Imaging in Radio-Oncology : Sometimes imaging = miraging Can you find the position of the camels? PD Dr Jean-François Germond SSRMP Medical Physicist Service de radiothérapie du DPO La Chaux-de-Fonds FMH resident physics training course in RO, PSI, 3/09/2007 Objectives Usage of imaging in medicine 1/2 1. Historical introduction 2. Anatomical imaging in RO localization 3. Functional imaging in RO localization 4. Immersion into 3D 5. 4D imaging in RO localization 6. Virtual simulation imaging 7. 3D guidance imaging Mechanical Waves Sound Ultrasound EM Potentials Biological Signals Medical imaging All types of waves Infrared Radio- Waves Magnet. Reson. Electromagnetic Waves Thermography E-Mgraphy Echography Stethoscopy Endoscopy Light Radiography X Rays Scintigraphy g Rays Low energies High energies Usage of imaging in medicine 2/2 Historical evolution of imaging in RO 4D CT Well established in diagnostic Extensively used for tumor detection and staging Used throughout radio-oncology: for planning for irradiation Reference imaging CT PET-CT MRI 2000 1990 2006 for verification Simulator 1980 1970 Cone beam 1960 Portal imaging Guidance imaging 1
Multimodal imaging Intensity modulation Virtual simulation Image guidance Block Imaging in conventional simulation 1/4 Imaging in conventional simulation 2/4 Therapy table Collimator X-Ray tube Simulated beam Digital camera Conventional simulator 1960 Width jaw Height jaw Block Simulator radiographs are used for drawing beams aperture Only based on D Bony structures Air cavities Contrast agent in specific organs Only 2D D Imaging in conventional simulation 3/4 Verification by portal film Imaging in conventional simulation 4/4 Still in use for palliative treatments => 30% of patients Still in use for pre-treatment verification of ports Reference image = Simulator image Portal film or electronic image used for verification Simulation film Verification film The 4 cornerstones of modern radio-oncology Features of imaging in RO Workflow 1. RO is image guided (IGRT) in three aspects: a. For the definition of volumes during planning (localization). Years3/4 of workflow 80 is imaging Years 2000 b. For the setup of fields in simulation (virtual simulation). c. For the verification of the ballistic during treatment (guidance). 2. Imaging for RO is multimodal (CT,MRI, PET, ) 3. Images RO must be stored, transferred and linked (DICOM) Years 2000 Years 90 2
Paradigm of target volumes definition GTV : Gross Tumor Volume CTV: ITV: PTV: ICRU 50 & 62 + subclinical envolvement + Internal margin + Setup margin GTV and CTV are oncological concepts based on a frozen anatomy Main issues of imaging in RO 1. Does the imaging modality allow to distinguish the target volume from its environment? Adjacent tissues with similar electronic densities in CT exams (atelectasy) 2. Does the imaging modality reflect the true anatomy? Artefacts 3. Does the image display allows the observer to perceive the target volume? JND index Definition: Visual perception of images The Just Noticeable Difference (JND) is the luminance difference that the average human observer can just perceive at a specified luminance level and viewing conditions Major steps in the RO workflow Localization Human eye is more sensitive in the low luminance range (Fechner law) Delineate in rooms with reduced ambient light (15-60 lux) 1 time Virtual simulation Guidance Irradiation DICOM GSDF conforming displays can compensate for this effect Dose planification Several times Major steps in the RO workflow Localization Virtual simulation Dose planification Guidance Irradiation Localization in RO Goal: Precisely define the position, shape and volume of organs Target (tumor, lymph nodes, tumor bed, ) At risk (spinal cord, lungs, bladder, rectum, ) Material: All 3D imaging modalities CT, MRI, PET, SPECT, MRS, Multimodalities registration and fusion Method: Organs delineation Manual contouring Automatic (thresholding, deformation, interpolation, atlas, ) 3
Anatomical imaging Information characteristics of imaging in RO Anatomical imaging Functional imaging => Structures and morphology => Biological and molecular abnormalities in tumors From Smith A. and K.S. Clifford Chao, Radiation Research, 2005 Paradigm of target volumes definition Historical evolution of imaging in RT 4D CT ICRU 50 & 62 GTV : Gross Tumor Volume CTV: + subclinical envolvement ITV: + Internal margin PTV: + Setup margin Reference imaging CT PET-CT MRI 2000 1990 2006 Anatomical imaging in RO is extensively used for delineating the GTV Simulator 1970 1980 Cone beam Portal imaging 1960 Guidance imaging CT equipment for RO Mode of acquisition Large gantry aperture Movable lasers Acquisition in helical (spiral) mode Table shift over 360 rotation(d) Pitch Collimation length(l) Flat table top Software + protocols Axial scan Pitch = 1 d = 1 x L Pitch = 1.5 d = 1.5 x L Pitch = 2 d = 2 x L Artist view, courtesy of Philips Medical Systems Full sampling for 360 recontruction Full sampling for 180 recontruction More complex for multislices CT 4
