Planning organ at risk volume margins for organ motion of the intestine

Materials and methods The present study was based on intestine contours outlined in a total of 149 CT scans of 20 male bladder cancer patients (20 planning scans, 129 during treatment). From these data, we created location probability maps of the intestine for each patient. A commercial treatment planning system was used to add 3D isotropic intestine PRV margins (from 5 to 30 mm, in intervals of 5 mm) around the intestine planning outline. We then derived the fraction of patients for which a given PRV encompassed various degrees of intestine motion (85%, 90% and 95% of volumes with different probabilities of intestinal occupancy). As a measure of the specificity of the PRV, we also derived the fraction of the PRV containing volumes with zero probability of intestinal occupancy. Results Isotropic margins of up to 30 mm are required to account for all intestine motion in 90% of the patients, while isotropic margins of 5–10 mm will encompass 85–95% of the volumes having a probability of intestinal occupancy of ⩾75% in the same fraction of patients. Intestine PRVs are not very specific and will also include volumes where the intestine will rarely or never be located. Conclusions Large intestinal motion was found, but isotropic PRV margins of 5–10 mm will include the major part of volumes with a large probability of intestinal occupancy in most patients. Keywords Pelvic radiotherapy Pelvic tumours RT Bowel Intestine Planning organ at risk volume (PRV) The small intestine is a radiosensitive organ, which is often the dose-limiting organ at risk (OR) in radiotherapy (RT) of pelvic tumour sites [10] . Decreasing the volumes of small intestine receiving high doses is therefore an important objective in pelvic RT [4,10,11,15] . However, the knowledge about the correlation between dose-volume parameters of the intestine and the risk of intestine adverse effects is less clear, probably due to the mobility of this organ [3,10] . Nuyttens and colleagues have reported considerable variation in the position of the small intestine both in the cranial–caudal and the anterior–posterior direction in patients with rectal cancer treated with preoperative and postoperative RT [21] . Our group has also documented large inter- and intra-patient variation in intestine position and volume when evaluated using weekly repeat CT scans of 20 bladder cancer patients [19] . Recently, Kvinnsland and Muren investigated the impact of pelvic organ motion on dose-volume histograms (DVH) of the intestine in ten of these bladder cancer patients [9] . In six patients, the volume occupied by the intestine in one scan only was larger than the volume occupied by the intestine in all scans. This mobility of the intestine was also reflected in large standard deviations in the DVHs for the individual patients. To account for internal organ motion and set-up uncertainties around ORs, the ICRU report no 62 introduced the planning organ at risk volume (PRV) [7] . The PRV includes the OR as well as safety margins around the OR, in analogy with the target volume definitions where a planning target volume (PTV) is used around the clinical target volume (CTV) to account for geometrical variation. In 2002, McKenzie and co-workers presented a method for determination of margins around ORs to account for both systematic (treatment preparation) and random (organ motion and set-up errors) geometrical uncertainties [13] . The PRV was defined such that the DVH of the PRV should not underestimate the high-dose components delivered to the OR, in 90% of cases. In our group, an purely empirical approach based on repeat CT scans of bladder cancer patients was applied, together with the McKenzie methodology to determine rectum PRV margins, and some limitations with the McKenzie et al., methodology when using it on complex organs such as the rectum were disclosed [16] . Recently, our group also investigated whether use of DVHs of various rectum PRVs (with different margins added) improved the correlation between DVH parameters and acute gastro-intestinal (GI) toxicity in a series of 132 prostate cancer patients [17] . Although a correlation was found for both the rectum only DVH [8,17] and the PRV DVHs [17] , 2–3 times (depending on margin size) as many dose levels were significantly related with toxicity for the PRVs compared to the rectum only [8,17] . In addition to potentially improving the predictive power of dose-volume statistics, PRVs may also be useful as a tool