See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/242271220

Digital Fluoroscopic Imaging: Digital Fluoroscopic Imaging: Acquisition,

Processing & Display Acquisition, Processing & Display

Article

CITATION READS

1 340

2 authors, including:

James Anthony Seibert

University of California, Davis
233 PUBLICATIONS   5,561 CITATIONS   

SEE PROFILE

All content following this page was uploaded by James Anthony Seibert on 21 November 2015.

The user has requested enhancement of the downloaded file.

 Digital Fluoroscopic Imaging:
Acquisition, Processing & Display

J. Anthony Seibert, Ph.D.

University of California Davis
Medical Center
Sacramento, California

Outline of presentation

• Introduction to digital fluoroscopy

• Digital fluoroscopy components
• Analog and digital image characteristics
• Image digitization (quantization/sampling)
• Image processing
• Summary

1
History of digital fluoroscopic imaging

• ……. mid 1970’s

– Modified II/TV system with “fast” ADC
– Temporal and energy subtraction methods

• ……. 1980’s
– Clinical DSA angiography systems
– Qualitative and quantitative improvements
– Image processing advances
– Temporal and recursive filtering

History of digital fluoroscopic imaging

• ……. 1990’s
– Quantitative correction of image data
– Rotational fluoroscopic imaging
– Micro-
Micro-fluoroscopic imaging capabilities
– CT fluoroscopy (using fan-
fan-beam scanners)
– Cone-
Cone-beam CT reconstructions

• ……. 2000 - present

– Introduction of real-
real-time flat-
flat-panel detectors

2
Why digital fluoroscopy / fluorography?

• Low dose fluoroscopic imaging

(digital averaging, last frame hold)

• Pulsed fluoroscopy and variable frame rate

• DSA and non-

non-subtraction acquisition and display

• Digital image processing and quantitation

• Image distribution and archiving, PACS

• Introduction to digital fluoroscopy

• Digital fluoroscopy components
• Analog and digital image characteristics
• Image digitization (quantization/sampling)
• Image processing
• Summary

3
 Fluoroscopic Acquisition Components
TV Camera
Side View: C arm System

C-Arm Image
Apparatus Intensifier TV Monitor

Peripherals
Cine Camera
Photospot Camera
Collimator Spot Film Device
Digital Photospot
DSA System

X-ray Tube

Image Intensifier - TV subsystem

Input phosphor Photocathode (- ) Housing
Aperture
Focusing (Iris) TV camera
electrodes
Evacuated Lens optics
Insert and mirror
e-
assembly
Anode (+)

e-

Output
Video or CCD
phosphor camera to ADC
to Digital Image
X-rays in
~25,000 Volts
acceleration
Grid -
e- e- e-
Light out → Recorder
e
-
e- e- e-
- ZnCdS:Ag
ZnCdS:Ag
ee output phosphor
CsI input SbCs3
phosphor photocathode Electrons → Light

X-rays → Light → Electrons ~5000 X amplification

4
Structured Phosphor: Cesium Iodide (CsI)
Crystals grow in long columns that act as light pipes

CsI

Light Pipe (Optical LSF

Fiber)

TV camera readout and output video

5
 TV camera specifications

• Low resolution:
– 525 line, interlaced, 30 Hz (RS-
(RS-170)

• High resolution:
– 1023 - 1049 line, interlaced, 30 Hz (RS-
(RS-343)

• Highest resolution
– 2048 line systems

• Progressive scan a must for short pulse-

pulse-width
digital applications

II-TV digital systems

• Two decades+ of availability

• Video signal is convenient for digitization
II’s: ↑SNR
• Low noise performance of II’s:
• Well-
Well-developed capabilities
– IA, DSA, digital photospot
– Rotational CT

• CCD camera implementations

• II is Big and bulky; image distortions prevalent

6
Flat-panel Fluoroscopy / Fluorography

• Based upon TFT charge storage and

readout technology

• Thin-Film-Transistor arrays
– Proven with radiography applications
– Just becoming available in fluoroscopy
• CsI scintillator systems (indirect conversion)
• a-Se systems (direct conversion)

