CECE-Therapy: Patient Motion: Adaptive RT Adaptive Management of Patient Motion in Radiotherapy Di Yan, D.Sc. William Beaumont Hospitals & Research Institute Educational Objectives I. To learn the options of 4D planning II. To understand the sensitivity of 4D planning on motion uncertainties, as well as the methods for uncertainty management III. To learn the key components of adaptive treatment process and their functions Outlines I. Geometry based and dosimetry based 4D planning for motion compensation II. Motion uncertainty and its dosimetric effect on 4D planning. The management options III. Key components and functions of adaptive treatment process Page 1 4D Planning: Geometry Based ITV Construction Patient specific target for motion compensation constructed using target motion excursion (Margin = 0.5*Excursion) ¤ determined using fluoroscopic image (Balter JM, et al. Med Phys 1994, 21:913) ¤ 4D CT or directly from a MPI image (Underberg Rene, et al. IJROBP 2005, 63:253-60) Purely geometric compensation, no dose distribution is used in the margin design Overestimate the target margins significantly Motion Effect in Dose Distribution Motion blur effect of dose distribution has been demonstrated long time ago using the convolution approach, spatially invariant dose distribution + motion pdf (Leong J. PMB 1987, 32:32737) In reality, one should also consider the effect of patient internal density variation & the leave interplay effect if dose is delivered using the MLC based IMRT (Chui CS, et al. Med Phys 2003, 30:1736-46, & Bortfeld T, et al. Med Phys 2002, 47:2203-20) 4D Dose Summation Tissue density distribution Machine output VoI subvolume position D (v ) = n ∑∫ i =1 t∈Ti dD v (x t ( v ), ρv t , uvt ) ⋅ dt dt In time domain =∫ (∫ D ( xv , ρv , uv ) ⋅ pdf ( xv , ρv ) ⋅ d xv ⋅ dρv )⋅ pdf ( uv ) ⋅ d uv =∫ ( = ) v v v v v v v ∫ D ( x , ρ M , u ) ⋅ pdf ( x ) ⋅ d x ⋅ pdf (u ) ⋅ d u v ∫ D (x, v v v v ρ M , u c ) ⋅ pdf ( x ) ⋅ d x In freq domain Apply the mean CT Constant output Page 2 4D Planning: Dose Based ITV Construction Courtesy Dr Liang from WBH pdf Dose convolution with motion pdf measured at treatment simulation Perform the margin calculation iteratively by adjusting beam aperture Effect of The Prescription Dose 3 cm Target margin is strongly dependent on the prescription dose point 85% of the iso M1 = 7.7 mm Therapeutic ratio could be future increased by reducing treatment beam aperture & allowing higher heterogeneity dose in the target (Engelsman M, et al. IJROBP 2001, 51:1290-8) 3 cm 70% of the iso M1 = 4.1 mm Inter-patient Heterogeneity Target margin depends on the dose distribution which greatly relies on the tumor location M1 = 7.5 mm M2 = 5.6 mm M1 = 2.9 mm M2 = 2.6 mm With respect to the prescription dose of 75% ~ 95%, Target Margin = 0.1 ~ 0.4*Excursion 2.5 cm Page 3 4D Planning: 4D Inverse Planning Similar to 3D Inverse Planning, but Include the motion pdf from m-phase 4D CT in the 4D dose summation (Alexei Trofimov, et al. PMB 2005, 2779-98) v v v v v v v D ( v ) = ∫ D ( x ( v ), ρ , uc ) ⋅ pdf ( x , ρ ) ⋅ dx ⋅ dρ = Opt v v v v v v v D ( x1 , ρ1 , uc ) + ⋅ ⋅ ⋅ + D ( xm , ρ m , uc ) m v F ( D ( v , u c ) , v ∈ VoIs ) { uc} Zhang P, et al. submitted to IJROBP 4D Planning Methods: Summary 4D Planning Methods for Motion Compensation Geometry based ITV (Margin = 0.5*Excursion), Dosimetry based ITV (Margin = 0.1 ~ 0.4*Excursion) 4D inverse planning (Margin = 2 mm) All 4D planning methods perform treatment planning adaptable to patient motion measured at the pretreatment simulation alone, but not those during the treatment delivery Page 4 Clinical Observations Significant inter-treatment baseline variation & breathing pattern (cycle to cycle) variation (in time domain), but relatively small variation in motion standard deviation (in frequency domain). (Geoff Hugo, et al. Radiother Oncol 2006, 78:326331. Jan-Jakob Sonke, et al. IJROBP 2008, 70:590-8) Intra-treatment baseline drift is limited within small group (5%~10%) of patients Dose response related variations (volume shrinkage, baseline position change, relative distance change, et al) could be significant after the first few weeks of treatment resulting significant dose variation in normal organs Uncertainties of Motion pdf Variations between the