SlideShare a Scribd company logo
1 of 39
MLC error prediction
using recurrent neural network (LSTM)
Wonjoong Cheon1), Kim Seong Jung2), Youngyih Han3),
Hyebin Lee4), Byung Jun Min4*), Heerim Nam4)
1) Department of Health Sciences and Technology, SAIHST, Sungkyunkwan University, Seoul, 06351, Korea.
2) Department of Computer Engineering, Yonsei University, 03722, Seoul, Korea.
3) Department of Radiation Oncology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, 06351, Korea.
4) Department of Radiation Oncology, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine, Seoul, 03181, Korea.
1
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Contents
I. Introduction
II. Material & Method
1) Convert Rtplan to Expected position of MLC
2) Prediction Actual position of MLC using artificial neural network
3) Statistical analysis
III. Results
IV.Conclusion
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅰ. Introduction
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅰ. Introduction
Multi-Leaf Collimator
160 MLC™
(Siemens)
Agility™
(Elekta)
• Intensity modulated radiation therapy (IMRT) technique uses MLCs for modifying the beam fluence in the same treatment field
in order to improve the conformity of prescribed dose distribution around the tumor region. (Mundt & Roeske, 2005)
• The volumetric modulated arc therapy (VMAT) technology is a novel delivery method that is capable of producing highly conformal
dose distributions through concomitant optimization of MLC shapes, dose rate, and gantry speed.(Schreibmann et al., 2009)
Millennium™ MLC
(Varian)
Figure resource: www.elekta.com, www.varian.com, www.healthcare.siemens.com
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Actual
MLC position
Expected
MLC position
Ⅰ. Introduction
MLC log files
Dicom RT
Eclipse™ Treatment Planning System, Varian
Dynalog file
Trilogy® System Linear Accelerator, Varian
• MLC leaf position uncertainties directly affect the dose delivered to tumor targets and sensitive
structures in IMRT. (Mu et al., 2008)
• Actual multileaf collimator (MLC) position data can be tracked throughout a treatment delivery in the
form of MLC log files. (Kerns, Childress, & Kry, 2014)
Mismatch (=Errors)
Figure resource: www.varian.com
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
 Radiation treatment plan
(RTplan)
- plans may contain
fractionation information,
and define external beams
and/or brachytherapy
 Expected position
- Position of each leaf
calculated and predefined
by RTplan file
 Actual position
- Position of each leaf
conducted and written to
dynalog file
- Containing machine errors
Ⅰ. Introduction
RT workflow
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
RT workflow
Ⅰ. Introduction
MLC positional error
• Most of the studies on MLC positional error — both random and systematic errors were analyzed —
have reported that the random errors were insignificant, while systematic errors have shown significant effects on dose distributions,
even with only 1 mm of positional errors applied for MLC positions. (Nithiyanantham et al., 2015)
• MLC position and velocity errors are affected by Friction and Gravity. (Wijesooriya, K., et al., 2005, Lee, Jeong-Woo, et al.,)
• Varian said that “The leaves have to fight gravity”
Figure reference : http://medicalphysicsweb.org/cws/article/research/44068
1
2
3
4
5
6
56
57
58
59
60
2
3
4
Friction Gravity
2
3
4
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅰ. Introduction
Aim of this study
• Prediction multi-leaf collimator position and velocity errors before radiation treatment
using radiation treatment plan (RTplan) file and artificial neural network (RNN:LSTM).
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Treatment
Planning
Computer
RT plan
Batch
Folder
LinacView
Computer
Logfile
Batch
Folder
Linac Console
RT Plans
Log files
RT Plans
Convert
RTplan to
Expected position
of MLC
MLC position
expected
Prediction
Actual position
of MLC
Artificial Neural NetworkStatistical
analysis
Probability
error function
1
2
3
Is this error
acceptable?
No
Warning
MLC error prediction
RT workflow
If I operate the this
linear accelerator
based on my
RTplan file, will it
work safely?
Ⅱ. Material & Method
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅱ. Material & Method
StepⅠ
 Convert Radiation treatment plan to Expected position of MLC
• Computer Language: Matlab
• Parameters: Monitor unit per second, Dose rate, Gantry angle
Step Ⅱ
 Prediction Actual position of MLC using Artificial neural network
• Computer Language: Python
• Type of neural net: RNN: LSTM
• Deep learning framework: Tensorflow (Google)
• Parameters: MLC position (expected, actual), Gantry angle, collimator angle, Beam on/off sign