Photodetector: a - Si TFT active matrix array

Photodiode:
Scintillator Light to electronic signal

X-rays to light Amplifiers – Signal out

TFT: Storage and readout

7
 Amorphous Silicon
TFT active matrix array
Gate
Amplifiers – Signal out

G1 switches
Active
Area Thin-
Thin-Film
Transistor
Dead G2
Zone Storage
Fill Factor = Active area ÷ (Active area + Dead Zone) Capacitor
G3
Large pixels: ~ 70%
Small pixels: ~ 30 % Charge
Collector
D1 D2 D3 Electrode
CR1 CR2 CR3
Analog to
Data lines Charge Digital
Amplifiers Converters

Amorphous Silicon
TFT active matrix array
Amplifiers – Signal out

G1 Expose to x-
x-rays

G2

Store the charge

G3

Active Readout
Activate gates
Amplify charge
Convert to Digital

8
 Cross section of detector:
a-Si TFT/ CsI phosphor

X-ray

Structured X-
X-ray
Light phosphor (CsI)

Source
Gate S G D +
Drain
TFT Adjacent gate line
Charge
Storage capacitor
Photodiode

X-rays to light to electrons to electronic signal:

Indirect digital detector

Flat panel vs. Image Intensifier

Flat
panel

II

Field coverage / size advantage to flat panel Image distortion advantage to flat panel

9
 Output Total over-framing Digital
phosphor sampling
image Maximum horizontal framing matrix

Maximum vertical framing

Framing of digital matrix:

FOV vs. spatial resolution vs. x-
x-ray utilization

framing FOV spatial resolution % recorded area

4:3 aspect ratio 23 cm nominal 512 × 480 matrix (% digital area used)
input diameter 1023 x 960 matrix

Maximum vertical 22 cm 0.46 mm 100 %

framing
1.09 lp / mm (41%)

Maximum horizontal 19 cm 0.43 mm 74%

framing
1.16 lp / mm (78%)

Maximum 15 cm 0.33 mm 61%

overframing*
overframing*
1.5 lp / mm (100%)

10
 Flat-panel fluoro detector:
efficient use of x-
x-ray detector / x-
x-ray field

Flat panel vs. Image Intensifier

II conversion gain: ~5000:1

-- Electron acceleration flux gain
-- Minification gain

FOV variability (mag

(mag mode) and sampling advantage to II
Gain / noise advantage to II

11
 Flat panel vs. Image Intensifier
• Electronic noise limits flat-
flat-panel amplification
(1-5 µR/frame)
gain at fluoro levels (1-

• Pixel binning (2x2, 3x3) lowers noise; “mag

“mag--
mode” equivalent changes pixel bin sampling

• Low noise TFT’s are being produced (low

yield); variable gain technologies are needed

• Prediction:
– II’s will likely go the way of the CRT…….

Interventional system digital hardware architecture

Display calibration

X-ray
system Arithmetic
Logic Unit
Analog
ADC
signal
Array DAC
Processor
Micro-
Processor
Display
Peripheral Processor
equipment
Video
Patient memory:
Image Workstation
monitor 64 MB to Digital Local Image Modality Interface
512 MB Disk Array Cache

DICOM HL-7
Interface Interface
System information (kV, mA, etc) Modality Worklist

Images Patient / Images

(XA objects)
PACS reconciliation

12
• Introduction to digital fluoroscopy
• Digital fluoroscopy components
• Analog and digital image characteristics
• Image digitization (quantization/sampling)
• Image processing
• Summary

Fluoroscopic Analog Image

• Continuous brightness variation
corresponding to differential x-
x-ray
transmission of the object

Uniformly irradiated
II with lead disk

13
 Conventional raster scan: RS-170
4:3 aspect ratio, 525 lines, 483 active

700 mV

voltage
image
height:
0 mV
3 39 µsec

-300 mV
sync signals determine
image location

image width: 4 33 msec Single horizontal video line

Digital Image Requirements

• Contrast resolution
– Ability to differentiate subtle differences in
x-ray attenuation (integer numbers)

