IGRT Solution for the Living Patient and the Dynamic Treatment Problem Lei Dong, Ph.D. Associate Professor Dept. of Radiation Physics University of Texas M. D. Anderson Cancer Center Houston, Texas Learning Objectives: Identify common site-specific anatomic changes during the course of external beam radiotherapy using conventional fractionation schemes. Illustrate dosimetric consequences due to changed anatomy or the positions of the targets. Suggest correction strategies based on observed patterns of site-specific anatomic variations. Conventional Workflow An implied assumption is that the shapes, volumes, and positions of an anatomic structure of interest will not change throughout entire radiotherapy treatment. Even if these changes exist, wide margins can be used to correct for them (ICRU #50 & #62). But, this may not be enough: IMRT is more sensitive to organ variations Further treatment optimization requires more careful considerations of site-specific or patient-specific anatomic variations. 1
Outline Review: types of uncertainties Site-specific issues in prostate Site-specific issues in head & neck Site-specific issues in thorax or abdomen Image-guided adaptive radiotherapy Why do we need to worry? The dose response curves are quite steep. There is a clinical evidence that a small change (7% to 10%) in dose can result in a change in tumor control probability (ICRU Report #24, 1976). With highly conformal radiotherapy (such as IMRT), the probability of geographic miss increases significantly. How to quantify inter- and intra-fraction organ motion or variations Systematic Group statistics Random Day-to-day variations Shape Variations Volume Deformation Trends Time-dependent variations 2
Random and Systematic Uncertainties Individual Patient Statistics σ i ( x x ) x i = ref / Ni 2 ( ( x xref ) / N ) = i Group Statistics μ = x i / N Σ = 2 ( ( x i μ) / N ) 2 σ rms = ( ( σ i ) / N ) Margin Considerations Marcel van Herk Errors and Margins in Radiation Therapy. Seminars in Radiation Oncology, Vol 14, No.1, 2004 Di Yan et al. CT-guided Management of Inter-fractional Patient Variation. Seminars in Radiation Oncology Large Σ and small σ: more beneficial for off-line correction Large σ: more beneficial for on-line correction How big is the population margin? Margin = 2.5* + 0.7 *σ (mm) σ (mm) Margins (mm) ROIs AP SI RL AP SI RL AP SI RL C2 2.3 1.6 2.6 1.7 1.8 1.5 6.9 5.3 7.6 C6 2.5 2.0 3.2 2.0 2.2 2.3 7.7 6.5 9.6 PPM 2.4 4.2 1.5 1.7 2.9 1.1 7.2 12.5 4.5 3
Population-derived margin will not work well (too conservative) The danger of systematic errors Consistently wrong Example de Crevoisier R, Tucker SL, Dong L, Mohan R, Cheung R, Cox JD, Kuban DA. Increased risk of biochemical and local failure in patients with distended rectum on the planning CT for prostate cancer radiotherapy. Int J Radiat Oncol Biol Phys 2005; 62 (4):965-973. Effect of Rectal Distension Hypothesis: If prostate motion is not taken into account, rectal distension on the planning CT may result in a systematic target miss and therefore an increased risk of biochemical failure. 4
Measures of Rectal Distension: Cross-Section Area (CSA) = rectal volume / length Study Population Distribution of Risk Groups Low Risk: N=26 (21%) Intermediate Risk: N=60 (47%) High Risk: N=41 (32%) > 2-yr follow-up (median 7 yrs; up to 10 yrs) No enema Setup with skin marks+weekly PF Biochemical Control by Risk Group: 5