in planning of RT, in particular in inverse planning of intensity-modulated RT (IMRT). Currently, we are using IMRT to treat pelvic lymph nodes in locally advanced prostate cancer. In the present study, we therefore quantified PRV margins for the intestine as an OR in pelvic radiotherapy. Materials and methods Patient material The present study was performed using 20 male patients with muscle invading transitional cell urinary bladder cancer. Out of these, 14 were referred for radical conformal radiotherapy at Haukeland University Hospital (HUH) in the period from January 2000 to October 2001, while the other 6 were treated at Edinburgh Cancer Centre (ECC) during 2003. For the whole group, age ranged between 58 and 87 years (mean age: 74 years). For the patients treated at HUH, weekly repeat computer tomography (CT) scans were acquired as close to the treatment session as practically possible. Patients were instructed to empty their bladder before all treatment and scanning sessions, and 70 ml of contrast was instilled into the bladder before the planning scan only. For the six patients from ECC, CT scans were acquired twice a week during the four-week treatment course, again as close to treatment session as practically possible. These six patients were scanned with an empty bladder, without instillation of contrast. Repeat CT data and intestine outlining Overall, 6–9 CT scans (totally 149 scans) were acquired for each of the patients. For the patients treated at HUH, the CT scans covered both the abdomen and pelvis, while the CT scans of the patients from ECC covered the pelvis up to the sacral promontory. All patients were scanned in supine position: The patients from HUH as well as one of the patients from ECC (i.e., 15 patients) were scanned with 5 mm slice thickness and 5 mm interval (5/5 slices) throughout the pelvis, while the five remaining patients from ECC were scanned with 3/3 slices. The repeat scans were registered to the planning scan using the Advantage Fusion software (v. 1.15; GE Medical Systems, Milwaukee, WI, USA) – initially by using an automatic procedure that primarily matched on bony anatomy, and further by defining six bony landmarks in the abdomen and pelvis. One of the authors (LPM) outlined all segments of the intestine located below the sacral promontory for the patients treated at HUH. For the patients treated at ECC, one of the authors (HL) outlined the intestine in the repeat scans, while another author (LBH) did the intestine outlining in the planning scans. The same instruction for outlining was followed for all cases, with each individual loop of the intestine (both small and large) being outlined. Only the part of the intestine located below the sacral promontory was included in the analysis. The organ outlines in the repeat scans were automatically transferred to the planning scan using the 3D image registration transform in the Advantage Fusion software and saved as separate DICOM RT Structure sets (DICOM RTSSs). These RTSSs were transferred to a PC, where further analysis was performed using specially designed software written in Interactive Data Language (IDL, Research Systems, Inc., Boulder, USA). Defining the intestine location probability map A 3D location probability (LP) map for the intestine was created for each patient, using eight intestine outlines for each patient. The maps had the same pixel resolution as the CT matrix, containing voxels with a value from 1/8 to 8/8, corresponding to how many times the voxel had been occupied by the intestine (i.e., as observed in the repeat scans). For patients with fewer than 8 repeat scans, a few randomly selected intestine outlines were used twice to create the location probability matrix. The probability of intestinal occupancy for each voxel in the location probability map was equal to the voxel value, and a location probability volume of ⩾LP% (with LP being 12.5–100% in intervals of 12.5%) was defined as the volume of the voxels with a value greater or equal to the LP. The planning outline of the intestine was not included in the location probability map. Calculation of volumes and volume comparison The volume of the intestine was calculated by multiplying the number of pixels within the contours in each slice of the CT scan with the pixel size and the slice thickness of the CT scan of each individual patient. The same definition was applied for the location probability maps. Since the patients treated at HUH had a small amount (70 ml) of contrast instilled into the bladder before