reference motion pdfr and those during the treatment deliveries, pdftx SI Displacement Tx 2 Tx n Tx 1 … µ1 µ µ2 µn Systemic Error, µ , for the entire treatment Systemic error, µk , for each treatment Uncertainties of Motion pdf Uncertainty depends on the motion management 25% R e fe re n c e M e a n C o r r e c t io n B o n e C o r r e c t io n N o C o r r e c t io n Motion PDF 20% 15% 10% 5% 0% -2 - 1 .5 -1 -0 .5 0 0 .5 1 1 .5 2 2 .5 S I D ir e c t io n ( c m ) Page 5 2% 0% 7.2 ∆ Dose -2% -4% MeanCorrection ∆µ (mm) ∆σ (mm) −0.2 0.0 BoneCorrection 1.0 0.2 NoCorrection 2.6 2.0 -6% -8% -10% -12% Target (SI Direction) Large Motion pdf Variation 25% R e fe re n c e M e a n C o r re c tio n B o n e C o rr e c tio n N o C o rre c tio n Motion PDF 20% 15% 10% 5% 0% -2 -1 0 1 2 3 S I D ir c tio n io n (c m ) 10% 5% 0% ∆ Dose -5% 7.2 7.7 8.2 8.7 9.2 9.7 -10% -15% ∆µ (mm) ∆σ (mm) -20% -25% MeanCorrection -30% BoneCorrection -35% NoCorrection 0.0 10.0 12.0 −0.2 0.5 0.1 -40% Target SI Directin Page 6 Motion Uncertainty Management Robust Planning (Timothy Chan, et al. PMB 2006, 51:2567-83) ¤ include the bounds of motion uncertainties (previously determined) in the pre-treatment planning ¤ If motion variations are within the bounds, the treatment plan needs no modification ¤ However, the treated volume can be quite large, if generic variation, specifically the systematic variation, is considered Change in The Mean Target Position (mm) Clinical Observation: Baseline Variation 20 P a tie n t P a tie n t P a tie n t P a tie n t P a tie n t 15 1 3 5 7 9 P a tie n t P a tie n t P a tie n t P a tie n t P a tie n t 2 4 6 8 10 10 5 0 -5 -1 0 σ = 6.7 mm -1 5 -2 0 S e ssio n 1 S e ssio n 2 S e ssio n 3 S e ssio n 4 S e ssio n 5 S e ssio n 6 S e ssio n 7 S e ssio n 8 G. Hugo, Radiother Oncol, 2006, 78:326-331 Clinical Observation: SD Variation Change in standard deviation (mm) 20 P a tie n t P a tie n t P a tie n t P a tie n t P a tie n t 15 1 3 5 7 9 P a tie n t P a tie n t P a tie n t P a tie n t P a tie n t 2 4 6 8 10 10 5 0 -5 -1 0 σ = 1.7 mm -1 5 -2 0 S e ssio n 1 S e s sio n 2 S e s sio n 3 S e s sio n 4 S e ss io n 5 S e ss io n 6 S e ss io n 7 S e ssio n 8 G. Hugo, Radiother Oncol, 2006, 78:326-331 Page 7 Motion Uncertainty Management Adaptive Management ¤ Modify the treatment plan to cope with the patient specific variations while the treatment is running ¤ Identify characteristics of individual motion to provide a proper decision for baseline correction or adaptive planning modification ¤ Dose feedback to support the adaptive planning evaluation & modification Model Identification Adaptive Control (MIAC) Radiotherapy Process ADJUSTMENT MECHANISM Goals ADAPTIVE PLANNING MOTION IDENTIFICATION Plan TREATMENT DELIVERY SYSTEM Delivered Dose feedback Adaptive Motion Management Motion Identification & Management ¤ Baseline Variation – detected and corrected using onboard CBCT directly (G. Hugo, IJROBP 2007, 69:1634-41) ¤ The pdf Pattern Variation – detected using fluoroscopic image, CB projection images or a surrogate such as surface motion detection Page 8 Online Motion Verification (CB Fluoro Imaging) kV Fluoro 10% -12 8% -8 Probability Sup.<--- Tumor Position (mm) --->Inf. Ref or DRF -4 0 4 8 Ref 0.0 (4.5) 6% Port -3.5(4.0) 4% 2% Ref 0.0 (4.5) Port -3.5 (4.0) 12 0 5 10 15 Time (second) 20 25 0% 30 10 5 0 -5 -10 Courtesy Dr Liang from WBH Online Motion Verification (CB Projection Imaging) CB Projection Courtesy Dr Hugo from WBH Adaptive Motion Management Adaptive Planning Modification ¤ Dose feedback + 4D robust inverse planning ¤ Include all pre-measured pdfs in the planning v Dn (v , u c ) = Dk ( v ) v v v v v v v + ( n − k ) ⋅ ∫ d ( x ( v ), ρ , u c ) ⋅ pdf 1− k ( x , ρ ) ⋅ d x ⋅ d ρ Opt v v F ( D n ( v , u c ) , v ∈ VoIs ) { uc } Page 9 Adaptive Motion Management: Summary The most important issue in managing patient motion is to eliminate the baseline variation Daily CBCT imaging localization & correction could be the most efficient method to perform this task Intra-treatment motion pdf variation can be detected using CB fluoro, CB projection imaging or (maybe) a surface surrogate. This detection is used to guide the selection of the adjustments Adaptive planning modification = Dose Feedback + 4D Robust Inverse Planning Page 10