Step Ⅲ
 Statistical analysis :Generate Probability error function
• Preprocess: Central limit theorem
• Model: Single Gaussian distribution (mean, std)
• Computer Language: Matlab
Radiation treatment plan file Dynalog file (expected position)
Ⅱ. Material & Method
Convert RTplan to Expected position using Matlab
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅱ. Material & Method
StepⅠ
 Convert Radiation treatment plan to Expected position of MLC
• Computer Language: Matlab
• Parameters: Monitor unit, Dose rate
Step Ⅱ
 Prediction Actual position of MLC using Artificial neural network
• Computer Language: Python
• Type of neural net: RNN: LSTM
• Deep learning framework: Tensorflow
• Parameters: MLC position (expected, actual), Gantry angle, collimator angle, Beam on/off sign
Step Ⅲ
 Statistical analysis :Generate Probability error function
• Preprocess: Central limit theorem
• Model: Single Gaussian (mean, std)
• Computer Language: Matlab
• RNN’s ability to anticipate also makes them capable of surprising creativity.
• You can ask them to predict which are the most likely next notes in a melody,
then randomly pick one of these notes and play it. Then ask the net for the next
most likely notes.
 How it is work ?
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅱ. Material & Method
Prediction Actual position using Artificial neural network
 Learning for prediction of actual position of MLC
Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60
1480Dynalog
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅱ. Material & Method
Prediction Actual position using Artificial neural network
 Window size: 7 time stamp
 Gantry angle: Ga(t)
 Collimator angle: Ca(t)
 Beam on/off signal: B(t)
 Leaf expected position: leafPosi(t)
 Leaf actual position:
Actual position of 2nd leaf
Time: t=8
Ga(t)Ca(t) B(t) leafPosi(t)
 Learning for prediction of actual position of MLC
Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60
1480Dynalog
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅱ. Material & Method
Prediction Actual position using Artificial neural network
 Window size: 7 time stamp
 Gantry angle: Ga(t)
 Collimator angle: Ca(t)
 Beam on/off signal: B(t)
 Leaf expected position: leafPosi(t)
 Leaf actual position:
Actual position of 2nd leaf
Time: t=9
Ga(t)Ca(t) B(t) leafPosi(t)
 Learning for prediction of actual position of MLC
Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60
1480Dynalog
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅱ. Material & Method
Prediction Actual position using Artificial neural network
 Window size: 7 time stamp
 Gantry angle: Ga(t)
 Collimator angle: Ca(t)
 Beam on/off signal: B(t)
 Leaf expected position: leafPosi(t)
 Leaf actual position:
Actual position of 2nd leaf
Time: t=10
Ga(t)Ca(t) B(t) leafPosi(t)
 Learning for prediction of actual position of MLC
Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60
1480Dynalog
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅱ. Material & Method
Prediction Actual position using Artificial neural network
 Window size: 7 time stamp
 Gantry angle: Ga(t)
 Collimator angle: Ca(t)
 Beam on/off signal: B(t)
 Leaf expected position: leafPosi(t)
 Leaf actual position:
Actual position of 2nd leaf
Time: t=11
Ga(t)Ca(t) B(t) leafPosi(t)
 Learning for prediction of actual position of MLC
Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60
1480Dynalog
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅱ. Material & Method
Prediction Actual position using Artificial neural network
 Window size: 7 time stamp
 Gantry angle: Ga(t)
 Collimator angle: Ca(t)
 Beam on/off signal: B(t)
 Leaf expected position: leafPosi(t)
 Leaf actual position:
Actual position of 2nd leaf
Time: t=12
Ga(t)Ca(t) B(t) leafPosi(t)
 Learning for prediction of actual position of MLC
Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60
1480Dynalog
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅱ. Material & Method
Prediction Actual position using Artificial neural network
 Window size: 7 time stamp
 Gantry angle: Ga(t)
 Collimator angle: Ca(t)
 Beam on/off signal: B(t)
 Leaf expected position: leafPosi(t)
 Leaf actual position:
Actual position of 2nd leaf
Time: t=13
Ga(t)Ca(t) B(t) leafPosi(t)
FC
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
𝑖𝑛𝑝𝑢𝑡𝐺𝑎𝑡𝑒(𝑡) = 𝜎(𝑊𝑥𝑖 ∙ 𝑥(𝑡) + 𝑊ℎ𝑖𝑇 ∙ ℎ(𝑡 − 1) + 𝑏𝑖)
𝑓𝑜𝑟𝑔𝑜𝑡𝐺𝑎𝑡𝑒(𝑡) = 𝜎(𝑊𝑥𝑓 ∙ 𝑥(𝑡) + 𝑊ℎ𝑓𝑇 ∙ ℎ(𝑡 − 1) + 𝑏𝑓)
𝑜𝑢𝑡𝑝𝑢𝑡𝐺𝑎𝑡𝑒(𝑡) = 𝜎(𝑊𝑥𝑜 ∙ 𝑥(𝑡) + 𝑊ℎ𝑜𝑇 ∙ ℎ(𝑡 − 1) + 𝑏𝑜)
• LSTM structure
• MLCLogNet structure
𝐿𝑜𝑛𝑔 𝑆ℎ𝑜𝑟𝑡 𝑡𝑒𝑟𝑚 𝑚𝑒𝑚𝑜𝑟𝑦 3 𝑐𝑒𝑙𝑙
𝐹𝑢𝑙𝑙𝑦 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑒𝑑 𝑙𝑎𝑦𝑒𝑟
ℎ𝑖𝑑𝑑𝑒𝑛 dim = 20
o𝑢𝑡𝑝𝑢𝑡 dim = 1
