BMED4962ECSE4962 Introduction to Subsurface Sensing and Imaging Systems

1
BMED-4962/ECSE-4962 Introduction to Subsurface
Sensing and Imaging Systems

• Lecture 28 Grand Summary - II
• Kai Thomenius1 Badri Roysam2
• 1Chief Technologist, Imaging technologies,
• General Electric Global Technology Center
• 2Professor, Rensselaer Polytechnic Institute

Center for Sub-Surface Imaging Sensing
2
Summary of Choices
Influence field
3
Steps for Building Sensing Imaging Systems
Step 1 Identify physical sources of
contrast Intrinsic Object and medium interact
differently with acoustic/EM wave Extrinsic
Contrast agent Step 2 Examine feasibility of
each source of contrast Can generate probe? Does
probe penetrate enough? Can detect returned
signal? Sufficient contrast? Sufficient
signal-to-noise ratio? Step 3 Study application
needs What to measure? Desired resolution, range,
specificity Constraints on power, weight,
toxicity,.. Step 4 Choose SSI
Architecture gtgtgtgtgtgtgt Next!
Time-Lapse (t)
Chemical Species
Dynamics
Things we can image
Physical Properties
Structure
Space (x, y, z)
4
SSI Architectures

• Once we have established the choice of probe,
  contrast source(s) and detectors, we can
  architect the SSI system
• The manner in which probes and detectors are
  positioned and controlled
• Spatially
• Temporally
• Spectrally
• Information extraction algorithms
• Lots of choices
• Luckily, there are common themes across seemingly
  diverse system architectures

5
Transmissive Architectures
Surface
Probe(s)
Object
Medium
Detector(s)

• Simplest architecture
• Only possible when both sides are accessible
• Ideal when scattering is minimal/negligible

Example Microscopy, X-ray radiography
6
Choice of Positioners

• Positioners determine how an x-ray system looks

7
Single-side Access Architectures
Probe(s)
Detector(s)
Surface
Medium
Object
Examples Ultrasound, Ground penetrating radar
General Imaging
Cardiac
OB/Gyn
8
Transmitter and Receiver Patterns
i
j
j
i
Probe Pattern Region in the medium occupied by
the probe wave, in the absence of the target
clutter
Detector Pattern Region in the medium to which
the detector is sensitive, I.e., , if a source
were to exist within this medium, it would be
detected.
9
Localized vs. Tomographic SSI
Localized SSI
Tomographic SSI
1 2 .. I
1 2 .. J
1 2 .. I
1 2 .. J
Probe System
Detector System
Probe System
Detector System
j
j
i
i
ij
ij
Different interaction regions ij are small do
not overlap
Different interaction regions ij are large
do overlap
Interaction region ij region in the medium to
which the signal of detector j is sensitive, when
probe wave i is active
10
Weak Scattering
Probe is scattered from each of the target points
independently. Each scattered wave is
unperturbed by the other points of the target.
Interaction region overlap of transmitter
pattern with receiver pattern.
11
Multi-view Tomography Weak Scattering Case
CT Scanning
12
Strong Scattering
RPIs Electric Impedance Tomograph
13
Multi-view Tomography Strong Scattering Case
Reflection
Transmission
Examples
Electrical Impedance/Resistance Tomography Ground
Penetrating Radar Diffraction Tomography Diffuse
Optical Tomography Elastography
14
Spatial Probing Detection Patterns

• Distributed Probing
• Localized Detection

ii) Localized Probing Distributed
Detection
Applications
Applications
Vision Photography Conventional Microscopy
Fluorescence Scanning Microscopy
15
Localized Probing, Distributed Detection

• Sound velocity profile reconstruction for
  environmental remediation

Multi-photon Microscopy
Good transverse resolution Poor axial resolution
16
Localized Probing, Distributed Detection
Sheet Illumination Localized along a line
Applications
Slit-Lamp Biomicroscope
17
Localized Probe, Localized Detection
Applications
Applications
Ultrasonic Harmonic Imaging
Scanning Confocal Fluorescence Microscopy
18
Point-by-point imaging Scanning
a) Conventional imaging
b) Scanning imaging system
Many choices of scanning systems (point scan,
line-scan, etc.)
19
Example Confocal Microscopy
Confocal Microscope
Point-by-point scanning, localized probing and
localized detection
20
Classification Based on Method of Axial
Localization
Optical Coherence Tomography (OCT)
Ultrasound
Microscopy
21
Interferometric Architectures
Differential Interference Contrast (DIC)
Microscopy
Phase Contrast Microscopy (PCM)
22
Optical Coherence Tomography
Optical Coherence Tomography (OCT)
Macular hole
http//www.neec.com/Glaucoma_OCT.html
23
Classification Based on Spectra
Detectors
Narrow-band
Wide-band
Narrow-band (e.g. laser)
Probes
Wide-band (e.g white light)

• Interesting case
• Multiple narrow-band Probes Detectors can be
  used to build multi- and hyper-spectral imaging
  systems

24
Narrow Probe, Narrow Detector Example
Blue blobs DAPI (DNA Stain) Red clouds Lewis X
(LeX) Green tubes GFAP Tourquoise (tubes and
blobs) Laminin (stains vessels and bulbs)
25
Wideband Probe, Narrow Detector Example

• Estimation of composition of substances with
  known spectral signatures at each position
• Feature detection at each position
• Wavelength ratiometric imaging