CT data acquisition CT data acquisition Sinogram Gantry angle Transmission profile Transmission profile Reconstruction by backprojection Example of a point object Image quality 0 15 24 180 24 Filtered Sinogram 15 The usefulness of images is affected by several of their characteristics: Artefacts Contrast Resolution Noise The image characteristics depend strongly on modalities. Transverse image D pretense of reality only! Image charateristics Artefacts CT Different structures seen PET Different structures seen Definition: An artefact is any visible structure which does not reflect the reality of the patient anatomy (wrong CT numbers) Artefacts may be patient related: a. Metallic objects inside the body (prothesis, clip, dental filling, ) b. Patient voluntary or involuntary motions Different resolution c. Field of view smaller than patient size (obese patients) Different noise From F. Schoenahl, Cours CFPFTRM, 2005 5
HN Artefacts Example of streaks due to metal inside the patient: Artefacts Example of shifts due to metal taped to the skin: Double hip prothesis dental filling Workaround: Reduce pitch or use special protocols Skin markers used for breast tangential fields virtual simulation Reconstructed below the skin D Workaround: Use special non metallic skin markers Hounsfield scale Windowing Hounsfield numbers (HN) = Tissue absorption coefficients compared to water Tissue Water HN 1000* => Normalized scale Water Increase contrast by mapping to 256 greyscale (8 bits) Choice of the level (L) and the window (W) from definite protocols 3000 Os compact 1000 2000 L/W : 1000, 2500 Bone window 800 80 600 70 Foie 400 200 0-200 Eau +4-4 Os spongieux 200 50 250 Graisse -80-100 60 50 40 30 Reins Pancréas Sang 1000 L/W : 50, 400 Mediastinal window 0 2 Exemple de la prostate à insérer -400-600 Poumons -550 20 10-1000 L/W : -600, 1600 Pulmonary window -800-1000 Air -990-1000 -950 0 Contrast adaptation L/W: 0, 1000, 2002000 (range (full in range) ROI) HN profile 1000 Resolution and contrast Limited resolution at high constrast 100 (HN HN ) GTV Surrounding (40 ( 90)) Calcul : %Contrast 13% 10 10 1. Dispay a full profile 2. Zoom in on VOI 3. Set L/W to average, range of VOI HN s Contrast 10 1 57 Resolution at low constrast 0.7 0.5 1 3 5 10 Object size in mm 6
Localization with MRI saggital slice Organ 2 Organ 1 Multiplanar views Partial volume effect Definition: The partial volume effect is the part of an organ which has a reduced constrast to finite extension in the cranio-caudal direction coronal slice Correct HN Slice thickness I worst resolution in cranio-caudal Reduced contrast Transverse slice Consequences: Partial volume effect Small objects like lymph nodes have too low contrast Organ borders can be blurred Organs delineation on CT 3-D paintbrush/eraser paradigm -GTV: Gross Tumor Volume (Target volume, ICRU 50) Example of fuzzy separation between right kidney and liver Transverse slice -Vessie - Rectum (Organs at risk, ICRU 50) Coronal reconstruction Historical evolution of imaging in RT 4D CT MRI equipment for RO Reference imaging CT MRI PET-CT 2000 2006 Movable lasers Magnetic field Antenna 1990 Simulator 1980 Flat table top Movable table 1970 Cone beam 1960 Portal imaging Guidance imaging Courtesy of Philips Medical Systems 7
Image fusion Localization with MRI Localization with MRI Localization with MRI Y gradient during waiting phase Localization with MRI Freq. MRI technics of spatial acquisition Phase Y i Frequence X i Contrast by T-weighting Relaxation times measured by special sequences T1: Longitudinal = spin-lattice Specific to tissue types T2: Transverse = spin-spin z gradient 1. Excite one plane 2. Select one raw 3. Select one column X gradient during reading phase Freq. Prostate example Examples of MRI artefacts Better definition (apex) Need training in fine anatomy details (nerves, muscles, ) Different volumes D Black hole in dental filling Courtesy of HUG Distortion at field edges Courtesy of HUG Courtesy of HUG Fusion of images Modalities other than CT are mostly used as complementary in RT: Question: How to used them combined for localization? Answer: by 2 successive electronic manipulations: 1. Registration of the 2 sets of images by a) Rigid method (ex: chamfer matching) b) Deformable method (ex: mutual information) 2. Display of the registrered images: Overlay of the 2 sets into one combined data set (fusion) Contour delineation on one modality and report in the other Constrained delineation 8
