1. Development of a compact time-resolved
two-dimensional dosimetry system
Wonjoong Cheon1), Hyunuk Jung2), Moonhee Lee1), Jinhyeop Lee1),
Sung Jin Kim3), Sungkoo Cho3), Youngyih Han4)*
1) Department of Health Sciences and Technology, SAIHST, Sungkyunkwan University, Seoul, 06351, Korea.
2) Department of Radiation Oncology, University of Texas Southwestern Medical Center, Dallas, TX 75235, USA
3) Department of Radiation Oncology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, 06351, Korea.
1
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville
3. 1. History of dosimetry using 2D scintillator
# Title Year Author Structure of detector
1 A fast 2D phantom dosimetry system for scanning proton beams 1998 S.N Boon and P. van Luijk 2D Scintillation sheet + Mirror + CCD camera
2 Performance of a fluorescent screen and CCD camera as a two-dimensional dosimetry system for dynamic treatment techniques 2000 S.N Boon and P. van Luijk 2D Scintillation sheet + Mirror + CCD camera
3 Development and characterization of a tissue equivalent plastic scintillator based dosimetry system. 2006 Petric 2D Scintillation sheet + Mirror + CCD camera
4 A new scintillating fiber dosimetry using a single optical fiber and a CCD camera 2006 Ferelin 1D Scintilation fiber + Mirror + CCD camera
5 The DosiMap, a new 2D scintillating dosimeter for IMRT quality assurance: characterization of two Cerenkov discrimination methods. 2008 Ferelin 2D Scintillation sheet + Mirror + CCD camera
6 The DOSIMAP, a high spatial resolution tissue equivalent 2D dosimeter for LINAC QA and IMRT verification 2009 Collomb-Patton 2D Scintillation sheet + Mirror + CCD camera
7 2D dosimetry in a proton beam with a scintillating GEM detector 2009 E seravalli 2D Scintillation sheet + Mirror + CCD camera
8 A new water-equivalent 2D plastic scintillation detectors array for dosimetry of megavoltage energy photon beams in radiation therapy 2011 Guillot 2D Scintillation fiber + Mirror + CCD camera
9 Performance assessment of a 2D array of plastic scintillation detector 2013 Guillot 2D Scintillation fiber + Mirror + CCD camera
10 https://www.iba-dosimetry.com/solutions/radiation-therapy/particle-therapy-qa/lynx-pt/ 2015 IBA_Dosimetry 2D Scintillation sheet + Mirror + CCD camera
11 Development and characterization of a 2D scintillation detecotr for quality assuracne in scanned carbon ion beam 2016 A. Tamborini 2D Scintillation sheet + Mirror + CCD camera
12 Characterization of a GEM-based scintillation detector with He-CF4 gas mixture in clinical proton beams. 2016 Nichiporov D, 2D Scintillation sheet + Mirror + CCD camera
13 Characterization of a commercial scintillation detector for 2-D dosimetry in scanned proton and carbon ion beams 2017 S. Susso 2D Scintillation sheet + Mirror + CCD camera
Table 1. Researches of 2D scintillation dosimetry are summarized since 1998
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
4. To measure two-dimensional dose distribution
by using “Time-resolved” “Mirror-less” scintillation detector
H/W
Time-resolved mirrorless scintillation detector
S/W
Visible
light
Dose
distribution
Dose
profile
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
5. Conventional
Scintillation detector system
(in-house)
Too much Heavy
Too much Bulky
Specification
Conventional
Scintillation detector system
Size (L x W x H) 60.7 x 37.0 x 33.5 cm3
Weight 18.26 kg
Effective field of view 20.0 x 20.0 cm2
Specification
Mirror-less Compact
Scintillation detector system
Size (L x W x H) 40.5 × 32.5 × 15.5 cm3
Weight 1.74 kg
Effective field of view 17.0 x 17.0 cm2
Time-resolved mirrorless
Scintillation detector system
5Radiation Protection Dosimetry, Vol. 131(1), 2008
Compact: Without Mirror and Optical fibers
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
6. 10 X 10 cm2 open field measured by GoPro Hero5 Black camera without the mirror
3. Geometric distortion
• Lens distortion
• Perspective distortion
2. Noise (hot-pixel)
4. Dependency of luminance
• Output factor of field size
• Dependency with distance
5. Penumbra
1. Optimization of camera parameters
• ISO value
• Shutter speed
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
8. 1. Optimization of GoPro Hero5 Black camera parameters
Optimization of
camera parameters
Noise reduction
based on frames
Geometric correction Penumbra region correction
CCD camera specification
• Model: GoPro Hero5 Black
Weight : 118 g
• Parameters:
- Resolution: 4K/1440/1080
- Frame Rates: 24
- Field Of View: Wide/Narrow/Medium/Linear
- ProTune: On/Off
- Shutter speed: Auto/FPS-1
- Exposure value compensation: -2.0 to 2.0
- ISO limit: 6400/3200/1600/1200/800/400
But, to use GoPro5 as for radiation dosimetry, Setting of camera parameters is necessary.