Ⅱ. Material & Method
Prediction Actual position using Artificial neural network
 Learning for prediction of actual position of MLC
Structure of learning frame
 Validation of MLCLog Net
Log 1
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Log 2 Log 3 Log 4
Log 175 Log 176 Log 178
Log 1 Log 2 Log 3
Training log set Test log set
Ⅱ. Material & Method
Prediction Actual position using Artificial neural network
Log 30
Ⅱ. Material & Method
StepⅠ
 Convert Radiation treatment plan to Expected position of MLC
• Computer Language: Matlab
• Parameters: Monitor unit, Dose rate
Step Ⅱ
 Prediction Actual position of MLC using deep learning
• Computer Language: Python
• Type of neural net: RNN: LSTM
• Deep learning framework: Tensorflow
• Parameters: MLC position (expected, actual), Gantry angle, collimator angle, Beam on/off sign
Step Ⅲ
 Statistical analysis :Generate Probability error function
• Preprocess: Central limit theorem
• Model: Single Gaussian (mean, std)
• Computer Language: Matlab
Ⅱ. Material & Method
Statistical analysis:Generate probability error function
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
𝑷𝒐𝒔𝒊𝒕𝒊𝒐𝒏𝑬𝒓𝒓𝒐𝒓 𝒏, 𝒕 = 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑𝑃𝑜𝑠𝑖𝑡𝑖𝑜𝑛 𝑛, 𝑡 − 𝐴𝑐𝑡𝑢𝑎𝑙𝑃𝑜𝑠𝑖𝑡𝑖𝑜𝑛(𝑛, 𝑡)
𝑽𝒆𝒍𝒐𝒄𝒊𝒕𝒚𝑬𝒓𝒓𝒐𝒓 𝒏, 𝒕 = 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑛, 𝑡 − 𝐴𝑐𝑡𝑢𝑎𝑙𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦(𝑛, 𝑡)
𝐴𝑐𝑡𝑢𝑐𝑎𝑙𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑛, 𝑡
=
)−𝐴𝑃 𝑡 + 2, 𝑛 + 8 ∗ 𝐴𝑃 𝑡 + 1, 𝑛 − 8 ∗ 𝐴𝑃 𝑡 − 1, 𝑛 + 𝐴𝑃(𝑡 − 2, 𝑛
𝑑𝑒𝑙𝑡𝑎 𝑡 ∗ 12
𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑛, 𝑡
=
)−𝐸𝑃 𝑡 + 2, 𝑛 + 8 ∗ 𝐸𝑃 𝑡 + 1, 𝑛 − 8 ∗ 𝐸𝑃 𝑡 − 1, 𝑛 + 𝐸𝑃(𝑡 − 2, 𝑛
𝑑𝑒𝑙𝑡𝑎 𝑡 ∗ 12
Central limit theorem
Central limit theorem
[mm]
[mm/s]
Ⅲ. Results
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
StepⅠ
 Convert Radiation treatment plan to Expected position of MLC
Step Ⅱ
 Prediction Actual position of MLC using Artificial neural network
Step Ⅲ
 Statistical analysis :Generate Probability error function
Ⅲ. Results
Footnote: Validation of calculated Expected position from RTplan Footnote: Comparison calculated expected position and expected
position extracted from dynalog file (30 th leaf of MLC)
Ⅲ. Results
Convert RTplan to Expected position using Matlab
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Plot the cost value of l2 loss function Prediction actual position of MLC using
recurrent neural network (LSTM)
Ⅲ. Results
Prediction Actual position of MLC using Artificial neural network
Plot the cost value of l2 loss function Prediction actual position of MLC using
recurrent neural network (LSTM)
Ⅲ. Results
Prediction Actual position of MLC using Artificial neural network
Ⅲ. Results
Statistical analysis :Generate Probability error function
[mm] [mm/s]
Ⅰ. Introduction
Aim of this study
• Prediction multi-leaf collimator position and velocity errors before radiation treatment
using radiation treatment plan (RTplan) file and artificial neural network (RNN:LSTM).
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Treatment
Planning
Computer
RT plan
Batch
Folder
RT PlansRT Plans
Convert
RTplan to
Expected position
of MLC
MLC position
expected
Prediction
Actual position
of MLC
Artificial Neural NetworkStatistical
analysis
Probability
error function
1
2
3
Is this error
acceptable?
No
Warning
MLC error prediction
[mm] [mm]
Ⅳ. Conclusion
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Ⅳ. Conclusion
• Multi-leaf collimators have had a major impact on the development of radiation therapy such
as IMRT, VMAT and IMPT.
• Small defects of the multi-leaf collimator can have a large effect on the results of radiation
therapy.
• We build workflow for predicting mechanical errors of MLC before radiation treatment.
• Consequently, our team developed a method of prediction multi-leaf collimator position
and velocity errors before radiation treatment
using radiation treatment plan (RTplan) file and artificial neural network (RNN:LSTM).
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
V. Further study
• Actual dose prediction using recurrent neural network (LSTM)
• Expected dose distribution • Actual dose distribution
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Thank you for your attention.
2017/11/10
대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
Acknowledgement:
This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government
(MSIP) (No. NRF-2017R1C1B2011257)
ERROR RANGE 1
Max acc 49.49 Max acc 55.65
With BCGWithout BCG
ERROR RANGE 2
Max acc :84.28 Max acc :84.74
Max acc :96.33
ERROR RANGE 3
Max acc :97.19
Max acc :99.45
ERROR RANGE 4
Max acc :99.60
Max acc : 1.00
ERROR RANGE 5
Max acc : 1.00