the planning scan, it was investigated whether the volume of the intestine from the planning CT scan differed from the volumes of the intestine from repeat CT scans. This analysis was performed separately for the patients from HUH and ECC, since the latter had no contrast instilled during the planning CT session. We also compared the ratio between the location probability volumes and the average intestine volume of each patient. Analysis of intestine PRVs The PRV concept was used to quantify various degrees of intestine motion relative to the intestine planning outline in each patient. Using the 3D margin tool of the Eclipse treatment planning system (v. 6.5, Varian Medical Systems, Palo Alto, USA), we created intestine PRVs, by adding isotropic margins in the range of 5–30 mm (in intervals of 5 mm) to the planning outline of the intestine in all directions (superior/inferior/left/right/anterior/posterior). Also two combinations of anisotropic margins of 5 mm left/right/anterior/posterior and 10 mm superior/inferior as well as 10 mm left/right/anterior/posterior and 15 mm superior/inferior were applied to the intestine planning outline. All PRVs were finally transferred to a PC in the form of DICOM RTSSs, and further analysed using the in-house made IDL software. For each PRV for each patient, a matrix with the same dimensions as the location probability map was created, with each voxel either defined as being inside the PRV (assigned a value of 1), or outside the PRV (assigned a value of 0). By using the intestine location probability map of each patient and the corresponding PRV matrices, we found the fraction of patients of which a given PRV encompassed 85%, 90% and 95% of the different location probability volumes. To investigate the specificity of the PRV, we also derived the average volume fraction of the PRVs that had no probability of being occupied by the intestine (LP = 0%). For this analysis, we only included the part of the PRV that was located inside the patient outline. Statistics The SPSS software (v. 13.0, SPSS Inc., Chicago, Illinois, USA) was used for a two-way analysis of variance when comparing the volume of the intestine planning outline with repeat outlines. Results An example of an individual probability distribution of where the intestine was located – the basis of our empirical analysis – is shown in Fig. 1 . Volume analysis The individual intestine volume of the planning scan was not significantly different from the intestine volumes in the repeat scans, neither for the patients from HUH ( p = 0.7), which had bladder contrast instilled before the planning scan, nor for the patients from ECC ( p = 0.3). The overall volume visited by the intestine in the repeat scans (LP ⩾ 12.5%) was on average twice as large as the individual average intestine volume ( Table 1 ), which varied between 141 and 569 cm 3 . The volume with LP ⩾ 50% was about the size of the individual average intestine volumes and volumes with a high location probability (LP ⩾ 75%) were smaller than this average ( Table 1 ). Quantification of PRVs to account for various degrees of intestine motion The fraction of patients where various degrees of intestine motion (in fractions of 85%, 90% and 95% of different LP volumes) were within a certain margin level is shown in Fig. 2 , each diagram representing a certain LP level. Looking at 90% of the patients and three out of these location probability levels (i.e., LP ⩾ 12.5%, LP ⩾ 50%, and LP ⩾ 75%) we found the margins presented in Table 2 . To encompass 85–95% of all intestine motion in 90% of the patients, an isotropic PRV of 15–25 mm was required. However, if we instead seek to encompass 85–95% of the volume with high location probability (LP ⩾ 75%) in 90% of the patients, an isotropic PRV of 5–10 mm was sufficient. The fraction of the intestine planning outline that had no probability of being occupied by the intestine (LP = 0%) in the repeat scans varied from 1% to 32%, with an average of 8% ( Table 3 ). This average volume fraction increased for the PRV along with the size of the margins applied, from 20% for the isotropic 5 mm PRV margin to 50% for the isotropic 15 mm margin. Discussion In the present study, we have analysed the ICRU planning organ at risk volume (PRV) concept for the intestine [7] . Our empirical approach was based upon intestine location probability maps for each individual patient derived from outlines of the intestine as observed in repeat CT scans. As reported in previous studies [9,19,21] , we found a considerable