Image fusion Image fusion Image fusion Image fusion iteration Image fusion Image fusion a) Rigid 3D registration Mathematics of rigid registration Finds the best rigid transformation between 2 modalities (chamfer matching for bones) Reference image I ref ( r) Chamfer matching algorithm CT data set MRI data set Contours of interest C N p p Bone( I ref ) 3 translations Image of the day I day (r) 3 rotations Cost function Change T 1 cost N ps N T : Isocenter shift I day ( T r ) Injected CT T2 weighted CHUV Minimal cost I day ( T min r) I ( r) ref b) Deformable 3D registration Display of 2 data sets by colors superposition Maximalize similarity measure (mutual information) CT data set MRI data set morphing HUG Example of CT- MRI prostate fusion: Usages: Manual registration Quick check of registration quality --- Indications: HUG Morphological changes HUG Respiratory movements x Rectum filling Display of 2 data sets by colors addition Modality 1: greycale mapped to RGB pink = (255,0,255) Modality 2: greycale mapped to RGB green = (0,255,0) Overlay on screen mapped to the sum of pixel intensities: Only modality 1 = pink Only modality 2 = green Identical : (255,0,255) + (0,255,0) = (255,255,255) = white Display of 2 data sets by B&W + color Example of PET- CT fusion: Usages: Assignement of CT segmentation to tumor Example of CT- CBCT fusion: Usages: Visualization of morphology changes Tumor flashing Localization CT in pink CBCT of the day in green Changes in rectum filling Image from S. Senan S and D. De Ruysscher, Oncology Hematology, 2005 9
Functioning imaging Localization with MRI Localization with MRI Image fusion Localization with MRI HUG Display of 2 data sets side by side Example of CT- MRI prostate registration: HUG Organs delineation on CT - MRI Example of prostate : HUG HUG Usages: Synchronized windows Sharing cursor Simultaneous delineation MRI data set From J. Dipascale et al., SASRO meeting,geneva, 2004 CT data set Organs delineation on CT - MRI Example of H&N : Organs delineation on CT - MRI Example between different institutions : CHUV 10/3/06 7/4/06 From Smith A. and K.S. Clifford Chao, Radiation Research, 2005 Time evolution of edema? Geometrical distortions? Adapt CTV margin Paradigm of target volumes definition ICRU 50 & 62 GTV : Gross Tumor Volume CTV: + subclinical envolvement ITV: + Internal margin PTV: + Setup margin Functional imaging in RO can be used for delineating the GTV and possibly the CTV 10
Localization with SPECT Localization with SPECT Localization with SPECT Localization with SPECT Localization with MRS Localization with MRS Dose maximalization MRS spectroscopy in gliomas Technics able to detect specific compounds of tissue other than water and fat (choline, creatine, N-AcetylAspartate) Gives information about the level of alteration of metabolites in tumor MRS spectroscopy in prostate High concentration of choline inside the prostate = tumor signature Used for dose painting Peak Cho >Cr and NAA in brain tumors REF Tumor [Chang, Med.Phys.06] SPECT equipment SPECT data acquisition Rotation Photo HNe Head 2 Gantry angle H&N mask system Head 1 Emission profile SPECT data acquisition SPECT data acquisition Sinogram Gantry angle Gantry angle Emission profile Emission profile 11
Localization with PET Localization with PET Localization with SPECT Localization with SPECT Localization with SPECT SPECT- CT fusion Organs delineation on SPECT - CT Example in H&N (Parathyroid Adenoma) : Example in glioma: Tc-mici Lenox Hill Hospital, NY Need specific calibration for quantitative use Recidive not seen on CT GTV delineated on mici captation volume SPECT GTV mapped to CT data set Perfusion segmentation on SPECT Historical evolution of imaging in RT 4D CT Perfusion withtc-99m High value of perfusion = lung zone OK Mc Guire et al. (2006) Reference imaging MRI PET-CT 2006 CT 2000 1990 Simulator 1980 1970 Cone beam Volume to spare in dose calculation 1960 Portal imaging Verification imaging PET gantry PET- CT equipment PET data acquisition Positron emission and annihilation ± movable lasers Sinogram Flat table top CT gantry Gantry angle Courtesy of Philips Medical Systems Emission profile 12