• Spatial Resolution
– Ability to discriminate and detect small objects
(typically of high attenuation)

14
 Digital Image Matrix
700 mV

voltage

0 mV
39 µsec

-300 mV

Rows and columns define

useful matrix size across Single horizontal video line
active field of view. For
RS-170 standard, this 23 68 145 190 238 244 249 150 38 31 30 35 43 159 232 241 239 182 131 33
corresponds to
~480 x 480. Digitized video signal corresponding to horizontal line

A better match now often available is 640x480 (VGA)

Digital Acquisition Process

• Conversion of continuous, analog signal

into discrete digital signal

• Digitization
– Sampling (temporal / spatial)
– Quantization (conversion to integer value)

15
 Digital Image Characteristics

• Advantages
– Separation of acquisition and display
– Image processing applications
– Electronic display, distribution, archive

• Disadvantages: noise and data loss

– Quantization
– Sampling
– Electronic (shot)

Consequences of digitization
• Negative:
– Loss of spatial resolution

– Loss of contrast fidelity

– Aliasing of high frequency signals

• Positive:
– Image processing and manipulation

– Electronic distribution, display and archive

– Quantitative data analysis

16
• Introduction to digital fluoroscopy
• Digital fluoroscopy components
• Analog and digital image characteristics
• Image digitization (quantization/sampling)
• Image processing
• Summary

Acquisition Processing Display

Computer
Fluoro unit Softcopy
hardware
Peripheral ADC and DAC CRT or
components software
FlatPanel
algorithms
Analog to Digital to
digital analog
conversion conversion

RAID-
RAID-5
online

Storage / Archive

17
 Analog to Digital Conversion:
Digitization

• Sampling: measuring the analog signal at

discrete time intervals
– @ 2x frequency of video bandwidth

• Quantization: converting the amplitude of

the sampled signal into a digital number
– Determined by the number of ADC bits

Sampling

• Signal averaging within detector element (del)

area = ∆x × ∆y

• Cutoff sampling frequency = 1 / ∆x

2∆x
• Nyquist frequency = 1 / 2∆

• Minimum resolvable object size (mm)

= 1 / (2 × Nyquist frequency)

18
 Sampling: discrete spatial measurement
infinite bits, 3 samples / line

Input

Sampling aperture Sampling points relative error

infinite bits, 7 samples / line

Input

relative error
Sampling aperture Sampling points

Resolution and digital sampling

Detector
Element,
“DEL” MTF of pixel (sampling) aperture
1000 µm 500 µm 200 µm
1

0.8
Modulation

0.6

0.4

0.2

0
0 1 2 3 4 5 6
Frequency (lp/mm)

Cutoff frequency = 1 / ∆x
Sampling Sampling
pitch aperture MTF of sampling aperture

1/2∆x, when pitch = aperture

Nyquist frequency = 1/2∆

19
 Phase Effects
Input signal equal to Nyquist frequency

in phase 180° phase shift

Bar pattern

pixel matrix

good signal modulation no signal modulation

sampled output signal

Aliasing: Insufficient sampling

Pixel Sampling

Low frequency

> 2 samples/ cycle

High frequency

Assigned (aliased) frequency < 2 samples/ cycle

20
 Aliasing effects:
Input signal frequency, f > Nyquist frequency, fN

input f = 1.5 fN input f = 2.0 fN

output f = 0.5 fN output f = 1.0 fN

Aliasing
Input signal frequency spectrum, fin

Input signal BW
Sampling BW
amplitude

-fN 0 fN fS 2fS
Frequency

Higher frequency overlapping sidebands

reflect about f to lower spatial frequencies
N

21
 How important is aliasing?
• Most objects have relatively low contrast

• High frequency noise lowers DQE(f) in the clinically

useful frequency range

• Clinical impact is probably minimal, except with

stationary anti-
anti-scatter grids and sub-
sub-sampled images

• Image size reduction can cause aliasing

– Subsampling retains high frequencies, violating Nyquist limit

Resolution and image blur

• Sources of blur
– Light spread in phosphor
– Geometric blurring: magnification / focal spot
– Pixel aperture of detector and display