Biochemical Control by Rectal Distension: Biochemical Control by Rectal Distension For Intermediate Risk Patients: (large rectum) Biochemical Control by Rectal Distension For High Risk Patients: 6
Multivariate Analysis: Risk Factor Hazard Ratio p-value High Risk Disease 2.28 0.028 CSA > 11.2 cm 2 3.12 0.018 Rectal Diameter > 6cm 2.44 0.043 No correlation between rectal distension (CSA) and: Risk group PSA T-stage Gleason score Conclusion of this study There is clinical evidence that rectal distension on the treatment planning CT scan decreased the probabilities of biochemical control, local control and rectal toxicity in patients who are treated without daily image-guided prostate localization, presumably due to geographic misses. Therefore, an empty rectum is warranted at the time of simulation. 7
Int. J. Radiation Oncology Biol. Phys., Vol. 67, No. 5, pp. 1418 1424, 2007 N=549 78 Gy vs. 68Gy Tumor control was significantly decreased in a subgroup of patients with a large rectum filling visible on the planning CT scan and who had an estimated risk of SV involvement of 25%. Example of setup error and organ variation during the course of prostate radiotherapy Contours from treatment planning CT are overlaid as reference Patient aligned with BBs Prostate Position Relative to Skin Marks 15 10 AP movement (mean+/-1sd) Positions (mm) 5 0-5 -10-15 1 2 3 4 5 6 7 8 9 1011121314151617181920212223242526 Patients Systematic 1SD = 3.7 mm Random 1SD = 3.3 mm 8
Comparison of Bony and Direct Target Localization Bony Registration Direct Target Localization freezing the prostate! % Rectal Wall Volume Dosimetric Impact of Varying Anatomy (Pat07, Average Rectal Wall DVH +/- 1SD) 100% 90% 80% 70% 60% 50% 40% 30% 20% Average 10% Planning 0% 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 Dose (cgy) Error bar (+/- 1SD) was calculated based on 24 serial CT images. Ratio Relative to Simulation CT Time-Trend for Bladder Volume 90% of daily bladder volumes are smaller than the volume at the simulation! 2.6 2.4 2.2 2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 0 7 14 21 28 35 42 49 56 63 Elapsed Days pat01 pat02 pat03 pat04 pat05 pat06 pat07 pat08 pat09 pat10 pat11 pat12 pat13 pat14 pat15 9
% Bladder Volume 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% Dosimetric Impact of Varying Anatomy (Pat07, Average Bladder DVH +/- 1SD) Average Planning 0% 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 Dose (cgy) Error bar (+/- 1SD) was calculated based on 24 serial CT images. Can I use 0-margin for PTV when I use IGRT? Anatomic and Dosimetric Analysis of Intra-fractional Motion during an IMRT Treatment Fraction A Melancon*, R de Crevoisier, L Zhang, J O'Daniel, D Kuban, R Cheung, A Lee, R Mohan, L Dong, Univ. of Texas M. D. Anderson Cancer Center, Houston, TX 10
Before Treatment After Treatment (20 minutes) Contours overlaid from before treatment CT images after bony registration Before Treatment After Treatment (20 minutes) Volume Change of the Rectum for 45 patients Volume Change of Rectum in cc 80 60 40 20 0 Gaseous Build-up -20 1 4 7 10 13 16 19 22 25 28 Patient # 31 34 37 40 43 Effect of Bladder Filling to Prostate Positions MICHAEL PINKAWA et al. from Department of Radiotherapy, Rheinisch-Westfälische Technische Hochschule Aachen, Aachen, Germany. Int. J. Radiat. Oncol. Biol. Phys., Vol. 64, No. 3, pp. 856 861, 2006 30 patients CT scanned with full and empty bladder at the beginning, middle, and the end of RT. Prostate position is barely affected by the bladder filling. Rectum volume has the major influence on prostate position. Support to treat patients in supine position with full bladder. 11
