Some Essential Physics of Flat-Panel Detectors

Transcription

1 AAPM/RSNA Tutorial on Equipment Selection: FlatPanel Detectors Some Essential Physics of FlatPanel Detectors Jeff Siewerdsen, PhD Associate Professor, Department of Biomedical Engineering Johns Hopkins University Johns Hopkins University Schools of Medicine and Engineering Acknowledgements Collaborators John Carrino (Johns Jerry Prince (Johns Russ Taylor (Johns Rebecca Fahrig (Stanford University) Angel Pineda (CSU Fullerton) Art Burgess (Harvard Medical School, ret.) Funding Support NIH R01CA NIH R01CA Siemens Healthcare (Erlangen) Carestream Health (Rochester NY) Disclosures Synergy IGRT, Elekta Oncology Systems Scientific Advisory Board, Carestream *OffLabel and investigational use of medical devices will be discussed in this talk. FlatPanel Detectors (FPDs) FPDs are xray detectors typified by: Largearea (up to ~43 x 43 cm 2 ) Active matrix readout using asi:h technology (no mechanical scanning) Compact form factor Digital xray image signal No geometric distortion Robustness to magnetic fields Main components of FPDs: Xray converter Direct (Photoconductor: ase, PbI 2, ) Indirect (Scintillator: CsI:Tl, Gd 2 O 2 S:Tb, ) Pixel component for charge integration / readout Photodiode or storage capacitor or diode configuration Readout electronics / digitization Row drivers ( switching rows ) Readout amplifiers (ASIC columns ) Image processing / display Image Formation: Indirect Detection Imaging Chain Incident xrays Interacting xrays Conversion to optical photons A: Complete absorption B: K xray escape C: K xray reabsorption Spread of optical photons Coupling of optical photons Integration by pixel aperture Sampling of pixel matrix Readout with additive noise scintillator A B C

2 Image Formation: Direct Detection Imaging Chain Incident xrays Interacting xrays Conversion to electronhole pairs A: Complete absorption B: K xray escape C: K xray reabsorption Spread of eh pairs (~negligible) Coupling of eh pairs Integration by pixel aperture Sampling of pixel matrix Readout with additive noise photoconductor A + B + C + +HV E HV + XRay Converters Conventionally, there is a tradeoff between: Thickness (thicker improved xray absorption) Spatial resolution (thinner improved spatial resolution) Thin Phosphor X This tradeoff is relaxed in modern FPDs: Structured scintillators: CsI:Tl Photoconductors: Efield collection of eh pairs X Thick Phosphor Photoconductor Structured Scintillator X X ~5 µm Pixel Matrix (Array) Gate Line PD Row Drivers (Gate Drivers) Gate Line PD Integrating Amplifiers (1 / ) + ADC

3 PD Gate Line Hi Q V ADU Q V ADU PD Gate Line PD Gate Line Hi Q V ADU Q V ADU PD Gate Line

4 Pixel Matrix (Array) 8/30/2010 Active Matrix Readout Some observations Signal spread beyond pixel apertures in scintillator ( indirect detection) But no charge sharing between pixels (in discrete PD array) Finite fill factor (in discrete PD arrays) Charge trapping effects in asi:h sensor and Realtime Readout ~2µm Frame rate dictated in large part by number of ROWS Typical frame rates: ~1 fps (radiography) ~30 fps (fluoro) Glass Substrate Pixel Pitch µm Advanced pixel architectures Various switch configurations: or Diode(s) switch (may require optical reset) Other base materials: asi:h, polysi, ase Continuous PD designs Advanced pixel architectures: Onpixel amplification Image adapted from R. A. Street (Xerox PARC) Imaging Performance Imaging Performance PixelBased Metrics Dark current and drift Dark noise Dynamic range Linearity Image lag and ghosting PixelBased Metrics Dark current and drift Dark noise Dynamic range Linearity Image lag and ghosting ImageBased (Fourier Domain) Metrics Modulation transfer function (MTF) Noisepower spectrum (NPS) Detective quantum efficiency (DQE) Noiseequivalent quanta (NEQ) ImageBased (Fourier Domain) Metrics Modulation transfer function (MTF) Noisepower spectrum (NPS) Detective quantum efficiency (DQE) Noiseequivalent quanta (NEQ) Pixel and line defects Timing artifacts EM interference Saturation Postprocessing (tone scaling) Pixel and line defects Timing artifacts EM interference Saturation Postprocessing (tone scaling) 4

5 Imaging Performance PixelBased Metrics Dark current and drift Dark noise Dynamic range Linearity Image lag and ghosting Imaging Performance PixelBased Metrics Dark current and drift Dark noise Dynamic range Linearity Image lag and ghosting ImageBased (Fourier Domain) Metrics Modulation transfer function (MTF) Noisepower spectrum (NPS) Detective quantum efficiency (DQE) Noiseequivalent quanta (NEQ) Pixel and line defects Timing artifacts EM interference Saturation Postprocessing (tone scaling) ImageBased (Fourier Domain) Metrics Modulation transfer function (MTF) Noisepower spectrum (NPS) Detective quantum efficiency (DQE) Noiseequivalent quanta (NEQ) Pixel and line defects Timing artifacts EM interference Saturation Postprocessing (tone scaling) MTF Imaging Performance PixelBased Metrics Dark current and drift Dark noise Dynamic range Linearity Image lag and ghosting ImageBased (Fourier Domain) Metrics Modulation transfer function (MTF) Noisepower spectrum (NPS) Detective quantum efficiency (DQE) Noiseequivalent quanta (NEQ) Spectral Density (mm 2 ) Pixel and line defects Timing artifacts EM interference Saturation Postprocessing (tone scaling) Imaging Performance PixelBased Metrics Dark current and drift Dark noise Dynamic range Linearity Image lag and ghosting ImageBased (Fourier Domain) Metrics Modulation transfer function (MTF) Noisepower spectrum (NPS) Detective quantum efficiency (DQE) Noiseequivalent quanta (NEQ) Pixel and line defects Timing artifacts EM interference Saturation Postprocessing (tone scaling)

