Advanced imaging technologies a revolution in diagnosis and treatment Optimization of image quality through online motion correction Oline Vinter Olesen December 10, 2010 Medical Solutions, Siemens Institute of Informatics and Mathematical Modeling, Technical University of Denmark Dept. of Clinical & Physiology and Nuclear Medicine & PET and Cyclotron Unit, Rigshospitalet Supervisors Professor, PhD, Rasmus Larsen, DTU Associate professor, PhD, Rasmus R. Paulsen, DTU Head of Simens Healthcare, Bjarne Roed, SIEMENS Head of Department, Professor, dr. med, Liselotte Højgaard, RH
Advanced imaging technologies a revolution in diagnosis and treatment Outline Patient motions induced image degeneration Motion compensation in PET brain imaging New motion tracking approach A phantom study Appetizer
Conventional PET Annihilation Modificeret fra Bailey et al., 2005 Detection Domino portræt, Ken Knowlton.
High Resolution Scanners HRRT PET scanner Octogonal geometry 2 layers of crystals 120.000 crystals Spatial resolution down to 1.4 mm
Problem Head motions end up counteracting the technological advances of high resolution scanners High Resolution Research Tomograph (HRRT), Siemens Conventional clinical PET scanner, GE Advance
Image degeneration Worst case scenario from Amsterdam, Marc C. Huisman et al. IEEE MIC 2010 Motion increases Simulation study from Manchester, J. M. Anton Rodriguez et al. IEEE MIC 2010
Something has to be done! An 15 O water study An 15 O water study
Solution! Increasingly expectations Decreasing resources Smart and effektive solutions Low cost & Time saving
State of the art: Motion tracking in PET
Serious Problems 1. Fixation of the tool 2. Stability of the system 3. Registration of the tool 4. Time consuming
New Tracking Approach Aims Design a tracking system with the following requirements: Stability No markers Sub millimeter precision Comfortable Matches the scanner geometry No interference with the PET acquisition Automatic Future perspective Having a system that can be: build into future scanners used for other applications (e.g. other image modalities and the radiation therapy)
Structured Light System System: One miniaturized DLP projector and two BW CCD cameras Capturing: 15 fps captured with two cameras tracking with 3 4 Hz
Stereo Vision System
Tracking Principle 1. Image capture (Three cosine patterns) 2. Reconstruction of 3D point clouds (Phase shifting interferometry) 3. Alignment of point clouds to triangulated reference target (Specialized ICP alg.) 4. Reference target is pre generated (Markov Random Field surface recon) Output: Rigid transformations with 6 degrees of freedom 5. Compute the rigid body transformation (Close form solution)
Comparison Experiment with the Polaris Vicra Simultaneous image capture and Polaris tracking Controlled movements in steps of 5 o from 15 o to 15 o Images were captured while the head was stationary O. V. Olesen, R. Paulsen, L. Højgaard, B. Roed, and R.Larsen, Motion tracking in narrow spaces: A structured light approach, Medical Image Computing and Computer Assisted Intervention (MICCAI), 2010, p. 253 260
Tracking Experiments Scans from the left camera at the seven positions 1. Surface of reference scan was reconstructed (0 degrees) 2. The other scans were aligned to the reference scan using ICP 3. The angles of rotations were estimated and compared with the results Position at 15 0 of Polaris and the performed motions Alignment of scan at 15 0 into the reference surface Paulsen, R., Bærentzen, J., Larsen, R.: Markov Random Field Surface Reconstruction, IEEE TVCG, 2009
Results Angle of rotation Absolute error RMSE SL = 0.11 o RMSE POL = 0.08 o Similar accuracy of the two systems 10 cm fra rotationsaksen svarer 0,1 o til 0,2 mm
NIR Structured Light System NIR light source & human scans
Advanced imaging technologies a revolution in diagnosis and treatment There are relevant motions to correct for in PET brain imaging A new tracking approach was presented designed for the HRRT PET scanner; a miniaturized structured light system A study comparing our system with a commercial system. The performed motions showed similar accuracy of the two systems Oline Vinter Olesen ovol@imm.dtu.dk
Advanced imaging technologies a revolution in diagnosis and treatment There are relevant motions to correct for in PET brain imaging A new tracking approach was presented designed for the HRRT PET scanner; a miniaturized structured light system A study comparing our system with a commercial system. The performed motions showed similar accuracy of the two systems Oline Vinter Olesen ovol@imm.dtu.dk