8
Light dependency
correction
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
9. Shutter speed
: The time the shutter was opened and closed
ISO Limit
: Level of sensitivity of a camera to available light.
ISO LIMIT: 800 ISO LIMIT: 1600 ISO LIMIT: 3200ISO LIMIT 9
Trial #
Parameters
1 2 3 4 5 6 7 8 9 10
Resolution
Frame rates
Field of view
Protune OFF
Shutter -
Exposure -
ISO value - 800 1600 3200 800 1600 3200 800 1600 3200
-2 2-
4K
24
Wide Angle
ON
24 AUTO
1. Optimization of GoPro Hero5 Black camera parameters
Optimization of
camera parameters
Frame-based noise
reduction technique
Geometric correction Penumbra region correction
Light dependency
correction
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
10. 𝐢𝐟
𝒗𝒇𝒏 = 𝒏
𝒏+𝒌
𝑷 𝒗𝒇𝒏, 𝒊, 𝒋 ≥ 𝒕𝒉𝒓𝒆𝒔𝒉𝒐𝒍𝒅 𝒗𝒂𝒍𝒖𝒆 , 𝐭𝐡𝐞𝐧 𝑷 𝒗𝒇𝒏, 𝒊, 𝒋 = 𝟎
𝐟𝐨𝐫 𝐚𝐥𝐥 𝒊 , 𝒋 𝐢𝐧 𝐩𝐢𝐱𝐞𝐥 𝐜𝐨𝐨𝐫𝐝𝐢𝐧𝐚𝐭𝐞𝐬
Frame number= n Frame number= n+1 Frame number= n+2 Frame number= n+k
Radiation directly interact with sensor
“Noise” (Hot-pixel)
10
• n : selected frame for applying noise reduction technique
• k : The frame receive a scan for detecting hot-pixel
• vfn : video frame number
• i, j : pixel(image) coordinates
To eliminate noise, the above procedure was applied to over all frames
2. Frame-based noise reduction technique
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
Optimization of
camera parameters
Frame-based noise
reduction technique
Geometric correction Penumbra region correction
Light dependency
correction
11. Original image
captured by GoPro5
Correction of
Lens distortion
Correction of
perspective effect
(Problems caused by the angle of the camera)(Problems caused by fish eye lens)
11
3. Geometric correction: Lens correction and Perspective correction
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
Optimization of
camera parameters
Frame-based noise
reduction technique
Geometric correction Penumbra region correction
Light dependency
correction
12. Horizontal and vertical profile
Field size: 20.0 X 20.0
Correction map (2D)
Distance dependency correction map
Inverse 1/r square law !
12
4. Distance dependency correction
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
Optimization of
camera parameters
Frame-based noise
reduction technique
Geometric correction Penumbra region correction
Light dependency
correction
13. 5. Penumbra region correction using combination of kernels
The kernel optimized through CNN network
Optimization of
camera parameters
Frame-based noise
reduction technique
Geometric correction Penumbra region correction
Light dependency
correction
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
15. 1. Linearity and Repeatability
TRMLSD vs EBT3
Figure 1. (a) Linearity: Plot for scintillation intensity corresponding to absorbed dose; red
line is linear fit to measured date. (b) Repeatability: Measured absorbed dose of TRMLSD
(red star) and Gafchromic EBT3 (blue square) are plotted.
- 20 cGy
- 200 cGy
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
16. 2. Percent depth dose profile (PDD):
TRMLSD vs EBT3 vs Ionization chamber
FIG. 2. Percent depth dose profile measured by TRMLSD (red star), EBT3 film (blue
square), and ionization chamber (blue dash) for 6MV photon beam. TRMLSD and
EBT3 data were normalized at depth 5.0 cm
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
17. 3. Profile (TG-142):
TRMLSD vs EBT3 vs Ion chamber
Measurement
tool
Flatness (%) Symmetry (%) Penumbra (cm)
Ion Chamber 1.78 0.83
L: 0.85
R: 0.85
EBT3 film 1.62 0.85
L: 0.42
R: 0.46
TRMLSD 1.88 0.86
L: 0.49
R: 0.49
Table 1. Beam profile verification of Novalis Tx accelerator: 6MV photon beam,
reference depth 5.0 cm.