6 Imaging Performance PixelBased Metrics Dark current and drift Dark noise Dynamic range Linearity Image lag and ghosting ImageBased (Fourier Domain) Metrics Modulation transfer function (MTF) Noisepower spectrum (NPS) Detective quantum efficiency (DQE) Noiseequivalent quanta (NEQ) Pixel and line defects Timing artifacts EM interference Saturation Postprocessing (tone scaling) Primary Application A Few Makes and Models Pixel Format Area (cm 2 ) Pixel Pitch (µm) Fill Factor Converter Material Model Manufacturer Radiography 2022 x x ~0.8 CsI:Tl Definium 5000/8000 General Electric 3001 x x ~0.7 CsI:Tl Pixium 4600 Trixell 2688 x x NA Gd 2O 2S:Tb or CsI:Tl CXDI40G/C Canon Fluoroscopy 2048 x x ~0.7 CsI:Tl PaxScan 4030A Varian 2022 x x ~0.8 CsI:Tl Innova4100 General Electric 2880 x x >0.9 effective ase Safire Shimadzu 2881 x x NA CsI:Tl Pixium 4343RF Trixell Mammography 3584 x x MV Portal Imaging >0.9 effective 2048 x x >0.9 effective 2400 x x NA CsI:Tl 1024 x x 41 Gd 400 ~0.8 2O 2S:Tb or CsI:Tl ase Selenia Hologic ase LMAM Anrad Senographe Essential RID 1640A General Electric PerkinElmer 4030CB A Few Makes and Models 4030R 2520E 127 µm pixel 2520E Portable Veterinary Radiography, Security 2520V 2520D Industrial Dental CBCT 1313 NDT and Dental 1308 Dental 4030E Veterinary Rad A Few Makes and Models Portable FlatPanel Detectors 35 x 43 cm Thickness = cassette 2520V Courtesy of Dr. R. Colbeth (Varian Imaging Products) 194 µm 139 µm D D 4030A 4030CB 4336R 4343R Cardiac Mobile Carms General RF Mobile Carms, Dental, CBCT General RF, Vascular, CBCT Portable Radiography Fixed Radiography Varian 4336R 194 µm pixel pitch Wireless + rechargeable battery Retrofit to filmscreen and CR Carestream DRX1 139 µm pixel pitch Courtesy of Dr. R. Colbeth (Varian Imaging Products) Courtesy of Dr. J. Yorkston (Carestream Health Inc.)

7 Example Applications and Images Chest Imaging Example Applications and Images CSpine Hopkins Images University) courtesy of Dr. J. Yorkston (Carestream Health Inc.) Hopkins Images University) courtesy of Dr. J. Yorkston (Carestream Health Inc.) Example Applications and Images Extremities Imaging Advanced Applications Dualenergy imaging Tomosynthesis Conebeam CT PreProcessed (Linear Window/Level) Processed for Display (Nonlinear Tonescale) Hopkins Images University) courtesy of Dr. J. Yorkston (Carestream Health Inc.)

8 DualEnergy Imaging DualEnergy Imaging HighkVp SoftTissue Image HighkVp BoneOnly Image kvp I High w w s = High µ Bone µ Bone kvp I High w w b = High µ Soft µ Soft I ln Soft High ( I ) = ln( I ) w ln( I ) s I Bone High ( ) = ln( I ) w ln( I ) ln I + b Example DE Images ROC Performance DE SoftTissue Image DE Bone Image Conventional CR Image Small Nodules (<1 cm) Right Upper Lung DE DE Sensitivity DR DR AUC : DE DR (p = 0.05) AUC : DE DR (p = 0.003) 1Specificity 1Specificity H. Kashani et al. (Academic Radiology 2009)

9 Projection data ~1060 projections over ~10 o 60 o ThickSlice Reconstruction 3D filtered backprojection Primary applications 3D breast imaging Diagnostic chest imaging Tomosynthesis Projection ConeBeam CT Tomosynthesis (30 o arc) ConeBeam CT (360 o arc) Yoon et al. (SPIE Physics of Medical Imaging 2009) Projection data Multiple projections (~100500) over ~180 o 360 o Volume reconstruction 3D filtered backprojection (FDK algorithm) Image Quality: Key Characteristics Example FPDCBCT Systems Breast Imaging Dental / Maxillofacial ImageGuided Interventions ImageGuided Radiotherapy Large volumetric FOV Single orbit about the patient SubMillimeter Spatial Resolution SoftTissue Visibility JM Boone et al. (UC Davis) P Sukovic JH Siewerdsen (Xoran) (Johns Hopkins Univ.) DA Jaffray (Princess Margaret Hosp.)

10 Summary and Conclusions FlatPanel Detectors Important base technology for digital xray imaging Realtime readout without mechanical scanning Cost still a major factor; improving with volume and yield Imaging Performance Factors Pixelbased performance (drift, linearity, lag, ) Fourierbased performance (MTF, NPS, DQE, NEQ) Image processing is key to clinical image quality Burgeoning scope of applications Radiography, fluoroscopy, Dental, NDT, Portable, veterinary, Advanced: DE imaging, tomosynthesis, conebeam CT Thank you and enjoy your week at RSNA! WANTED: Research Scientist CONTACT: Jeffrey jeff.siewerdsen@jhu.edu H. Siewerdsen (Johns