate the treatment effect (both in tumors and normal tissues) following RT. This is well illustrated in the work by Sjostedt et al. [22], where cell survival following exposure to inhomogeneous radiation doses (cold spots) of human prostate adenocarcinoma cultured in vitro was studied. Briefly, the authors found that for a culture dish irradiated largely with 2 Gy, but with a local cold spot of 1.6 Gy, about the same level of cell survival for the cells given 2 Gy and 1.6 Gy (in the same dish) was obtained. Furthermore, this survival level was roughly the same as for a homogenous irradiation with 2 Gy, indicating that intercellular communication (“bystander” effects) in the inhomogeneously irradiated cell population play a role. This effect is not incorporated in standard mathematical models of cell inactivation, where clonogenic cells are considered to be non-communicating entities only responding to a given dose. The study emphasizes that the calculated clinical effect obtained from inhomogeneous radiation dose distributions, such as obtained from tumor boosting, may not be correctly estimated by established models.

In conclusion, the current issue of Acta Oncologica shows that advanced cancer therapy and medical imaging can be successfully integrated. Together with better understanding of both radiobiological effects and the biology reflected in medical images, this is expected to result in improved outcomes for future patients.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.