More Related Content

What's hot

Emergency in HDR Brachytherapy
Emergency in HDR Brachytherapy Emergency in HDR Brachytherapy
Emergency in HDR Brachytherapy Monik Patel
 
Imrt Treatment Planning And Dosimetry
Imrt Treatment Planning And DosimetryImrt Treatment Planning And Dosimetry
Imrt Treatment Planning And Dosimetryfondas vakalis
 
SBRT: The Importance of Continuous Motion Management​
SBRT: The Importance of Continuous Motion Management​SBRT: The Importance of Continuous Motion Management​
SBRT: The Importance of Continuous Motion Management​SGRT Community
 
Radiation Shielding for Mega-voltage Photon Therapy Machines
 Radiation Shielding for Mega-voltage Photon Therapy Machines  Radiation Shielding for Mega-voltage Photon Therapy Machines
Radiation Shielding for Mega-voltage Photon Therapy Machines Daryoush Khoramian
 
SRS & SBRT - Unflattened Beam
SRS & SBRT - Unflattened BeamSRS & SBRT - Unflattened Beam
SRS & SBRT - Unflattened BeamKothanda Raman
 
Unscheduled rt interruptions
Unscheduled rt interruptionsUnscheduled rt interruptions
Unscheduled rt interruptionsHEBAGOMAA1984
 
SGRT for DIBH: from zero to hero 
SGRT for DIBH: from zero to hero  SGRT for DIBH: from zero to hero 
SGRT for DIBH: from zero to hero  SGRT Community
 
Immobilization in radiotherapy
Immobilization in radiotherapy Immobilization in radiotherapy
Immobilization in radiotherapy prabhurt
 
Radiotheray transition from 2D to 3D Conformal radiotherapy(3D-CRT)
Radiotheray transition  from 2D to 3D Conformal  radiotherapy(3D-CRT)Radiotheray transition  from 2D to 3D Conformal  radiotherapy(3D-CRT)
Radiotheray transition from 2D to 3D Conformal radiotherapy(3D-CRT)Gebrekirstos Hagos Gebrekirstos, MD
 
Quality Assurance in Radiotherapy and Dosimetry
Quality Assurance in Radiotherapy and DosimetryQuality Assurance in Radiotherapy and Dosimetry
Quality Assurance in Radiotherapy and Dosimetrymahbubul hassan
 
Commissioning of Truebeam LINAC
Commissioning of Truebeam LINACCommissioning of Truebeam LINAC
Commissioning of Truebeam LINACVIneeth C
 
The Radiobiology Behind Dose Fractionation.pdf
The Radiobiology Behind Dose Fractionation.pdfThe Radiobiology Behind Dose Fractionation.pdf
The Radiobiology Behind Dose Fractionation.pdfministry of health
 