Localization with PET Localization with PET Localization with PET Localization with PET Localization with PET Localization with PET PET CT fusion Organs delineation on PET - CT Example of glioma: CHUV FDG But boundary depends on grey level intensities GTV? Source: R. Bridwell, PETLinQ, 2006 Topography of structures are difficult to localize on pure PET images but much easier to read on PET- CT fused images Cavity correctly mapped between PET and CT GTV delineated on CT correspond to FDG fixation Increase CTV to involve all the FDG fixation volume Standart Uptake Value (SUV) Semiquantitative index based on ROI values Comparison between PET and CT scales Example for lung tumors: Formula: SUV (Activity inroi / Volume of ROI) Injectedactivity / patientweight FDG SUV = 1 if radiotracer is uniformly distributed within the organism Calculation of SUV requires corrections for: Photons attenuation and scatter Organ motion Partial volume effect Present recommandations for GTV delineation are SUV > 2.5 SUV > 40% of maximum SUV But sites and D technics dependent Images from S. Senan S and D. De Ruysscher, Oncology Hematology, 2005 PET : CT : Standard Uptake Value SUV Houndsfield numbers HN Concentration of radiotracer Tissue absorption coefficients Scale: 0 to 14 Scale: -1000 to + 1000 Organs delineation on PET - CT Example in oropharynx : Limitation of delineation on PET - CT Example for lung tumors: FDG CT SUV 40% SUV 2.5 D Corrections for the determination of radiotracer concentration (FDG) Images from U. Nestle et al., J. Nuclear Medicine, 2005 D Modelisation of the fixation metabolism (FDG different from glucose) From Smith A. and K.S. Clifford Chao, Radiation Research, 2005 D PET-CT is not yet the panacea for delineation 13
Localization in 3D Localization in 3D Localization in 3D Localization in 3D 3D virtual patient 3D model of images 3D virtual patient Slice séparation Example: 100 images, 512 * 512 pixels = 26.2 millions of voxels!! D frozen Voxel definition Surface rendering 3D model of contours Isocenter The contoured organs are becoming 3D objets The isocenter (barycenter) of the target volume (GTV) is defining the position of the objects system The isocenter is spotted by an orthogonal system of coordinates Patient marking The lasers system mimics the orthogonal system of coordinates The patient is tattoed at the lasers intersections with the skin 14
4D Localization 4D Localization 4D Localization 4D imaging 4D Localization Paradigm of target volumes definition Respiratory motion artefacts 2D MPR 3D rendering ICRU 50 & 62 GTV : Gross Tumor Volume CTV: + subclinical envolvement ITV: + Internal margin PTV: + Setup margin 4D imaging in RO can be used for assessing the ITV Helical CT: 1 turn / s Frozen virtual patient Inter-slices artifacts Internal organs movements Historical evolution of imaging in RT 4D CT Respiration affects the localization of many organs: Reference imaging MRI PET-CT 2006 CT 2000 1990 Simulator 1980 1970 Cone beam Courtesy of Philips Medical Systems Courtesy of Philips Medical Systems 1960 Portal imaging Verification imaging How to manage respiratory artefacts? Equipment for measuring respiratory phases Slow scanning technics Blurring PET 4D scanning Prospective (gating) Retrospective (sorting) Requires the measurement of the respiratory phases Pneumatic bellow RPM from Varian Spirometer Dyn R Spirometer ABC Elekta 15
Virtual simulation Virtual simulation Virtual simulation 4D Localization 4D Localization Multislices axial 4D-CT Organs delineation on 4D-CT CT images sorted by respiratory phases Low pitch spiral 4D-CT Organs delineation on 4D-CT CT images sorted by respiratory phases 90% 90% Example of lung tumor Colors = time spent in this position 10% Example of lung tumor 1 GTV per respiratory phase Internal Target Volume(ITV) 10% Courtesy of Philips Medical Systems Major steps in the RO workflow Localization Virtual simulation Dose planification See also delivery and verification session Guidance Irradiation Virtual simulation in RO Goal: Find the optimal balistic for irradiation which cover the target volume, Which spares the organs at risk. Material: Computer softwares(3d virtual simulation) Digitally Reconstructed Radiographs (DRR) Beam Eye View (BEV) Method: Beams conformation to target volume Choice of gantry angles Adaptation of the multileaves collimator (MLC) Technics of DRR s reconstruction Digitally Recontructed Radiographs(DRR) Reconstruction algorithm: Abrams et Goitein, 1983 Ray tracing : Siddon s method Interpolation : Trilinear or cubic Absorption : Curved integration Windowing : Remapping LUT table Technics of DRR s reconstruction Digitally Recontructed Radiographs(DRR) A exp i i dli i HNi i 1 water 1000 Hounsfield numbers HN i LUT i i voxel i D Pure mathematics Reconstruction algorithm: Abrams et Goitein, 1983 Ray tracing : Siddon s method Interpolation : Trilinear or cubic Absorption : Curved integration Windowing : Remapping LUT table 16