• Goal: match detector element size with

anticipated spread to optimize sampling
process

22
 FOV and digital sampling
12 cm
24 cm
12 cm

1k x 1k: 120 µm

24 cm
~4 lp/mm

1k x 1k : 240 µm
~2 lp/mm

2 k x 2k: 120 µm
~4 lp/mm

Sampling and spatial resolution

1000 samples 500 samples 250 samples 125 samples

23
 Quantization: conversion to digital number
2 bits (4 discrete levels) and infinite sampling
3

input signal ramp quantized output relative error

3 bits (8 discrete levels) and infinite sampling

7
6
5
4
3
2
1
0

input signal ramp quantized output relative error

350 mV
Reference
voltage, V 710 mV
3 bit Analog to
Video
Digital Converter
input Comparators
R
+
7V -
8
R Digital
+ Output
6V
-
8
R MSB
Successive +
fractional 5V -
0
voltage at each 8
R
comparator + Priority
4V Encoder 1 8 discrete output values
-
8 Logic
R
+
3V 1
-
8
R LSB
+
2V
-
8
R
+
V -
8

24
 Quantization

• Threshold to next level is ½ step size

• Larger # bits provide better accuracy

• Quantization noise causes “contouring”

• Typical bit depths:

– Fluoroscopy: 8 bits
– Angiography: 10 – 12 bits
– CR / DR: 10 – 14 bits

Quantization Effects

8 bits 4 bits 3 bits 2 bits

“Contouring” is a problem in areas slowly varying in contrast.

25
 Dynamic range considerations

• Maximum usable signal determined by:

– Saturation of detector (TV camera)
– Light aperture (determine entrance exposure)
– Analog to digital converter (ADC)

• Minimum usable signal determined by:

– Number of bits in ADC
– Quantum noise bits graylevels
– System noise 8 256
– Electronics 10 1024
12 4096
14 16384

Resolution and Image Size

• 2 bytes / pixel uncompressed for digital fluoro

• 512 x 512 matrix (1/2 MB/image, 15 MB/s*)
• 1024 x 1024 matrix ( 1 MB/image, 30 MB/s*)
• 2048 x 2048 matrix (4 MB/image, 120 MB/s*)
– *At 30 frame/s acquisition rate

• Overall storage requirement / Interventional

Angiography study: 200 to 1000 MB
– Image compression; selected key images

26
 Digital Image Display

• Digital to Analog Converter (DAC)

• Estimate of original analog signal amplitude

• Image fidelity determined by

– Frequency response (bandwidth)
– Number of converter bits (usually 8 or 10 bits)
– Image refresh rate (# updates / sec)

Digital to Analog Converter: DAC

Reference voltage =710 mV

355 mV
MSB Ref / 2
1 178 mV
Ref / 4
0 89 mV
Ref / 8 Voltage
0 44 mV adder
Ref / 16 432 mV
Digital 1 22 mV
input Ref / 32
1 11 mV
Ref / 64
Voltage
1 out
6 mV
Ref / 128
0 3 mV video
Ref / 256 synchronization
0 electronics
LSB source gate drain
Transistor (switch)

27
 MSB
0 0 0 0 0
0 1 1 0 0
000 0
0 0 0
0 1 1 1
0 0 0 0 1 1 1 1
0 0 Bit depth
0 0 0 0 0 0 1 1 1 1 1
0 0 0 0 0 0 1 1 1 1 1 1
0 0 0 0 0 1 1 1 1 1 1 1
0 0 LSB
0 0 0 1 1 1 1 1 1 1 1
0 0 0
0 0 0 0 0 1 1 1 1 1 1 1 1 1
0
0 0 0 Numerical
0 0 0 0 1 1 1 1 1 1 1 1 1
0 0 0 representation
0 0 0 0 0 1 1 1 1 1 1 1 1 1
0 0 0
0 0 0 0 0 1 1 1 1 1 1 1 1 1
0 0 0
0 0 0 0 0 1 1 1 1 1 1 1 1 1
y
0 0 0
0 0 0 0 0 1 1 1 1 1 1 1 1 1
0 0 0
0 0 0 0 1 1 1 1 1 1 1 1 1
0 0
0 0 0 1 1 1 1 1 1 1 1 1
0 0 1 1 1 1 1 1 1 1 1
Image bit 0
0 1 1 1 1 1 1 1 1 1
planes
x Linear DAC