Image guided radiation therapy (2011)
Image guided radiation therapy (2011)Image guided radiation therapy (2011)
Image guided radiation therapy (2011)Parminder S. Basran
 

What's hot (20)

Flattening filter Free
Flattening filter FreeFlattening filter Free
Flattening filter Free
 
Emergency in HDR Brachytherapy
Emergency in HDR Brachytherapy Emergency in HDR Brachytherapy
Emergency in HDR Brachytherapy
 
Imrt Treatment Planning And Dosimetry
Imrt Treatment Planning And DosimetryImrt Treatment Planning And Dosimetry
Imrt Treatment Planning And Dosimetry
 
SBRT: The Importance of Continuous Motion Management​
SBRT: The Importance of Continuous Motion Management​SBRT: The Importance of Continuous Motion Management​
SBRT: The Importance of Continuous Motion Management​
 
Radiation Shielding for Mega-voltage Photon Therapy Machines
 Radiation Shielding for Mega-voltage Photon Therapy Machines  Radiation Shielding for Mega-voltage Photon Therapy Machines
Radiation Shielding for Mega-voltage Photon Therapy Machines
 
SRS & SBRT - Unflattened Beam
SRS & SBRT - Unflattened BeamSRS & SBRT - Unflattened Beam
SRS & SBRT - Unflattened Beam
 
Immobilizations devices
Immobilizations devicesImmobilizations devices
Immobilizations devices
 
AAPM REPORT 59
AAPM REPORT 59AAPM REPORT 59
AAPM REPORT 59
 
Unscheduled rt interruptions
Unscheduled rt interruptionsUnscheduled rt interruptions
Unscheduled rt interruptions
 
TomoTherapy
TomoTherapyTomoTherapy
TomoTherapy
 
4 D Igrt
4 D Igrt4 D Igrt
4 D Igrt
 
Radiotherapy technique
Radiotherapy techniqueRadiotherapy technique
Radiotherapy technique
 
Gap correction
Gap correctionGap correction
Gap correction
 
SGRT for DIBH: from zero to hero 
SGRT for DIBH: from zero to hero  SGRT for DIBH: from zero to hero 
SGRT for DIBH: from zero to hero 
 
Immobilization in radiotherapy
Immobilization in radiotherapy Immobilization in radiotherapy
Immobilization in radiotherapy
 
Radiotheray transition from 2D to 3D Conformal radiotherapy(3D-CRT)
Radiotheray transition  from 2D to 3D Conformal  radiotherapy(3D-CRT)Radiotheray transition  from 2D to 3D Conformal  radiotherapy(3D-CRT)
Radiotheray transition from 2D to 3D Conformal radiotherapy(3D-CRT)
 
Quality Assurance in Radiotherapy and Dosimetry
Quality Assurance in Radiotherapy and DosimetryQuality Assurance in Radiotherapy and Dosimetry
Quality Assurance in Radiotherapy and Dosimetry
 
Commissioning of Truebeam LINAC
Commissioning of Truebeam LINACCommissioning of Truebeam LINAC
Commissioning of Truebeam LINAC
 
The Radiobiology Behind Dose Fractionation.pdf
The Radiobiology Behind Dose Fractionation.pdfThe Radiobiology Behind Dose Fractionation.pdf
The Radiobiology Behind Dose Fractionation.pdf
 
Image guided radiation therapy (2011)
Image guided radiation therapy (2011)Image guided radiation therapy (2011)
Image guided radiation therapy (2011)
 

Similar to Error of Multileaf collimator prediction using recurrent neural network (LSTM)

2009 HEP Science Network Requirements Workshop Final Report
2009 HEP Science Network Requirements Workshop Final Report2009 HEP Science Network Requirements Workshop Final Report
2009 HEP Science Network Requirements Workshop Final Reportbutest
 
Thermatic simulation platform for nano materials design in kist
Thermatic simulation platform for nano materials design in kistThermatic simulation platform for nano materials design in kist
Thermatic simulation platform for nano materials design in kistKIST
 
NIST-JARVIS infrastructure for Improved Materials Design
NIST-JARVIS infrastructure for Improved Materials DesignNIST-JARVIS infrastructure for Improved Materials Design
NIST-JARVIS infrastructure for Improved Materials DesignKAMAL CHOUDHARY
 
Machine Learning for Molecules
Machine Learning for MoleculesMachine Learning for Molecules
Machine Learning for MoleculesIchigaku Takigawa
 
Morgan uw maGIV v1.3 dist
Morgan uw maGIV v1.3 distMorgan uw maGIV v1.3 dist
Morgan uw maGIV v1.3 distddm314
 