geometrical variation of the intestine in each patient as well as between patients. The overall volume visited at least once by the intestine was on average twice as large as the individual average intestine volume ( Table 1 ). Isotropic margins of up to 30 mm were required to encompass all intestine motion in 90% of the patients ( Table 2 ). However, in volumes where the intestine has a high location probability (LP ⩾ 75%), isotropic margins from 5 to 10 mm will cover the major part of the volume in most patients. For instance, when looking at 95% of the volume with location probability ⩾ 75% ( Fig. 2 ), the anisotropic 5/10 mm margin increases the fraction of patients enclosed by the PRV from 0% (with no margin) to 80%. If 90% of the patients are to be included, margins of 10 mm are needed. The selection of which intestine PRV margin to use depends on what degree of intestine motion the PRV should account for, and in what fraction of patients. Another key factor when determining the required size of margins to account for intestine motion is the specificity of the PRVs. Our analysis showed that intestine PRVs will also include volumes where the intestine will rarely be located (i.e., volumes with no intestine occupancy). On average, 50% of the 15 mm PRV does not include any intestine at all from repeat scans ( Table 3 ). The PRV specificity decreases along with the margin size and the selection of a margin size therefore represents a trade-off between the degree of intestine motion encompassed by the PRV and the acceptable lack of specificity of the PRV. The low specificity of intestine PRVs also implies that the use of margins around the intestine is not an ideal method to accounting for intestine motion. As mentioned earlier by Muren et al., a probability distribution approach would probably be more suitable [19] . Margins around organs at risk may serve two purposes – first, they may improve the correlation between DVH parameters and adverse effect risks, and second, they may be a useful tool in IMRT optimisation to better avoid critical organs [13,17] . PRVs are recommended in the DAHANCA IMRT guidelines for head and neck cancer, where a slightly higher max dose is accepted in the PRVs (for the spinal cord, brain stem, etc.) than in the OR itself [1,2,12] . When used for IMRT planning the PRVs could be adjusted to not go beyond anatomical boundaries (i.e., pelvic bones, muscles) and possibly also exclude the target volumes (by using logical functions and/or manual adjustments). A drawback of the PRVs when used for IMRT planning is that all elements of the PRV will have the same importance in the optimisation, independent of the likelihood of observing the intestine there. There is also the risk that if the PRVs are too large, this would reduce the degrees of freedom in the IMRT optimisation process. Factors such as bladder filling [5,18,20] , prior pelvic surgery [4] and gender are believed to have an impact on intestine location and mobility, and the findings from the present study may therefore not necessarily be transferred to other groups of patients. Furthermore, the intestine outlines used in the present study included both large and small intestine segments, and we could therefore not provide any quantitative information about the degree of motion in the different parts of the intestine. It would be interesting to investigate the motion of the different intestine segments and also assess whether there is a correlation between intestine segments that are believed to be less mobile (such as the lower ileum and the lower sigmoid colon) and volumes with high location probability. These segments are the most important to avoid in RT, since adverse effects often occur in these parts of the intestine [14] . The location probability distribution of each patient used in the intestine margin analysis of the present study was based upon repeat CT scans acquired once or twice a week. They therefore gave no indication of the short-time intestine variation. This could however be investigated using MRI and/or high frequency ultrasound techniques. Although the ICRU have recommended the use of PRVs [7] , it is at present unclear how to deal with this concept [13,16] (i.e., what kind of ORs’ PRVs should be applied for and what degree of the geometrical uncertainties should the margins around ORs account for). This will be addressed in an up coming ICRU report [6] . In the present study, we have provided the data required to selecting intestine PRVs accounting for various degrees