Summary for prostate patients Systematic uncertainties are similar or greater than random uncertainties Off-line correction protocol may be beneficial. Daily IGRT would be ideal. The planning CT may not be a typical CT during treatment Large uncertainties in DVHs for both rectum and bladder The volumetric change of prostate gland during the course of radiotherapy is small and may not have significant clinical impact. Setup Uncertainties In Head & Neck Treatment Elapsed Treatment Days 19 Treatment CT scans acquired during the course of head & neck radiotherapy Planning First Fraction Last Fraction An example of increasing room inside a thermoplastic facemask due to tumor shrinkage as treatment progressing. Near the end of treatment, the lower neck was not centered on the headrest, presumably due to patient s self-adjustment to the relatively roomier mask. 12
Planning Planning Daily Daily Figure 8. Examples of anatomy rotation between the planning CT (first row) and the daily CT (second row). The axial CT images shown on the left indicate a roll in the patient s head; the coronal CT images on the right show yaw (rotation of the spinal column). Change in Neck Curvature Planning CT Daily Cone-beam CT with planning contour overlay 2D planar x-rays = volume alignment? Traditional alignment assumes rigid-body for the entire H&N region Can you combine multiple 2D ROI for a 3D ROI? 2 x 2D = 3D? 13
PPM C2 C6 L. Zhang et al. "Multiple regions-of-interest analysis of setup uncertainties for head and neck cancer radiotherapy," Int. J. Radiat. Oncol. Biol. Phys. 64 (5), 1159-69 (2006). 1.50 Correlation in AP direction Correlation on absolute setup shifts between C2 and C6 and between C2 and PPM C6/PPM Shifts (cm) C6/PPM Shifts (cm) C6/PPM Shifts (cm) C2C6AP C2PPMAP C2C6SI C2PPMSI 1.00 0.50 0.00-0.50-1.00-1.50-1.50-1.00-0.50 0.00 0.50 1.00 1.50 1.50 Correlation in SI direction 1.00 0.50 0.00-0.50-1.00-1.50-1.50-1.00-0.50 0.00 0.50 1.00 1.50 1.50 Correlation in RL direction 1.00 0.50 0.00-0.50 C2C6RL C2PPMRL -1.00-1.50-1.50-1.00-0.50 0.00 0.50 1.00 1.50 C2 Shifts (cm) PPM did not match after C2 alignment Daily Treatment CT Planning CT 14
C6 did not match after C2 alignment Daily Treatment CT Planning CT Evidence of Anatomic Variations During Treatment Course Planning CT Three Weeks into RT Barker et al. Int J Radiat Oncol Biol Phys 2004;59:960-970. Significant Anatomic Variations Planning CT During Treatment RTOG H0022 Case CTV is in the air! 15
140 Primary Tumor Response To RT Volume (cc) 120 100 80 60 40 20 HN01 HN02 HN03 HN04 HN05 HN06 HN07 HN08 HN09 HN10 HN13 HN14 HN15 0 0 7 14 21 28 35 42 49 Elapsed Treatment Days Barker et al. Int J Radiat Oncol Biol Phys 2004; 59 (4):960-970. Displacement (mm) 10 8 6 4 2 0-2 -4-6 -8 Center of Mass Parotid Shifts Relative To C2 Bony Reference "+" Lateral Shifts "-" Medial Shifts -10 0 7 14 21 28 35 42 49 Elapsed Treatment Days Median Line Barker et al. Int J Radiat Oncol Biol Phys 2004; 59 (4):960-970. Summary for H&N patients Setup uncertainties were found larger than traditionally expected with portal films. A collection of bones is not a rigid body. Need to worry about residual shifts at different parts of the H&N anatomy. Immobilization technique is important. Systematic uncertainties appear to be larger than random setup uncertainties. Offline correction strategy is beneficial. Trends in volumetric and positional variations Adaptive radiotherapy strategy 16