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
18. 4. Simple IMRT plan (time-resolved dosimetry):
TRMLSD vs EBT3 film
Field type 1 [%]
Accumulation time (sec)
Criterion 81.6 110.4 127.2
2% / 2mm 91.59 93.47 92.93
3% / 3mm 94.72 97.84 96.89
4% / 4mm 99.93 99.99 99.86
Figure 3. Signal of scintillation over time for simple intensity
modulated plan (field type-1). The accumulated two-dimensional (2D)
dose distribution of TRMLSD at (a) 81.6, (b) 110.4, and (c) 127.2 sec.
2D dose distribution of EBT3 film (d)
Table 2. Gamma analysis between TRMLSD
and EBT3 film for different gamma criterion.
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
19. 5. Clinical IMRT plan (Prostate: 7 beam with 50 segments):
TRMLSD vs TPS (Pinnacle)
2% / 2 mm 3 % / 3 mm 4 % / 4 mm
TRMLSD 93.62 96.89 99.43
EBT3 film 94.76 97.34 99.63
1 2 3 4 5 6 7 Avg.
2% / 2mm 92.51 92.75 92.95 92.29 92.42 93.29 93.28 92.78
3% / 3mm 97.38 97.42 97.28 96.44 96.31 97.02 95.81 96.81
4% / 4mm 99.38 99.42 98.44 99.03 99.48 99.17 99.37 99.18
Table 3. Gamma passing rate with diverse gamma criterion for
comparing two-dimensional dose distribution of
TRMLSD and EBT3 film with RTP
Table 4. Gamma passing rate with diverse gamma criterion between two-dimensional dose distribution of
TRMLSD and RTP.
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
20. Summary
We developed Times-resolved mirrorless compact scintillation detector.
The weight is 10 times lighter than conventional scintillation detector (compact)
It could analyze the two-dimensional dose distribution of plan
as per segment, beam. (time-resolved)
The effective field size is 17.0 x 17.0 cm2 in 2000 x 2000 pixel resolution
(high-resolution)
The accuracy of system for Clinical IMRT plan is about 96 percent using gamma analysis
method with 3% / 3mm gamma criterion
In addition, we will perform a dose evaluation with time-resolved using TRMLSD
for VMAT, dynamic scanning method for proton and carbon therapy, where
treatment parameters change with time.
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
21. Thank you for your
attention
wonjoongcheon@gmail.com
https://www.cakeresume.com/wonjoong-cheon
https://github.com/wjcheon
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
23. 𝐢𝐟
𝒗𝒇𝒏 = 𝒏
𝒏+𝒌
𝑷 𝒗𝒇𝒏, 𝒊, 𝒋 ≥ 𝒕𝒉𝒓𝒆𝒔𝒉𝒐𝒍𝒅 𝒗𝒂𝒍𝒖𝒆 , 𝐭𝐡𝐞𝐧 𝑷 𝒗𝒇𝒏, 𝒊, 𝒋 = 𝟎
𝐟𝐨𝐫 𝐚𝐥𝐥 𝒊 , 𝒋 𝐢𝐧 𝐩𝐢𝐱𝐞𝐥 𝐜𝐨𝐨𝐫𝐝𝐢𝐧𝐚𝐭𝐞𝐬
Frame number= n Frame number= n+1 Frame number= n+2 Frame number= n+k
Radiation directly interact with sensor
“Noise” (Hot-pixel)
23
• n : selected frame for applying noise reduction technique
• k : The frame receive a scan for detecting hot-pixel
• vfn : video frame number
• i, j : pixel(image) coordinates
To eliminate noise, the above procedure was applied to over all frames
2. Frame-based noise reduction technique
60th Annual Meeting & Exhibition: AAPM 2018 @ Nashville, W. Cheon
Optimization of
camera parameters
Frame-based noise
reduction technique
Geometric correction Penumbra region correction
Light dependency
correction
24. IF, Scanning this frame with k = 5. the maximum threshold value is 5
5
3
signal
Noise(hot-pixel)
25. IF, Scanning this frame with k = 5. the maximum threshold value is 5
5
3
signal
Noise(hot-pixel)
Editor's Notes
This is the list of two-dimensional, 2D, scintillator studies.
One of a common features of the previous studies employing a CCD camera is that a mirror has been used to deflect lights from a scintillation plate to a CCD camera.
