Aperture Based Inverse Planning AAPM Summer School 2003

Transcription

1 Aperture Based Inverse Planning AAPM Summer School 003 D.M. Shepard, M.A. Earl, Y. Xiao, C.X. Yu Acknowledgements Ziping Jiang Allen Li Shahid Naqvi James Galvin Di Yan Prowess, Inc. University of Maryland School of Medicine Thomas Jefferson University Outline Overview of aperture-based planning techniques -- Yu Direct Aperture Optimization -- Shepard For Step & Shoot IMRT For IMAT Contrast to the Beamlet-based Model BEV divided into a grid of beamlets Optimized Fluence Map

2 Optimized Fluence Map Beamlet-Based Based Inverse Planning Leaf Sequencing Two-step approach to treatment planning:. Fluence map optimization Delivery constraints ignored. Leaf sequencing Accounts for delivery constraints Employed by nearly all commercial vendors: Corvus (NOMOS). Pinnacle (ADAC). Plato (Nucletron( Nucletron). Focus (CMS) Theraplan (MDS Nordion) Deliverable Apertures Beamlet-based Inverse Planning The separation of intensity map optimization and leaf sequencing results in a large number of segments with the following consequences:. Inefficient delivery (large # of MUs, long delivery time). Requires extremely high geometric accuracy and intensive QA efforts 3. A dramatic change in practice Aperture-Based Optimization A One-step process. Contour-based planning Anatomy contour-based Isodose contour-based. Direct Aperture Optimization Contour-based using anatomy Prostate Case Contour-based using anatomy Prostate Case PTV PTV - rectum

3 3 RTOG #H00 Oropharyngeal Cancer st aperture nd aperture Targets 66 Gy Primary Yellow 60 Gy Subclinical Green 60 Gy Lt Nodes Blue 54 Gy Rt Nodes Teal Critical Structures Rt Parotid Brown Lt Parotid Purple Spinal Cord Pink Brainstem Red Covers all targets Covers boost volume 3rd aperture 4th aperture TJU - Oropharyngeal Protects spinal cord Protects Rt. Parotid Isodose Green 66 Gy Light Blue - 60 Gy Red - 54 Gy Blue - 45 Gy Structures Orange Parotid Red PTV66 Green PTV60 Blue PTV54 Purple PTV60 nodes TJU - Oropharyngeal TJU - Oropharyngeal ABIP standard MLC (green) ABIP standard MLC (green) ABIP standard MLC (green)

4 4 Anatomy Contour-Based Anatomy Contour-Based Automatic segment generation based on anatomy. Optimize segment weights SOWAT Approve plan for delivery Approve plan for delivery Anatomy Contour-Based Anatomy Contour-based Results Segment Outline and Weight Adapting Tool w/o SOWAT w/ SOWAT Aperture Optimization Isodose Based Isodose Contour-Based Breast IMRT at William Beaumont Used at William Beaumont Hospital Used mainly for Breast

5 5 Whole Breast IMRT - Open Field MLC Lung Block Segment Lung Treatment Volume 05% Isodose MLC Block 0% Isodose MLC Block 5% 0% 05% 00% 95% 90% Lavender=5% Green=0% Yellow=05% Red= 00% Blue=95% Light Blue=90% 5% Isodose MLC Block Lateral IMRT Segments 5% 0% 05% 00% 95% 90%

6 6 After definition of segments, optimizing the weights. 5% 0% 05% 00% 95% 90% Wedges Comparison with 3DCRT Transverse View IMRT 5% 0% 05% 00% 95% 90% Wedges Comparison with 3DCRT Sagittal View IMRT Clinical Results 450 Patients Treated Through April 00 (6/99-4/0) Median Patient Age - 59 Years Stage 0-7% Stage I - 56% Stage II - 7% Planning and Delivery Time Median Time For Treatment Delivery Minutes (depends upon # of MLC segments) Median IMRT Planning Time - 45 Minutes 3D Planning Time - Median 40 Minutes Contour-based planning Summary Anatomy contour-based Isodose contour-based

7 7 Advantages Aperture shapes are intuitive Small number of apertures = quick delivery. No leaf-sequencing. No need to recompute dose after optimization. Plan quality does not degrade. Disadvantages Difficult to make robust for all target sizes and shapes: typically used in a site specific fashion The solution space is limited before optimization. Plan quality may not rival that of beamlet- based optimization. Direct Aperture Optimization (DAO). Inverse planning technique where aperture shapes and weights are optimized simultaneously.. All of the MLC delivery constraints are included in the optimization. 3. The number of apertures per beam angle is specified in the prescription. Optimization via Simulated Annealing ) Pick a parameter (leaf position, aperture weight) randomly ) Change the parameter by a random amount 3) Calculate objective function based on the new dose distribution 4) Objective function lower: accept change 5) Objective function higher: accept change with certain probability Prescription: 3 apertures per angle Begin with 3 identical copies Pick an Aperture and Make a Change Aperture Leaf pair 6 Left leaf position Move leaf in cm

8 8 Keep or Reject the Change MLC Constraints Some sample Elekta constraints: Based on:. MLC constraints.. Cost function & Annealing Rules. ) Opposed leaves cannot come closer than -cm from one-another < cm ) Opposed-adjacent leaves cannot come closer than -cm from one-another < cm Not allowed Not allowed Final intensity map from DAO After numerous iterations... Add them up along with their weights Small number of apertures can produce large number of intensity levels Example: 3 apertures/angle Small number of apertures can produce large number of intensity levels N n = n N = Number of intensity levels n = Number of apertures 3 separate weights For 3 apertures, 7 intensities For 4 apertures, 5 intensities For 5 apertures, 3 intensities For 6 apertures, 63 intensities

