Crucial points in the use of FDG-PET for radiotherapy planning are patient positioning and image coregistration. As the quality of image fusion has been shown to be significantly better when CT and PET have been acquired in identical position, it is mandatory, that a FDG-PET acquisition for RT treatment planning must be done in RT treatment patient position (Coffey and Vaandering 2010). Here, photo documentation and laser systems are of great help. Irrespective of the kind of scanner used (PET or PET/CT), the correct coregistration of PET data with the CT dataset used for RT-planning must be verified. It must be kept in mind that patient position may change, not only between acquisitions on stand-alone PET and CT-scanners, but also during PET/CT acquisitions. Not correcting for the consequent differences in tumor localisation leads to a geographical miss, if PET-derived GTVs are transmitted to CT datasets without critical evaluation of the quality of coregistration. This is best done by comparing anatomical landmarks detectable by both imaging techniques like carina tracheae, lung apices, spine, sternum, thoracic wall and—with care due to breathing mobility—diaphragm (Grgic et al. 2008).

Non-rigid coregistration algorithms may solve a part of the positioning problems. However, although these methods have been proposed for the use in radiotherapy (Ireland et al. 2007) the resulting image datasets have not been evaluated concerning the correct position of pathologic tissues, especially tumors until now. It may well be, that the deformation of image data caused by non rigid algorithms may result in geometrical inexactnesses or geographical misses, predominantly in cases, where the tumours are not unequivocally depicted by the morphological method (Grgic et al. 2009). Unfortunately, the integration of FDG-PET into RT planning is especially needed for those cases. Therefore, to date the use of rigid coregistration algorithms after careful patient positioning in all modalities is the method of choice.

Determination of a volume from FDG-PET is a critical step. Various basic approaches were reported in the literature to accurately contour FDG-based GTVs (Erdi et al. 1997; Knight et al. 1996; Pötzsch et al. 2006). Easily applicable from a technical point of view is the visual contouring by the experienced physician, in analogy to the method used for CT-based contouring. However, a significant inter-observer variation (IOV) remains (Pötzsch et al. 2006). To improve this IOV, clinical protocols have been proposed (MacManus et al. 2007; Macmanus and Hicks 2008) and succeeded in a significant convergence of FDG-based GTVs contoured by different observers.

However, visual contouring remains observer dependent and by further distribution of the method into clinical practice, the varying experience of the radiation oncologist with PET will influence the quality of visual GTV-contouring. Therefore other methods for automatic and/or semi-automatic threshold contouring of the—often high contrast—FDG-accumulations have been evaluated.

Herefore, other than absolute or relative SUVs, contrast oriented algorithms appear quite robust under clinical conditions. However, the calibration to the PET and RT-planning system used is mandatory (Schaefer et al. 2008). As technical factors like the methods of reconstruction, attenuation correction of PET images and the change of data formats in transfer processes between different software systems do influence PET image data (Jaskowiak et al. 2005; Schoder et al. 2004; Tarantola et al. 2003), these methods and processes have to be defined before and kept constant after calibration.

Recently, gradient-based and statistical contouring methods have been proposed by some authors (Geets et al. 2007; Graves et al. 2007; Zaidi and El Naqa 2010). These approaches also appear promising, as they may not be susceptible to fluctuating maximum values within tumors and to varying background intensities. However, some of these methods require a high grade of user interaction and computing power and the robustness against intra-tumor inhomogeneities in clinical use, the usefulness for the contouring of low-contrast lesions and the comparison with other methods and gold standards has to be awaited.

Beyond all that, the preferable method for target volume delineation may depend from the tracer used due to widely differing contrast. Example shows the pathologic accumulations of brain tumors in amino acid PET less than twice the tracer uptake seen in normal brain tissue (Kracht et al. 2004), while with FDG, the tracer uptake of tumors may be 10–50-fold higher as compared to normal tissue. In low contrast situations, GTV contouring is often performed visually in cooperation of the radiation oncologist with the experienced nuclear medicine physician. For AA-Pet in glioma, some groups have published threshold values to be used for semi-automatic contouring before visual correction. Proposed values for this purpose are the 1.7-fold uptake compared to normal tissue (Grosu et al. 2005a, b), or the 2.0-fold uptake for intensely accumulating and 1.3-fold for moderately accumulating lesions compared to contralateral normal brain (Miwa et al. 2008). Other groups use thresholds related to the maximum uptake of the lesion, e.g., 75 % (Plotkin et al. 2006) for delineating glioma tissue in MET-PET.

Overall, it must be kept in mind, that the choice of the method for GTV contouring may have significant impact on the size of the GTV (Nestle et al. 2005) and that the most important factor in PET-based GTV delineation is the close collaboration of nuclear medicine and radiotherapy departments on the side of the medical as well as on the side of the physical and technical staff.

In percutaneous RT, due to the fixed gantry of the linear accelerators, the movement of an organ relative to the planning position is a parameter of interest (Langen and Jones 2001). In principle, three different categories of motion are of relevance. Organ motion related to patient position changes, motion that occurs in between fractions, and motion which occurs during fractions. Breathing movements of thoracic and abdominal tumors pose a serious problem both for imaging and for radiotherapy treatment application. The problems associated with motion include limitations of image acquisition, of treatment planning and of relevant dose delivery (Keall et al. 2006).

Usually the situation in diagnostic CT is the following: CT is performed in breath-hold condition and may then provide an image of the tumour representative of its instantaneous position and not of its different locations during the whole breathing cycle. Therefore, radiotherapy planning CTs are acquired in flat breathing in order to represent the situation on the treatment couch better. During the process of treatment planning, safety margins are added to the tumor volume determined in CT in order to adjust for respiratory motion. The ICRU recommends the CTV to be expanded to the “ITV” with an internal margin (IM) taking into account for tissue deformations and physiologic movements due to respiration, cardiac motion and peristalsis (ICRU 1993, 1999). For the definition of the IM different methodologies can be used: the application of population-based margins (Langen and Jones 2001); fluoroscopy (Ekberg et al. 1998; Shirato et al. 2004); multiple CT acquisitions (Ross et al. 1990); slow-CT scan (de Koste et al. 2003) or 4D-imaging (Nehmeh et al. 2004).