Courtesy Luis Jiminez at UPRM
26
Subsurface Intervention Methods
Invasive Access (e.g., surgery)
Surface 1
Clutter
object
Medium
Clutter
object
Medium
Limited Access
Surface 2
Treatment Map
Clutter
object
Beam Intervention
Medium
27
Beam Interventions

• Common Beam Types
• Optical (laser)
• Acoustic (HIFU)
• Radiation (x-ray, gamma ray, proton beam, )
• Common Objectives
• Achieve desired spatial dose distribution
• Full coverage of pathology
• Minimize dose to background
• Motion and deformation compensation
• Conforming the beam to the treatment map
• Integrated approaches to diagnosis, planning,
  execution, and follow up
• Examples
• Laser retinal surgery
• Radiation treatment of intra-ocular tumors
• Radiation treatment of cancer of lung, prostate,

Surface 1
Clutter
object
Medium
Surface 2
Treatment Map
Beam Intervention
28
Beam Interventions (contd)

• Pervasive issues
• Availability/otherwise of implantable marker
  targets
• Possibility/otherwise of labeling the pathology
  itself
• Relating several coordinate systems
• Imagery coordinates of the surface and the
  subsurface
• Marker targets
• Imaging and Beam coordinate systems
• Identification of delicate neighboring structures
  to be avoided
• Achieving adequate spatio-temporal sampling of
  moving 3-D structures
• Minimizing damage from imaging probes

Surface 1
Clutter
object
Medium
Surface 2
Treatment Map
Beam Intervention
29
Example Image-Guided Radiotherapy

• Examples
• Prostate, lung, retina
• 4D Systems needed
• Visualization
• Segmentation
• Planning Tools
• Delivery Tools
• Multiple motions and deformations
• Intra-fractional motion (lung, abdomen)
• Inter-fractional motion (organ deformations
  change)
• Motion of beam delivery system

30
Current Practice
Linac
3D CT

• Patient in restraint
• 3D CT du jour taken
• Patient-specific fractional dose calculations
  adjustments
• Transfer patient to Linac for dose delivery
• Conformal radio-therapy
• Intensity-modulated radio-therapy

Multi-leaf Collimator
31
Challenges Appearance Shape Variability
Bladder
Prostate
Rectum
Intra-patient variability, 10 days apart
Inter-patient variability
32
Engineering Implications

• Problem
• Given a new daily CT scan, just prior to a
  radiation fraction, segment the prostate,
  bladder, rectum, etc. for IGT and conformal
  avoidance
• Approach
• Construct effective low-dimensional models for
  intra- and inter-patient variability
• Robust fitting of models to image data
• Update the treatment plan
• Current Limitations
• Timeliness and verifiable accuracy of the model
  fitting
• Expert manual segmentation takes 15-45 mins

33
Mathematical Modeling of Variations
Describe variations of prostate, bladder, and
anterior rectal wall
Courtesy Prof. Richard Radke
34
Fitting the Model to a Image du jour
initial guess
35
Model Fitting

• Automatic fitting is 5 - 10X faster than manual
  contouring
• Comparable Accuracy
• Can be edited as needed
• Will enable accurate radiation delivery on day
  of treatment

36
Towards 4D Systems
4D CT Imaging
Optimization Planning
Linac
Note Single cycle looped!
37
Lung Tumor Motion
Patient 1
Peak to Peak 1.3 cm
Note Single cycle looped!
38
Challenges and Opportunities from 4D

• Artifact-free 4D Imaging
• Need for motion-compensated reconstruction
• Effective 4D visualization tools
• Effective and timely segmentation
• Variability
• Low contrast, poor edges
• Massive data volume!
• Verification/visualization tools
• Relating external markers
• 4D Dose calculation tools
• 4D Dose delivery systems
• Deformable registration with change intelligence

4D Image Browser
39
Where is this field headed?

• Accelerated building of new better systems
• Exploit commonalities and lessons learned from
  diverse existing SSI systems
• Biggest issues going forward
• Measure chosen properties with high specificity
• Molecular / chemically-specific imaging
• Develop more and better contrast agents
• Fusion of multiple modalities
• Information extraction algorithms
• Image understanding systems
• Computer-aided diagnostics
• Change detection
• Image-guided interventions
• Fall 2007 Course
• Cell and Tissue Image Analysis

40
Acknowledgments

• Many slides adapted from Prof. Bahaa Saleh at BU
  Prof. George Chen at Harvard/MGH

41
Instructor Contact Information

• Badri Roysam
• Professor of Electrical, Computer, Systems
  Engineering
• Office JEC 7010
• Rensselaer Polytechnic Institute
• 110, 8th Street, Troy, New York 12180
• Phone [phone]
• Fax [phone]
• Email roysam_at_ecse.rpi.edu
• Website http//www.ecse.rpi.edu/roysabm
• NetMeeting ID (for off-campus students)
  128.113.61.80
• Secretary Laraine, JEC 7012, [phone],
  michal_at_.rpi.edu

42
Instructor Contact Information

• Kai E Thomenius
• Chief Technologist, Ultrasound Biomedical
• Office KW-C300A
• GE Global Research
• Imaging Technologies
• Niskayuna, New York 12309
• Phone [phone]
• Fax [phone]
• Email thomeniu_at_crd.ge.com, thomenius_at_ecse.rpi.edu
  
• Secretary TBD