Digital image self-adaptive acquisition in medical x-ray imaging
1. Digital image self-
adaptive acquisition in
medical x-ray imaging
Bao Jie, Gao Jun et.al.
Lab on Image Information Processing
Hefei University of Technology , China
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 1
2. Content
What is X-ray fluoroscopy system and
digital acquisition system
The principle and implementation of self-
adaptive digital acquisition
Experiment and Conclusions
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 2
3. 1. What is X-ray fluoroscopy
system and digital acquisition
system?
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 3
4. What’s X-ray fluoroscopy system?
X-ray fluoroscopy system is a system for
medical diagnosing that can render image of
the body of patient by convert X-ray which
pass through and attenuated by the body into
visible light and record it on film or other
media. It’s a very common method for
examination in hospitals.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 4
5. Construction of X-ray fluoroscopy system
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 5
6. Why study the digital acquisition of
X-ray fluoroscopy system?(1)
The digitalization of x-ray imaging is very
important for PACS (Picture Archiving and
Communication system); high-quality digital
X-ray medical images are indispensable for
PACS data source.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 6
7. Why study the digital acquisition of
X-ray fluoroscopy system?(2)
There are three ways to digitalize x-ray imaging
Computed Radiography (CR)
Digital Radiography (DR)
Video digital acquisition.
Advantages of video digital acquisition : ability to
see dynamic change of organs, device simplicity,
operating convenience, and low-cost
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 7
8. The main difficulties in video
digital acquisition
x-ray fluoroscopy image detection noise
and digital quantum noise
Adjusting imaging contrast and
resolution
Device background signal
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 8
9. How to deal with them?
choose a appropriate working point automatically
and suppressed background signal by software
Improve hardware quality of x-ray imaging
system
Choose grabber board with high quantization
precision
Voltage stabilization and electromagnetic
shielding
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 9
10. Digital video processing system(1)
Enhancement
Annotation
Display
Diagnose
manual
Information
navigation
x-ray video NSPgrabber board
Host
Report
Aided
diagnose
Aided ment
treat
Control Archiving and backup
Query and management
PACS
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 10
11. Digital video processing system(2)
Host should analyze the input signal while sampling and
quantization to adjust grabber board setting for valid
signal to utilize the dynamic range sufficiently, and to make
device working in linear range.
The grabber board we used is NSP (Native Signal Process)
frame-grabber board DT3153-LS, it can adjust reference,
offset, gain, black level and white level by software, which
make it possible for self-adaptive acquisition by software.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 11
12. 2. The principle and
implementation of self-adaptive
digital acquisition
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 12
13. Self-adaptive digital acquisition
To resolve problems brought forward in
section 1, we use digital subtraction technique
to realize background removing for self-
adaptive acquisition, and monitor the
dynamic range of image valid region to search
for the best acquisition working point
automatically.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 13
15. 3.1Valid region recognition
The acquired image is not entirely valid.
Generally speaking, the valid region is a circle.
We should only count on valid region while
removing background and analyzing the
image feature to adjust acquisition
parameters, so we must recognize the valid
region at first.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 15
16. Valid observe region
(a) Whole valid observe
region. White line is (b) Valid observe region
detected region edge by with occlusion
improved seed algorithm.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 16
17. Valid region detection algorithm (1)
1. Compute the histogram of left and right narrow
edges of the image, the gray-level corresponding
to histogram peak value is the gray-level of
invalid region.
2. Perform median filtering to remove noise.
3. Grow region using classical seed growing
algorithm starting from any invalid point.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 17
18. Valid region detection algorithm (2)
4. Generate initial mask(bilevel ) image of valid
region. Perform Sobel operator to this image to
extract its edge.
5. Detect circle by general Hough transform; get the
radius and the center of the circle.
6. Generate valid region mask using result of step 5.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 18
19. 3.2 Background removing
Nonuniform background will affect image quality
and the computing of image characteristic to
adjust acquisition parameters.
So a digital subtraction will remove background
signal while keep the validity of information.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 19
20. Background removing algorithm
1. Acquire and save device background signal (I1)
when device is idle.