Image
representation
digital number 0 255
appearance: dark bright

Display adjustments

• LUT: Look up table

– Dynamic conversion of digital data through a
translation table

– Non-
Non-destructive variation of image brightness and
contrast

– Reduced display dynamic range requires

compression of image range data (to 8 bits)

28
 Display of digital data
Look-up-table 8 bit output
(LUT)
255 255
Logarithmic
transform Linear
transform

WL WW

Exponential
transform

0 0

4095 2048 0
8 bit output
12 bit input display range

Grayscale Processing

• Look-up-table Transformation
– Window (contrast, c) and level (brightness, b)

Iout (x,y) = c × Iin (x,y) + b

• Histogram equalization
– Redistribution of grayscale frequencies over
the full output range

29
 Window Width / Window Level

Contrast Resolution
• Fluoroscopic Speed
– Dependent on light-
light-limiting aperture (f-
(f-stop)
– Variable for digital flat-
flat-panel detectors
– ? secondary quantum sink at higher frequencies

• Electronic noise
– shot noise, dark noise, fixed pattern noise

• Structured noise
– Anatomy, overlying objects

• “Useful” dynamic range

– minimum detectable contrast with additive noise

30
 Low Contrast Resolution

Temporal
No Temporal
Averaging
Averaging
4 frames

1 mR 0.1 mR 0.01 mR
Image subtraction low contrast phantom

Noise Sources
• Digital acquisition: SNR-
SNR-limited detection
– quantum mottle and secondary quantum sink
– fixed pattern (equipment) structured noise
– electronic and shot noise
– digitization: sampling and quantization noise
– anatomic (patient) noise

• Imaging system should always function in

x-ray quantum-
quantum-limited range

– With II/TV, gain is sufficient

– With flat-
flat-panel, electronic noise is limiting factor

31
• Introduction to digital fluoroscopy
• Digital fluoroscopy components
• Analog and digital image characteristics
• Image digitization (quantization/sampling)
• Image processing
• Summary

Image Processing

• Reduce radiation dose through image averaging

• Enhance conspicuity of clinical information

• Provide quantitative capabilities

• Optimize image display on monitors

32
 Image Processing Operations

• Point
– Pixel to pixel manipulation

• Local
– Small pixel area to pixel manipulation

• Global
– Large pixel area to pixel manipulation

Temporal Averaging

Iout(x,y) = N Σ Ii(x,y)

• Reduces noise
fluctuations by N 0.5

• Increases SNR

• Decreases temporal
resolution

33
 Image Subtraction (DSA)

• Pixel by pixel operation:

out(i) (x,y) = Im(x,y) – Ii(x,y) + offset

Iout(

• Time dependent log difference signal

• Window / level contrast enhancement

Logarithmic amplification
• Linearizes exponential x-
x-ray attenuation

• Difference signal is independent of incident x-

x-ray flux

− µ bg tbg
Mask image: I m = N 0e

− µ vessel t vessel − µ bg tbg

Contrast image: I c = N 0e

Subtracted image: I s = ln( I m ) − ln( I c ) = µ vessel tvessel

34
 Linear to Log LUT
10 bit to 8 bit
250
Output Digital Number

200

150

100

50

0
0 200 400 600 800 1,000
Input Digital Number

Digital Subtraction Angiography

• Temporal subtraction sequence

– First implemented mid 1970’s

• Eliminate static anatomy

– Increase conspicuity

• Isolate and enhance contrast

– Lower contrast “load”