(PhD Dissertation Defense) Theoretical and Numerical Investigations on Crysta...
(PhD Dissertation Defense) Theoretical and Numerical Investigations on Crysta...(PhD Dissertation Defense) Theoretical and Numerical Investigations on Crysta...
(PhD Dissertation Defense) Theoretical and Numerical Investigations on Crysta...James D.B. Wang, PhD
 
Graphs, Environments, and Machine Learning for Materials Science
Graphs, Environments, and Machine Learning for Materials ScienceGraphs, Environments, and Machine Learning for Materials Science
Graphs, Environments, and Machine Learning for Materials Scienceaimsnist
 
Materials Design in the Age of Deep Learning and Quantum Computation
Materials Design in the Age of Deep Learning and Quantum ComputationMaterials Design in the Age of Deep Learning and Quantum Computation
Materials Design in the Age of Deep Learning and Quantum ComputationKAMAL CHOUDHARY
 
Multivariate dimensionality reduction in cross-correlation analysis
Multivariate dimensionality reduction in cross-correlation analysis Multivariate dimensionality reduction in cross-correlation analysis
Multivariate dimensionality reduction in cross-correlation analysis ivanokitov
 
Autonomous experimental phase diagram acquisition
Autonomous experimental phase diagram acquisitionAutonomous experimental phase diagram acquisition
Autonomous experimental phase diagram acquisitionaimsnist
 
Predicting local atomic structures from X-ray absorption spectroscopy using t...
Predicting local atomic structures from X-ray absorption spectroscopy using t...Predicting local atomic structures from X-ray absorption spectroscopy using t...
Predicting local atomic structures from X-ray absorption spectroscopy using t...aimsnist
 
The Transformation of Systems Biology Into A Large Data Science
The Transformation of Systems Biology Into A Large Data ScienceThe Transformation of Systems Biology Into A Large Data Science
The Transformation of Systems Biology Into A Large Data ScienceRobert Grossman
 
2018 KOSRO-Oral presentation-Wonjoong Cheon
2018 KOSRO-Oral presentation-Wonjoong Cheon2018 KOSRO-Oral presentation-Wonjoong Cheon
2018 KOSRO-Oral presentation-Wonjoong CheonWonjoongCheon
 
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...Globus
 
What can we learn from molecular dynamics simulations of carbon nanotube and ...
What can we learn from molecular dynamics simulations of carbon nanotube and ...What can we learn from molecular dynamics simulations of carbon nanotube and ...
What can we learn from molecular dynamics simulations of carbon nanotube and ...Stephan Irle
 
A multi-sensor based uncut crop edge detection method for head-feeding combin...
A multi-sensor based uncut crop edge detection method for head-feeding combin...A multi-sensor based uncut crop edge detection method for head-feeding combin...
A multi-sensor based uncut crop edge detection method for head-feeding combin...Institute of Agricultural Machinery, NARO
 

Similar to Error of Multileaf collimator prediction using recurrent neural network (LSTM) (20)

2009 HEP Science Network Requirements Workshop Final Report
2009 HEP Science Network Requirements Workshop Final Report2009 HEP Science Network Requirements Workshop Final Report
2009 HEP Science Network Requirements Workshop Final Report
 
Thermatic simulation platform for nano materials design in kist
Thermatic simulation platform for nano materials design in kistThermatic simulation platform for nano materials design in kist
Thermatic simulation platform for nano materials design in kist
 
NIST-JARVIS infrastructure for Improved Materials Design
NIST-JARVIS infrastructure for Improved Materials DesignNIST-JARVIS infrastructure for Improved Materials Design
NIST-JARVIS infrastructure for Improved Materials Design
 
Machine Learning for Molecules
Machine Learning for MoleculesMachine Learning for Molecules
Machine Learning for Molecules
 
Morgan uw maGIV v1.3 dist
Morgan uw maGIV v1.3 distMorgan uw maGIV v1.3 dist
Morgan uw maGIV v1.3 dist
 
Apt thomas kelly
Apt thomas kellyApt thomas kelly
Apt thomas kelly
 
(PhD Dissertation Defense) Theoretical and Numerical Investigations on Crysta...
(PhD Dissertation Defense) Theoretical and Numerical Investigations on Crysta...(PhD Dissertation Defense) Theoretical and Numerical Investigations on Crysta...
(PhD Dissertation Defense) Theoretical and Numerical Investigations on Crysta...
 