Image guidance Image guidance Image guidance Virtual simulation Beam Eye View method of virtual simulation Example of conformation on the prostate : Fusion of the projection of structures with DRR s Major steps in the RO workflow Localization Virtual simulation Guidance Anterio-Posterior beam DRR MLC GTV L-R Lateral beam Dose planification See treatmentirradiation planning sessions Major steps in the RO workflow Paradigm of target volumes definition Localization Virtual simulation See also delivery and verification session Guidance GTV : Gross Tumor Volume CTV: ITV: PTV: ICRU 50 & 62 + subclinical envolvement + Internal margin + Setup margin Dose planification Irradiation Image guidance in RO can be used for assessing the setup margin Goal: Image guidance in RO Check the treatment position of the patient Position of the target volume Position of the organs at risk Material: Imaging inside the treatment room With the iradiation beam (portal images) With an additional 3D imaging system (Echography, CBCT, ) Method: Registration with reference images DRR from virtual simulation(2d) CT slices from localization (3D) Various implementations of image guidance Electronic portal imaging : Set of 2D orthogonal portal images Golden markers MV CBCT Embarked kv imaging systems : Radiography Fluoroscopy CBCT Topometrical positioning systems Tomotherapy Stereoscopic kv imaging CT on rails Echography O-ring linac MR-Linac 17
CBCT guidance CBCT guidance CBCT guidance CBCT guidance CBCT guidance Historical evolution of imaging in RT 4D CT Cone Beam CT equipment Reference imaging MRI PET-CT 2006 X-ray tube 40-150kV CT 2000 Simulator 1980 1990 Collimator Filter 1970 Cone beam am-si panel Carbon fiber table top 1960 Portal imaging Verification imaging Elekta Synergy system in operation at Rotation to 260 Irradiation from 260 to 100 Irradiation from 260 to 100 Irradiation from 260 to 100 18
CBCT guidance CBCT guidance CBCT guidance CBCT guidance CBCT guidance CBCT guidance Irradiation from 260 to 100 Irradiation from 260 to 100 CBCT 3D reconstruction Examples of CBCT images 360-630 frames Frames @ isocenter P ( y y ( y z y Flex, z Flex Flex, z z Flex ) ) :Correction isocenter FDK algorithm Skull H&N Lung Scatter corrections P y, z) P ( y, z) 0.33 P ( y, z) ( Frame Bladder Verterbras Prostate Filtration p( y, z) dz d P ( y, z) h( z z) 2 2 2 d y z d :SAD 100 cm; h( z) : Median filter 5 Back-projection 2 1 d f r (, ) 2 2 0 0 4 d p y z d r xˆ y f ( r):3d patient density dy drzˆ 0, z0 ( d rxˆ) ( d rxˆ ) Clinical images from Clinical uses of CBCT images 1/2 Setup margins verification: PTV Inter-fractions motion: translations rotations deformations Clinical uses of CBCT images 2/2 Morphological changes : Inter-fractions variations: Tumor response Weight loss ITV Intra-fraction motion measured by 4D IGRT: breathing Courtesy of AVL Intra-fraction variations gaz Swallowing 19
Major steps in the RO workflow Localization Virtual simulation See delivery and verification session Guidance Dose planification Irradiation Take home message 1/2 Imaging is present in 75% of the RO workflow Localization serves as reference for planning and irradiation delivery Delineation by radio-oncologist ICRU as normative framework Beware of the intrinsic limitations of modalities The evolution of localization involves: Fusion of modalities Functional imaging 4D imaging Always use your judgment expertise to accept delineation in fused set of data Imaging during treatment (guidance) involves the registration of the image of the day with a localization image used as reference 2D: matching of portal imaging with DRR s 3D: registration between CBCT and CT The evolution of guidance involves: Fusion of images Tracking of organs 4D guidance Take home message 2/2 Always use your judgment expertise to accept patient setup adaptation Special acknowledgements Myriam Ayadi (CLB, Lyon) Giovanna Pasquale (HUG, Genève) Francis Verdun (IRA, Lausanne) Frédéric Corminboeuf (Inselspital, Bern) Markus Notter, Pascal Baudet (, La Chaux-de-Fonds) Stéphane Montandon (PMS, Gland) Carlos Rodriges (PMS, Fitchburg) Sometimes imaging = miraging Thank you for your attention? 20