35
Digital
Fluoro

Mask Contrast Image Subtraction

Image
Contrast agent

Time-dependent subtraction (DSA)

Subtracted
images

36
 DSA examples

DSA image manipulation / quantitation

• Pixel shifting
(correct for
misregistration)

• Add anatomy
(visualize
landmarks)

• Measurements /
densitometry

37
 Matched Filtration
C(t)
Cmax
Cavg

time
Average ROI signal in image i.

ki = C(t) - Cavg
+

-
time

Image sequence and ROI Image weighting coefficients, ki

Matched Filtration

k6 × I6(x,y)

k5 × I5(x,y)

k4 × I4(x,y)

k3 × I3(x,y) +
k2 × I2(x,y)

k1 × I1(x,y) Single averaged output image

High SNR at ROI position

Scaling factor ki

38
 Image comparisons

Contrast Mask subtract

Image Image

Matched filter Selective dye

Image Image

Recursive filtration
• Digital image buffer adds a fraction, k, of the incoming
image to the previous output image; temporal averaging
with exponentially decreasing signal

Iout(n) = k Iin(n) + (1-

(1-k) Iin(n- (1-k)2 k Iin(n-
(n-1) + (1- (n-2) +….

Iin(x,y) ×k

+ Iout(x,y)

× (1-
(1-k) feedback

Image
Memory
Buffer

39
 Image Processing Operations

• Point
– Pixel to pixel manipulation

• Local
– Small pixel area to pixel manipulation

• Global
– Large pixel area to pixel manipulation

Spatial Filtration

• Low pass (smoothing)

• High pass (edges)

• Bandpass (edge enhancement)

• “Real-
“Real-time” filtration uses special hardware
and filter kernels of small spatial extent

40
 Convolution
• Pixel by pixel multiplication and addition
of filter kernel with image:

( N −1)/ 2
I out ( x ) = ∑ g(i ) I
i =− ( N −1)/ 2
in ( x + i)

I out ( x ) = g ( −1) × I in ( x − 1) + g( 0 ) × I in ( x ) + g(1) × I in ( x + 1)

I out ( x ) = g ( x ) * I in ( x )

Point sampling aperture:

frequency response

MTF
LSF
width: ∆ x ~ 0 1
0.8
Modulation

0.6
0.4
height: 0.2
1/ ∆x
0
-0.2
0 0.5 1 1.5 2 2.5 3
Frequency
(units of 1/ ∆x)

41
 Finite sampling aperture:
frequency response

MTF

1
sinc (x)
Single element LSF 0.8

Modulation
width: ∆x 0.6
0.4
0.2
height: 0
1/ ∆x
-0.2
0 0.5 1 1.5 2 2.5 3
Frequency
fN fS (units of 1/ ∆x)

Filter kernels

Single element LSF Frequency response

width: ∆x 1 and 3 element equal weight kernel
1 MTF
1 element
height: 0.8
1/ ∆x
Modulation

0.6
0.4
3 element
0.2
Three element LSF
width: 3 ∆x 0
-0.2
height:
1/(3∆x)
1/(3∆ 0 0.2 0.4 0.6 0.8 1 1.2
Frequency
Units of 1/ ∆x

42
 Low pass filtration – smoothing

• Convolve “normalized” filter kernel with image

• Reduces high frequency signals

• Reduces noise variations

• Reduces resolution

2D Low pass filter kernel

• Convolve “normalized” filter kernel with image
Input Output

1 1 1 10 10 10 1 1 4 7 10 10
1 1 1 1 1 1 10 10 10 1 1 4 7 10 10
1 1 1 1 1 1 10 10 10 1 1 4 7 10 10
**
1 1 1 1 1 1 10 10 10 1 1 4 7 10 10
1 1 1 10 10 10 1 1 4 7 10 10
÷9
1 1 1 10 10 10 1 1 4 7 10 10