2. Acquire images to be observed (I2).
3. Perform image operation in valid region :
I3=I1-I2 ; I4=NOT I3;
4. I4 is the image signal removed of background.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 20
21. 3.3 Setting acquisition working point
After above-mentioned pre-processing, we will
adjust black level, white level, gain, reference
and offset automatically based on histogram
analysis of image valid region to obtain best
acquisition quality.
Black level = - offset
White level = reference / gain -offset
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 21
22. Meaning of offset, gain and reference
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 22
23. Meaning of black level and white level
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 23
24. Working point setting rule
Decreasing offset will shift image to light zone,
increasing offset will shift image to dark zone,
namely offset behaves as brightness adjusting;
decreasing reference will compress image to
light zone, increasing reference will compress
image to dark zone, namely reference behaves
as contrast adjusting.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 24
25. Dynamic range analysis of valid region
Analyze the proportion of dark zone and light zone in the
histogram of image valid region, the aim of adjusting is
to keep proper proportion of dark zone and light zone
for best image acquisition performance.
Setting brightness at first to ensure dark zone isn't too much
then setting contrast( that is, properly setting white level
by adjusting reference).
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 25
26. Self-adaptive
acquisition parameters setting
B e g in
R e a d in e s s ju d g m e n t
I m a g e a c q u is itio n
V a lid re g io n re c o g n itio n , b a c k g ro u n d re m o v in g
V a lid re g io n a n a ly s is
D y n a m ic ra n g e
T o o d a rk T o o b rig h t
W h ile ( d a r k z o n e is to o m u c h & & o ff s e t W h ile ( d a r k z o n e is to o f e w & & o ff s e t is
is n o t o u t o f lo w e r b o u n d ) not out of upper bound)
D e c r e a s e o ff s e t I n c r e a s e o ff s e t
I m a g e a c q u is itio n , b a c k g ro u n d I m a g e a c q u is itio n , b a c k g ro u n d re m o v in g
re m o v in g
V a lid re g io n a n a ly s is V a lid re g io n a n a ly s is
L ig h t z o n e a n a ly s is
L ess M o re
W h ile ( lig h t z o n e is to o m u c h & & W h ile ( lig h t z o n e is to o fe w & &
re fe re n c e is n o t o u t o f u p p e r b o u n d ) re fe re n c e is n o t o u t o f lo w e r b o u n d )
D e c re a se re fe re n c e In c re a se re fe re n c e
I m a g e a c q u is itio n , b a c k g ro u n d I m a g e a c q u is itio n , b a c k g ro u n d re m o v in g
re m o v in g
V a lid re g io n a n a ly s is V a lid re g io n a n a ly s is
End
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 26
27. Universal acquisition parameters choosing
It's very inefficient and unnecessary to setting best
working point every time we take fluoroscopy.
In practice, expert judgment and adjusting is used
to choose universal acquisition parameters.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 27
28. 3. Experiment and Conclusions
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 28
29. Run interface of self-adaptive
acquisition module
Run interface of
self-adaptive
acquisition
module in
ImagePro™
implemented by
Visual C++6.0
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 29
30. Valid region detection
Original Image Valid region mask image Sobel edge-detect image
integrated valid region Rim of Valid Region by improved
algorithm
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 30
31. Background removing
device acquired image after
background image with removing
signal nonuniform device
background background
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 31
32. self-adaptive adjusting(1):
Acquired image before
self-adaptive adjusting.
Black level=0V, white
level =0.7V, offset=0V,
gain=1, reference
=0.7V
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 32
33. self-adaptive adjusting(2):
Histogram of valid
region in (1). mean =
70.48, median value=54.
Image is too dark.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 33
34. self-adaptive adjusting(3):
Acquired image after
self-adaptive adjusting.
Black level =-0.042V,
white level =0.258V,
offset=0.042V, gain=2,
reference= 0.6V
scapula
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 34
35. self-adaptive adjusting(4):
Histogram of valid
region in (3). mean =
121.19. median value =
113.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 35
36. Conclusions
It's possible to implement self-adaptive acquisition of
medical video image automatically by integrating
various images processing method. The proposed
method has recognized the valid region of image and
removed the background, then adjusted acquisition
parameters by analyzing image dynamic range to
obtain best acquisition quality. But there still some
problem remained to be resolved.
Aug.22,2000,WCC2000 Jie BAO , ImageInfoLab , Hefei University of Technology 36