of internal intestine motion in male patients. Currently, also planning studies of the effect of using intestine PRVs when planning pelvic IMRT have been initiated. We are also considering to investigate the usefulness of intestine PRVs for adverse effect prediction in a prostate and bladder material from Aarhus with CT-based treatment plans and intestine follow-up data [3] . Such studies will provide useful information about applying PRV margins for the intestine. Acknowledgements The authors thank Anthony Thomas Redpath, Andrew Reilly, Duncan McLaren, Sara Erridge and Aileen MacLeod (all from Edinburgh Cancer Centre) for their contribution to the manuscript. References [1] S.L. Breen T. Craig A. Bayley Spinal cord planning risk volumes for intensity-modulated radiation therapy of head-and-neck cancer Int J Radiat Oncol Biol Phys 64 2006 321 325 [2] DAHANCA. The danish head and neck cancer study group (DAHANCA) guidelines: www.dahanca.dk , 2006. [3] L. Fokdal H. Honore M. Hoyer H. von der Maase Dose-volume histograms associated to long-term colorectal functions in patients receiving pelvic radiotherapy Radiother Oncol 74 2005 203 210 [4] M.J. Gallagher H.D. Brereton R.A. Rostock A prospective study of treatment techniques to minimize the volume of pelvic small bowel with reduction of acute and late effects associated with pelvic irradiation Int J Radiat Oncol Biol Phys 12 1986 1565 1573 [5] N. Gerstner S. Wachter T.H. Knocke The benefit of Beam’s eye view based 3D treatment planning for cervical cancer Radiother Oncol 51 1999 71 78 [6] Grégoire V, Mackie T. ICRU committee on volume and dose specification for prescribing, recording and reporting special techniques in external photon beam therapy: conformal and IMRT. 8th biennal ESTRO meeting on physics and radiation technology for clinical radiotherapy, Lisbon, Vol. 76, Supplement 2: Radiother Oncol, 2005: s71. [7] ICRU. Prescribing, recording and reporting photon beam therapy (Report 62, supplement to report 50): Internation Commission on Radiation Units and measurements, 1999. [8] A. Karlsdottir D.C. Johannessen L.P. Muren T. Wentzel-Larsen O. Dahl Acute morbidity related to treatment volume during 3D-conformal radiation therapy for prostate cancer Radiother Oncol 71 2004 43 53 [9] Y. Kvinnsland L.P. Muren The impact of organ motion on intestine doses and complication probabilities in radiotherapy of bladder cancer Radiother Oncol 76 2005 43 47 [10] J.G. Letschert The prevention of radiation-induced small bowel complications Eur J Cancer 31A 1995 1361 1365 [11] J.G. Letschert J.V. Lebesque B.M. Aleman The volume effect in radiation-related late small bowel complications: results of a clinical study of the EORTC Radiotherapy Cooperative Group in patients treated for rectal carcinoma Radiother Oncol 32 1994 116 123 [12] M.A. Manning Q. Wu R.M. Cardinale The effect of setup uncertainty on normal tissue sparing with IMRT for head-and-neck cancer Int J Radiat Oncol Biol Phys 51 2001 1400 1409 [13] A. McKenzie M. van Herk B. Mijnheer Margins for geometric uncertainty around organs at risk in radiotherapy Radiother Oncol 62 2002 299 307 [14] K. Meissner Late radiogenic small bowel damage: guidelines for the general surgeon Dig Surg 16 1999 169 174 [15] B.D. Minsky J.A. Conti Y. Huang K. Knopf Relationship of acute gastrointestinal toxicity and the volume of irradiated small bowel in patients receiving combined modality therapy for rectal cancer J Clin Oncol 13 1995 1409 1416 [16] L.P. Muren R. Ekerold Y. Kvinnsland A. Karlsdottir O. Dahl On the use of margins for geometrical uncertainties around the rectum in radiotherapy planning Radiother Oncol 70 2004 11 19 [17] L.P. Muren A. Karlsdottir Y. Kvinnsland T. Wentzel-Larsen O. Dahl Testing the new ICRU 62 ‘Planning Organ at Risk Volume’ concept for the rectum Radiother Oncol 75 2005 293 302 [18] L.P. Muren R. Smaaland O. Dahl Conformal radiotherapy of urinary bladder cancer Radiother Oncol 73 2004 387 398 [19] L.P. Muren R. Smaaland O. Dahl Organ motion, set-up variation and treatment margins in radical radiotherapy of urinary bladder cancer Radiother Oncol 69 2003 291 304 [20] J.J. Nuyttens J.M. Robertson D. Yan A. Martinez The influence of small bowel motion on both a conventional three-field and intensity modulated radiation therapy (IMRT) for rectal cancer Cancer Radiother 8 2004 297 304 [21] J.J. Nuyttens J.M. Robertson D. Yan A. Martinez The position and volume of the small bowel during adjuvant radiation therapy for rectal cancer Int J Radiat Oncol Biol Phys 51 2001 1271 1280