Profile before Profile after

43
 Variable weight low-pass filter kernel

Variable weight kernel

Frequency response
width: variable weight kernel
∆x height:
1
0.6 / ∆x
0.8 Combined response
0.2 / ∆x

Modulation
0.6
0.4
0.2
Break into parts:
0
-0.2
+
0 0.2 0.4 0.6 0.8 1 1.2
Frequency
1/∆x
Units of 1/∆

High pass filtration

• Low pass filtered signal subtracted from

original signal

• High frequencies (edges) remain in image

• Noise is increased

44
 High-pass filter kernel
Single kernel LSF
Frequency response
high-
high-pass filter
Highpass LSF 1
Difference
0.8

Modulation
+ 0.6
-
0.4
-
0.2
Lowpass LSF 0
-0.2
0 0.2 0.4 0.6 0.8 1 1.2
Frequency
1/∆x
Units of 1/∆

2D high pass filter kernel

•Convolve “normalized” filter kernel with image
Input Output

1 1 1 10 10 10 1 1 -26 35 10 10
-1 -1 -1 1 1 1 10 10 10 1 1 -26 35 10 10
-1 9 -1 1 1 1 10 10 10 1 1 -26 35 10 10
**
-1 -1 -1 1 1 1 10 10 10 1 1 -26 35 10 10
1 1 1 10 10 10 1 1 -26 35 10 10
1 1 1 10 10 10 1 1 -26 35 10 10

Profile before Profile after

45
 Example filtered images

Unfiltered Edge enhanced Smoothed

Image Processing Operations

• Point
– Pixel to pixel manipulation

• Local
– Small pixel area to pixel manipulation

• Global
– Large pixel area to pixel manipulation

46
 Global Image Processing

• Frequency domain processing

– Fourier transform of kernel and image
– Convolution → Multiplication
– More efficient for convolution kernels > 9x9

• Inverse filtering (deconvolution)

– e.g., veiling glare, scatter corrections

• Image translation, rotation and warping

– Correction of misregistration artifacts, pincushion
distortion, vignetting, non-
non-uniform detector response

Inverse filtering
• 2D – FT methods:
– Measure PSF
– Generate FT of inverse filter
– Multiply by 2D-
2D-FT of image
– Re-
Re-inverse transform
X-ray scatter PSF and inverse filter:

47
 Quantitative Algorithms

• Stenosis sizing: length, area, densitometry

• Distance measurements
• Density – time curve analysis
• Perfusion – functional studies
• Relative flow and volumetric assessment
• Vessel tracking
• CT with cone-
cone-beam reconstruction

Limits to Quantitation

• Non-
Non-linear / non-
non-stationary degradations
– Beam Hardening
– Scatter
– Veiling Glare
– Non-
Non-uniform bolus / diffusion

• Geometric effects
– Pincushion distortion
– Vignetting
– Rotational accuracy (CT)

48
 Summary

• Digital imaging is an essential part of

fluoroscopic and angiographic systems

• Limitations and advantages of fluoro digital

acquisition and processing must be
understood for maximum utilization

• DICOM standards are a must for the

integration of digital fluoroscopy in the clinical
environment and PACS

Summary

• Fluoroscopic / Fluorographic
image processing can provide

– Significant improvement of image quality

– Reduced dose (radiation and contrast)
– Enhanced image details
– DSA, roadmapping,
roadmapping, quantitative densitometry
– Functional imaging, cone-
cone-beam fluoro CT

49
 References / further information
• Seibert JA. Digital Image Processing Basics, in A
Categorical Course in Physics: Physical and
Technical Aspects of Interventional Radiology, Balter
S and Shope T, Eds,
Eds, RSNA Publications, 1995

• Bushberg et.al. Essential physics of Medical

Imaging, Lippincott,
Lippincott, Williams & Wilkens,
Wilkens, Philadelphia,
2002

• Balter S, Chan R, Shope T. Intravascular

Brachytherapy / Fluoroscopically Guided
Interventions, Medical Physics Monograph #28,
Medical Physics Publishing, Madison, WI, 2002.

……The End……

View publication stats

50