Squamous Cell Ca NOS T4 N2 M0 Planning CT One Month Into Treatment 4D CT Imaging Coronal Sagittal Non-ITV Approach (IMRT-8mm) 10 Gy 20 Gy 35 Gy 50 Gy 70 Gy 17
Planned vs. Delivered (IMRT-8mm) Planned 10 Gy 20 Gy 35 Gy 50 Gy 70 Gy After Accounting for Motion Effect Non-gated Approach (IMRT-ITV) Planned 10 Gy 20 Gy 35 Gy 50 Gy 70 Gy Non-gated Approach (IMRT-ITV) Planned 10 Gy 20 Gy 35 Gy 50 Gy 70 Gy Delivered 18
Impact of Organ Motion To Proton Dose Distributions Free breathing Treatment Proton vs. IMRT (motion effect to dose distribution) (Proton-ITV) (IMRT-ITV) ITV) 10 Gy 20 Gy 35 Gy 50 Gy 70 Gy 10 Gy 20 Gy 35 Gy 50 Gy 70 Gy Dong/MDACC Impact of Organ Motion on Proton Dose Distributions Prescription Dose Line Treatment planned based on single Free-breathing (FB) CT image (conventional approach) The same treatment plan calculated on 4D CT images Y. Kang et al. IJROBP. 67, 906-914 (2007). 19
AP wk0 wk1 wk2 wk3 wk5 wk6 wk7 wk8 Dong/MDACC 3D tumor traces on different weeks 0.4 Week 0 0.2 0.0-0.2-0.4-0.6-0.8-1.0 1.0 0.8 0.6 0.4 Week 0 Week 2 Week 3 Week 4 Week 5 Week 6 0.2 0.0-0.2-0.4-0.6-0.8-1.0 SI 0.8 0.6 0.4 0.2 0.0-0.2-0.4-0.6-0.8 RL Dong/MDACC Inter-fraction vs. Intra-fraction Skin Mark Alignment Fiducial Alignment C. Nelson/Starkschall/MDACC 20
Workflow of IGRT and Adaptive Radiotherapy CT scanner CT-Simulation Treatment planning system Treatment Planning On/Off-line Adaptive Radiotherapy LINAC console Console CT console Alignment Workstation LINAC Patient couch CT-on-Rails Treatment Room A workflow diagram for inroom CT-guided adaptive radiotherapy AJ Mundt/UCSD 21
Why Adaptive Radiotherapy (ART)? Live patient and dynamic treatment ART = Adapting to changes (anatomy, dose distributions, biology ) Corrections beyond simple translation shifts Image-guided Adaptive Radiotherapy Schemes Prospective Correction Plan Auto-segmentation Re-planning On-line Off-line Final Dose Map Deforming doses Retrospective Evaluation Dong/MDACC CT 1 (1/15/2003) CT 2 (2/21/2003) difference Example H&N CT-on-rails Studies Deformable image registration to quantify anatomic changes Auto-segmentation Cumulating doses Int. J. Radiat. Oncol. Biol. Phys. 61 (3), 725-735 (2005). Wang et al. 22
How to evaluate and approve these deformed contours? Auto-Segmentation of H&N Anatomy Benefit of IGRT Alone Mask Setup vs. C2-bone Setup BB Alignment Bone Alignment Patient #2 23
Evaluation of Cumulative Dose Distributions Fraction-by-Fraction Cumulative DVH after 33 fractions Solid BB alignment Dash Original Plan Patient #1 Cumulative DVH after 33 fractions Solid Bone alignment Dash Original Plan Patient #1 24
At the end of treatment DVH-BB Setup Patient #2 At the end of treatment DVH-Bone Setup Patient #2 At the end of treatment DVH-C2 Setup (Replan) Patient #2 25
Reducing CTV/GTV Uncertainties Target delineation uncertainties Better knowledge about the disease and spread pathway. Multi-modality imaging (PET, SPECT, fmri etc.) Inter-observer variations C CT only CT + PET SPECT PET only + SPECT only JD Cox Summary and Discussion The living patient and the dynamic treatment problem A computer treatment plan is not WYSIWYG Reducing systematic errors Improving simulation techniques Designing better immobilization devices Off-line correction Reducing day-to-day variations On-line image guidance Better immobilization devices Trends Volume effect Mid-course correction Adaptive Radiotherapy 26