Graphs, Environments, and Machine Learning for Materials Science
Graphs, Environments, and Machine Learning for Materials ScienceGraphs, Environments, and Machine Learning for Materials Science
Graphs, Environments, and Machine Learning for Materials Science
 
Materials Design in the Age of Deep Learning and Quantum Computation
Materials Design in the Age of Deep Learning and Quantum ComputationMaterials Design in the Age of Deep Learning and Quantum Computation
Materials Design in the Age of Deep Learning and Quantum Computation
 
Multivariate dimensionality reduction in cross-correlation analysis
Multivariate dimensionality reduction in cross-correlation analysis Multivariate dimensionality reduction in cross-correlation analysis
Multivariate dimensionality reduction in cross-correlation analysis
 
NNPDF3.1
NNPDF3.1NNPDF3.1
NNPDF3.1
 
Autonomous experimental phase diagram acquisition
Autonomous experimental phase diagram acquisitionAutonomous experimental phase diagram acquisition
Autonomous experimental phase diagram acquisition
 
Predicting local atomic structures from X-ray absorption spectroscopy using t...
Predicting local atomic structures from X-ray absorption spectroscopy using t...Predicting local atomic structures from X-ray absorption spectroscopy using t...
Predicting local atomic structures from X-ray absorption spectroscopy using t...
 
The Transformation of Systems Biology Into A Large Data Science
The Transformation of Systems Biology Into A Large Data ScienceThe Transformation of Systems Biology Into A Large Data Science
The Transformation of Systems Biology Into A Large Data Science
 
2018 KOSRO-Oral presentation-Wonjoong Cheon
2018 KOSRO-Oral presentation-Wonjoong Cheon2018 KOSRO-Oral presentation-Wonjoong Cheon
2018 KOSRO-Oral presentation-Wonjoong Cheon
 
Europy17_dibernardo
Europy17_dibernardoEuropy17_dibernardo
Europy17_dibernardo
 
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
Real-Time Analysis of Streaming Synchotron Data: SCinet SC19 Technology Chall...
 
What can we learn from molecular dynamics simulations of carbon nanotube and ...
What can we learn from molecular dynamics simulations of carbon nanotube and ...What can we learn from molecular dynamics simulations of carbon nanotube and ...
What can we learn from molecular dynamics simulations of carbon nanotube and ...
 
Fault tolerance in big data
Fault tolerance in big dataFault tolerance in big data
Fault tolerance in big data
 
A multi-sensor based uncut crop edge detection method for head-feeding combin...
A multi-sensor based uncut crop edge detection method for head-feeding combin...A multi-sensor based uncut crop edge detection method for head-feeding combin...
A multi-sensor based uncut crop edge detection method for head-feeding combin...
 

More from WonjoongCheon

3D isocenters quality assurance in radiation treatment room using a motion c...
3D isocenters quality assurance in radiation treatment room  using a motion c...3D isocenters quality assurance in radiation treatment room  using a motion c...
3D isocenters quality assurance in radiation treatment room using a motion c...WonjoongCheon
 
Time-resolved mirrorless scintillation detector @ KSMPRS2018
Time-resolved mirrorless scintillation detector @ KSMPRS2018Time-resolved mirrorless scintillation detector @ KSMPRS2018
Time-resolved mirrorless scintillation detector @ KSMPRS2018WonjoongCheon
 
3-D isotope position tracking system using portable gamma cameras; Feasibilit...
3-D isotope position tracking system using portable gamma cameras; Feasibilit...3-D isotope position tracking system using portable gamma cameras; Feasibilit...
3-D isotope position tracking system using portable gamma cameras; Feasibilit...WonjoongCheon
 
Lymphedema measurement using kinect volume reconstruction
Lymphedema measurement using kinect volume reconstructionLymphedema measurement using kinect volume reconstruction
Lymphedema measurement using kinect volume reconstructionWonjoongCheon
 
Development of gamma camera simulator using arduino capston design_yonseiuniv
Development of gamma camera simulator using arduino capston design_yonseiunivDevelopment of gamma camera simulator using arduino capston design_yonseiuniv
Development of gamma camera simulator using arduino capston design_yonseiunivWonjoongCheon
 
New Quality Assurance Method Using Motion Tracking for 6D Robotic Couches
New Quality Assurance Method Using Motion Tracking for 6D Robotic CouchesNew Quality Assurance Method Using Motion Tracking for 6D Robotic Couches
New Quality Assurance Method Using Motion Tracking for 6D Robotic CouchesWonjoongCheon
 
Prediction of patient respiratory signal using deep learning model: LSTM
Prediction of patient respiratory signal using deep learning model: LSTMPrediction of patient respiratory signal using deep learning model: LSTM
Prediction of patient respiratory signal using deep learning model: LSTMWonjoongCheon
 