However, we removed the mirror and developed a time-resolved mirrorless scintillation detector, named “TRMLSD”, for two-dimensional dosimetry. This includes new hardware design and software developments.
Removing a mirror make the dosimeter can be compact and light, which is convenient to be used as a portable detector.
Therefore, the TRMLSD consist of only a CCD camera and a scintillation sheet, thus the weight was reduced to 1.74 kg. but the camera was position with a certain angle to capture the lights from scintillation sheet.
The image for 10x10 cm2 radiation field taken by TRMLSD is shown in this slide.
As you notice, the geometry of the field is distorted and there were many problems to be resolved for an accurate two-dimensional dosimetry.
The first thing was determining camera parameters, …..
Firstly, let me introduce the CCD camera used. GoPro5, we usually called “Action cam” was used for TRMLSD, because, is lighter and relatively chipper than scientific CCD sensor. Additionally , This camera support 4k video resolution which is enough for high spatial resolution.
But, To use GoPro5 for dosimetry, selecting optimal camera parameters was necessary.
The Key parameters are Shutter speed and ISO Limit,
For shutter speed, The shortest shutter speed is best for time-resolved dosimetry.
As for the ISO Limit, It means that “Level of sensitivity of a camera to available light”
It controls the sensitivity as shown in bottom of this slide which recorded radiation with different ISO limit value.Finally, 4k video resolution, and the 1/24 second of shutter speed and the 800 of ISO limit were determined.
Next let me talk about the Frame based noise reduction technique.
When radiation directly interact with sensor, the hot-pixel was generated in each images instantaneously.
However, it is not permanent effect.
Thus, If the summation of pixel intensity(i,j) from n th frame to n+k frame is higher than threshold value(?) then, Pixel intensity (i,j) is treated as noise for all i, j in pixel coordinates .
To eliminate noise the above procedure was applied to all frames of images.
As for the geometric problem, firstly geometry correction due to a fish-eye lens of GoPro5 was done.
Second problem was “perspective effect” caused by the angled position of the camera without a mirror. And it was corrected as shown in the 3rd image.
Next, The distance dependancy was corrected which was caused by varying distance between the camera lens and the each pixel of light source.
You can see the vertical profile of image, on the right side of slide
Despite different size of homogenous radiation fields were irradiated, you can see that the asymmetric profile was displayed along the y direction.
To correct decreasing light intensity following inverse square of distance, 2D correction map was generated by 20 by 20 radiation field and applied to each pixel of captured image.
Lastly, penumbra region of the radiation fields was corrected using combination of different kernels
In order to find the optimal combination of correction kernels, we used Deep learning method rather than mathematical or experimental approaches.
on the right side of slide, you can see the profiles before- and after- correction .
Now let me move on Results.
In order to validate the performance of TRMLSD,
The Linearity and Repeatability were tested.
Response to the dose was linear as R-square value was very close to 1.0.
And the repeatability of response for 20 cGy, and 200 cGy was within 3 % of deviation.
PDD was measured by TRMLSD, and EBT3 film
In comparison with ion-chamber data,
The maximum difference of TRMLSD was 1.51 % on 2.5 cm depth.
In case of EBT3 film, 2.16 % on same depth.
Profiles of 6 MV X-ray of 10x10 cm2 field was measured and anlayzed.
The flatness, symmetry, and penumbra size measured with TRMLSD were in good agreement with Ion chamber and EBT3 flim measurements.
To test the feasibility of clinical field dosimetry,
A Simple IMRT plan consisted of three different size and MU of rectangular fields
was delivered and measured with TRMLSD.
All frames having signals were separately accumulated between time intervals for dosimetry of each field.
Comparing with EBT3 film, The gamma passing rates with 3%/3mm gamma criterion were 94, 97, 96 percent, for 1st, 2nd and 3rd field, respectively.
Lastly, we measured prostate plan composed of 7 beams with 50 segments.
and 2D dose distribution was analyzed per each beam.
The gamma analysis for total plan dose and per beam.
The gamma passing rate for each beam was over 95% with 3%/3mm criterion.
The averaged gamma passing rate was 96.81 percent.
Next let me talk about the Frame based noise reduction technique.
When radiation directly interact with sensor, the hot-pixel was generated in each images instantaneously.
However, it is not permanent effect.
Thus, If the summation of pixel intensity(i,j) from n th frame to n+k frame is higher than threshold value(?) then, Pixel intensity (i,j) is treated as noise for all i, j in pixel coordinates .
To eliminate noise the above procedure was applied to all frames of images.