PIERS, Hawaii, Oct Second Generation Microwave Imaging System: Phantom and Early Clinical Experience Paul M. Meaney, Dun Li, Qianqian Fang, Sarah A. Pendergrass, Margaret W. Fanning, Keith D. Paulsen

PIERS, Hawaii, Oct Outline Difficulties Overview of 2 nd Generation MIS Advances of Hardware Advances of Algorithms Phantom and Clinical Test Future works

PIERS, Hawaii, Oct First Generation Prototype Operating from 300 MHz to 1 GHz Sequentially Collecting data at 16 trans x 9 rec. sites

PIERS, Hawaii, Oct Challenges Difficulties in 1G Microwave Imaging System – high attenuation of coupling medium (saline) – limited measurement data (only up to 1GHz) – electronics efficiency (serial detection) – positioning (manually) – reconstruction efficiency

PIERS, Hawaii, Oct Overview of 2 nd Generation MIS Main Features 1. Parallel detection scheme 2. Broad operating band (WB source, components and connectors) 3. Automated positioning 4. Low contrast adjustable coupling medium 5. Efficient reconstruction algorithms 6. Streamlined DAQ and data processing

PIERS, Hawaii, Oct Comparisons Performance Matrix Old System New System Receivers 9 15 Operating Frequency 300MHz~1GHz 500MHz~3GHz Detection Mode serial parallel DAQ Speed 16 x 9 x 12 x 7 =12, min 16 x 15 x 20 x 7 =33, min Positioning manual automatic Coupling medium Saline Glycerin/Water mixture

PIERS, Hawaii, Oct Clinical Interface Patient Bed Electronics Illumination Tank Linear Actuator Control Terminal Antenna Array Fluid Management System

PIERS, Hawaii, Oct Operation Animation FMS Patient Bed Reservoir SterilizerMotor 1,11,1 2,22,2

PIERS, Hawaii, Oct Electronics Design

PIERS, Hawaii, Oct Coupling medium Benefits of Glycerin/Water mixture – Low attenuation – Adjustable dielectric properties – Better matching with breast tissue – Safe

PIERS, Hawaii, Oct Measured Properties

PIERS, Hawaii, Oct High Frequency Measurement Crosstalk between transeiver channels Amplitude response and phase error vs. input signal power

PIERS, Hawaii, Oct Reconstruction Framework 2D imaging (3D in developing,3D/2D,3D/3D) Iterative Gauss-Newton reconstruction Freq.-Domain/Time-Domain forward solvers High frequency data implementation Flexible regularization strategies Simultaneous multiple-frequency measurement synthesis

PIERS, Hawaii, Oct Algorithm Optimization Adjoint method (building Jacobian matrix) Dual-mesh (reducing problem size) Log-Magnitude Phase-Unwrapping 2 step reconstruction and conformal mesh technique (enhancing image quality) Spatial filter (stabilizing reconstruction) Pre-scaled system (fine tuning)

PIERS, Hawaii, Oct Adjoint method for dual-mesh Jacobian Matrix - Sensitivity Receiver JsJs Source Only Matrix-vector multiplications are needed, geometric related sparse matrix can be build independent of reconstruction

PIERS, Hawaii, Oct Two-stage Regularization Switching from L-M to Tikhonov Reg. with a priori estimates Step 1 Step 2

PIERS, Hawaii, Oct Phantom Experiment & Results Phantom Corn syrup phantom Diameter: 10cm Inclusion: saline Inclusion diameter: 3cm

PIERS, Hawaii, Oct Clinical Data Patient 1082, Age: 68, AP: 14cm 1 2 Left BreastRight Breast 1100MHz/2stage/conformal

PIERS, Hawaii, Oct Future works 3D data acquisition and reconstruction Multiple spectrum data analysis Faster forward/recon algorithm ->Real time MRI data validation System optimizations (optimal frequency(ies), transeiver num., spatial configurations, backgrd. contrast etc )

PIERS, Hawaii, Oct Questions?

PIERS, Hawaii, Oct Conformal Mesh Shirking the reconstructed unknown close to object based on scout images Saline inclusion Conformed domain & Imaging object Monopole antenna FE / BE region interface FE / BE region interface Concentric FE mesh Saline inclusion Imaging object Monopole antenna array