In contrast to routine CT, a routine PET scan is performed over a multitude of breathing cycles in free breathing condition. Therefore the PET study provides an image of the lesion representing the integral over the whole volume within which the lesion moves during patient respiration. The resulting image ideally shows an increase in lesion size representing the localisation of the tumor during the whole breathing cycle. Therefore, it has been proposed to use ungated PET data to define the ITV for lung tumors (Caldwell et al. 2003).

Beyond the accurate measurement of motion and consecutive definition of the ITV (Bettinardi et al. 2010), various approaches have been undertaken in order to minimize the ITV by influencing either the patients respiration or adapting the treatment application to the respiration cycle (Senan et al. 2004)

However, outcome data have to be awaited demonstrating a clearly reduced normal tissue toxicity or increased tumor control after gated radiotherapy application in order to justify the enourmous technical effort needed.

In recent years, numerous clinical studies and metanalyses have shown the superiority of FDG-PET over traditional imaging methods in lung cancer staging and its clinical consequences (Dwamena et al. 1999; Eschmann et al. 2007; Hellwig et al. 2007; MacManus et al. 2001; van Tinteren et al. 2002). The promising diagnostic performance of FDG-PET in non-small cell lung cancer (NSCLC) has led to the idea of integrating these data into RT-planning more than ten years ago (Table 1) (Munley et al. 1999; Nestle et al. 1999). The main motivation for these activities is the clearly superior diagnostic accuracy of FDG-PET compared to CT in the staging of NSCLC, which has meanwhile been accepted by the scientific community and has led to the standard use of FDG-PET with this indication in clinical routine.

However, the diagnostic accuracy of FDG-PET is not 100 %, but only between 85–90 %. This fact must be considered in clinical concepts dealing with the integration of PET-based target volumes in revised treatment algorithms, e.g., when discussing the need for the prophylactic irradiation of macroscopically unaffected lymph node regions.

To date (Nestle et al. 2002, 2006, 2009), over 20 studies in more than 600 patients have shown that the use of FDG-PET image data in radiotherapy planning may lead to an advantage for the patient. The main sources of this possible advantage are the better coverage of the primary tumor and the protection of healthy tissue. The FDG-based target volumes may be both smaller and larger compared to only CT-based ones. However, based on the surgical data regarding the specificity of FDG-PET for tumor staging it can be assumed, that FDG-based target volumes do include the tumor tissue more accurately than merely CT-based volumes. The high percentage of changes in target volumes due to FDG-PET reported in the literature (20–100 %) concerning various parameters of the planning process (field sizes, GTV, CTV, PTV, NTCP etc.) is mainly caused by two factors: the ability to distinguish the tumor from collapsed lung tissue (atelectasis, Fig. 3), and the higher accuracy of FDG-PET in lymph node staging compared to CT.

The highest possible benefit for the patients from FDG-based RT planning can only be gained, if, due to the exact depiction of tumor localisation by PET, the irradiation of large parts of normal tissues can be omitted. In lung cancer, this supports the concept of selective rather than elective irradiation of mediastinal lymph nodes when defining the CTV (Kiricuta 2001). This could lead to a significant protection of highly radiosensitive normal tissues, e.g., lung, with the consequence of obtaining higher irradiation doses in the tumor. First clinical data with (De Ruysscher et al. 2005) and even without FDG-PET (Rosenzweig et al. 2007), have shown that the risk of “out field”-recurrences after targeting the macroscopic tumor alone is small, much smaller than the risk for local (“in field”) tumor progression.

Regarding the literature data on cases with information on PET-, CT-findings and histology available, one may calculate the residual risk when following one or the other strategy. In a patient cohort published by Vanuytsel et al. (2000) 988 lymph node stations were evaluated by PET, CT and histology (Fig. 4). It can be calculated, that a targeting strategy in this patient group including CT-positive lymph node stations into the treatment fields only would lead to the inclusion of 96 lymph node stations while half of the stations (47 affected stations) would not be treated properly. Using the PET signal for targeting, 84 stations would be treated, while 25 histologically positive stations would not receive adequate treatment, leading to a residual risk for out field recurrences of 2.5 %. Adding both strategies would significantly increase the number of irradiated stations (n = 129; i.e., >50 % more than treating by PET alone) while reducing this residual risk only by 0.5 %. In the end, it is the question of the clinical situation, which will lead to the acceptance or rejection of risks. As locally advanced lung cancers unfortunately by radiochemotherapy can only be locally controlled to an extent of about 30–50 %, current strategies accept these risks in order to dose escalate to the tumor with the chance of a higher rate of tumor control.

However, prospective randomized clinical studies will have to show, that this policy is safe and beneficial for the patients.

In several studies the sensitivity and specificity of FDG-PET for the detection of lymph node metastases, unknown primary tumor or tumor recurrence after previous treatment in the head and neck region was higher in comparison to MRI and CT. Gambhir et al. (2001) summarized the data of eight studies (468 patients), which evaluated the impact of FDG-PET in staging of head and neck cancer. The average sensitivity and specificity for FDG-PET were 87 and 89 %, whereas for CT 62 and 73 %, respectively. For tumor diagnosis the sensitivity and specificity of FDG-PET, assessed in 7 trials incorporating 193 patient studies, were 93 and 70 %, in comparison to CT with 66 and 56 %, respectively.