9 9 Head and Neck Patient H&N results: axial 7 beam angles 3 apertures per angle H&N results: Sagittal H&N Results: DVH Lt. Mandible Cord Rt. Mandible Lung Patient Lung Patient beam angles 3 apertures per angle 7 beam angles 3 apertures per angle

10 0 Lung Patient Oropharyngeal - Axial Targets 66 Gy Primary 60 Gy Subclinical 60 Gy Lt Nodes 54 Gy Rt Nodes Critical Structures Rt Parotid Lt Parotid Spinal Cord Brainstem Oropharyngeal - Coronal Comparison with Beamlet-based Inverse Planning A direct comparison was made between DAO and Corvus for four clinical cases. In each case, we sought equivalent treatment plan quality. Treatment Site Head and Neck Lung Prostate Comparison with Beamlet-based Inverse Planning Corvus # of segments Corvus # of MU DAO # of segments 0 DAO # of MU Comparison with Beamlet-based Inverse Planning DAO reduced the number of segments by up to 95%. DAO reduced MU by up to 80%. Note: Results specific to Elekta and Corvus Pancreas

11 Direct Aperture Optimization Benefits. Highly conformal IMRT plans with only 3 to 5 apertures per beam.. Delivery in traditional 5 minute time slots. 3. The user has complete control over the complexity. Direct Aperture Optimization Benefits 4. Provides optimal aperture shapes and weights. 5. No leaf sequencing. 6. Can be used for IMAT treatment planning. IMAT Rotational IMRT is delivered on a standard treatment unit using a conventional MLC. A dynamic arcing approach is used where the shape of the MLC changes during the arced beam delivery. Multiple overlapping arcs are used to produce a modulated intensity pattern from each beam angle. IMAT Treatment Planning Each arc is approximated by a series of static beams typically separated by 0 degrees. During the arced beam delivery, the MLC interpolates between adjacent beam shapes. IMAT Constraint I Leaf motion is constrained by the maximum leaf travel speed and gantry speed. Gantry speed = 0 degrees/sec Leaf travel speed = cm/sec Maximum leaf travel between two adjacent angles = cm Gantry angle = 35 Not allowed Gantry angle = 45 IMAT Constraint II Gantry speed and dose rate must be constant throughout each arc. All beam weights within an arc must be equal. d = 0 cm d = 4 cm

12 Cylinder -Arc configuration IMAT - Planned overlapping arc 6 arcs total overlapping arc Cylindrical phantom delivery Planned Delivered overlapping 40 overlapping 6 arcs total arc arc

13 3 Sample case: prostate Arc configuration for prostate case Prostate IMAT axial slice overlapping -0 to -60 overlapping 0 to 60 8 arcs total PTV overlapping -0 to -40 overlapping 0 to 40 Purple: 95% coverage Prostate IMAT: coronal Prostate IMAT: sagittal Purple: 95% coverage PTV Purple: 95% coverage PTV

14 4 Sample arc field sequence - prostate treatment plan 0 degrees 30 degrees 40 degrees IMAT Summary 50 degrees 60 degrees Black part represents open field Each block = 0.5 cm Leaves do not travel more than -cm between adjacent angles Inverse planning for IMAT is significantly more complex than that for step-and and-shoot IMRT due to added MLC constraints. DAO can account for these constraints and provides efficient IMAT treatment plans. Aperture Optimization Summary Aperture optimization provide efficient high quality IMRT treatment plans. These plans are easy to delivery and easy to verify. Contour-based planning produces aperture shapes based on the patient s anatomy or isodose curves. Direct aperture optimization simultaneously optimizes the aperture shapes and weights. Isodose Green 66 Gy Light Light Blue - 60 Gy Red - 54 Gy Blue Blue - 45 Gy Structures Orange Orange Parotid Red PTV66 Green Green PTV60 Blue Blue PTV54 NOMOS CORVUS SYSTEM Breast Patient Purple Purple PTV60 nodes 5 beam angles 3 apertures per angle

15 5 Breast Patient - Axial Breast Patient - Sagittal Contour-based optimization using anatomy TJU has implemented a technique whereby additional aperture shapes are created based upon the hot spots and cold spots found after an initial optimization. Multiple optimizations are performed with additional aperture shapes included in each. TJU Algorithm. Scale the existing dose to meet all constraints on sensitive structures. Search for additional segments in a given beam direction by analyzing the missing target dose and allowed dose deposit with exhaust of sensitive structure limits (BEV dosimetry) 3. Re-optimize the beam weights with added apertures and delete unused apertures. 4. Iterate steps # to #3 Comparison of three inverse planning methods for patient with cancer of the tonsil Lateral IMRT Segments Number of MUs Geometric planning 506 Automated method 93 Fluence optimization 73 Number of Segments

16 6 Volume-Based Optimization Oropharyngeal - DVH Lateral IMRT Segments Direct Aperture Optimization for IMAT Each beam is initialized to match the beam s eye view of the tumor volume. The number of arcs, and the range of each arc is included in the prescription. DVH based penalties are used to define the treatment goals. Pancreas Patient After definition of segments, optimizing the weights. 7 beam angles 3 apertures per angles

17 7 Pancreas Patient Axial Slice Pancreas Patient Sagittal Slice DAO for IMAT Conclusions Direct aperture optimization provides a tool for inverse treatment planning for IMAT. This tool has been tested on patient cases and delivery verification has been performed using a cylindrical phantom.