Liver segmentation using U-net: Practical issues @ SNU-TF
Liver segmentation using U-net: Practical issues @ SNU-TFLiver segmentation using U-net: Practical issues @ SNU-TF
Liver segmentation using U-net: Practical issues @ SNU-TFWonjoongCheon
 
Time-resolved mirrorless scintillation detector @ AAPM2018
Time-resolved mirrorless scintillation detector @ AAPM2018Time-resolved mirrorless scintillation detector @ AAPM2018
Time-resolved mirrorless scintillation detector @ AAPM2018WonjoongCheon
 

More from WonjoongCheon (9)

3D isocenters quality assurance in radiation treatment room using a motion c...
3D isocenters quality assurance in radiation treatment room  using a motion c...3D isocenters quality assurance in radiation treatment room  using a motion c...
3D isocenters quality assurance in radiation treatment room using a motion c...
 
Time-resolved mirrorless scintillation detector @ KSMPRS2018
Time-resolved mirrorless scintillation detector @ KSMPRS2018Time-resolved mirrorless scintillation detector @ KSMPRS2018
Time-resolved mirrorless scintillation detector @ KSMPRS2018
 
3-D isotope position tracking system using portable gamma cameras; Feasibilit...
3-D isotope position tracking system using portable gamma cameras; Feasibilit...3-D isotope position tracking system using portable gamma cameras; Feasibilit...
3-D isotope position tracking system using portable gamma cameras; Feasibilit...
 
Lymphedema measurement using kinect volume reconstruction
Lymphedema measurement using kinect volume reconstructionLymphedema measurement using kinect volume reconstruction
Lymphedema measurement using kinect volume reconstruction
 
Development of gamma camera simulator using arduino capston design_yonseiuniv
Development of gamma camera simulator using arduino capston design_yonseiunivDevelopment of gamma camera simulator using arduino capston design_yonseiuniv
Development of gamma camera simulator using arduino capston design_yonseiuniv
 
New Quality Assurance Method Using Motion Tracking for 6D Robotic Couches
New Quality Assurance Method Using Motion Tracking for 6D Robotic CouchesNew Quality Assurance Method Using Motion Tracking for 6D Robotic Couches
New Quality Assurance Method Using Motion Tracking for 6D Robotic Couches
 
Prediction of patient respiratory signal using deep learning model: LSTM
Prediction of patient respiratory signal using deep learning model: LSTMPrediction of patient respiratory signal using deep learning model: LSTM
Prediction of patient respiratory signal using deep learning model: LSTM
 
Liver segmentation using U-net: Practical issues @ SNU-TF
Liver segmentation using U-net: Practical issues @ SNU-TFLiver segmentation using U-net: Practical issues @ SNU-TF
Liver segmentation using U-net: Practical issues @ SNU-TF
 
Time-resolved mirrorless scintillation detector @ AAPM2018
Time-resolved mirrorless scintillation detector @ AAPM2018Time-resolved mirrorless scintillation detector @ AAPM2018
Time-resolved mirrorless scintillation detector @ AAPM2018
 

Recently uploaded

Call Girls Service In Shyam Nagar Whatsapp 8445551418 Independent Escort Service
Call Girls Service In Shyam Nagar Whatsapp 8445551418 Independent Escort ServiceCall Girls Service In Shyam Nagar Whatsapp 8445551418 Independent Escort Service
Call Girls Service In Shyam Nagar Whatsapp 8445551418 Independent Escort Serviceparulsinha
 
Artifacts in Nuclear Medicine with Identifying and resolving artifacts.
Artifacts in Nuclear Medicine with Identifying and resolving artifacts.Artifacts in Nuclear Medicine with Identifying and resolving artifacts.
Artifacts in Nuclear Medicine with Identifying and resolving artifacts.MiadAlsulami
 
Call Girls Colaba Mumbai ❤️ 9920874524 👈 Cash on Delivery
Call Girls Colaba Mumbai ❤️ 9920874524 👈 Cash on DeliveryCall Girls Colaba Mumbai ❤️ 9920874524 👈 Cash on Delivery
Call Girls Colaba Mumbai ❤️ 9920874524 👈 Cash on Deliverynehamumbai
 
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls AvailableVip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls AvailableNehru place Escorts
 
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore EscortsCall Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escortsvidya singh
 
Low Rate Call Girls Pune Esha 9907093804 Short 1500 Night 6000 Best call girl...
Low Rate Call Girls Pune Esha 9907093804 Short 1500 Night 6000 Best call girl...Low Rate Call Girls Pune Esha 9907093804 Short 1500 Night 6000 Best call girl...
Low Rate Call Girls Pune Esha 9907093804 Short 1500 Night 6000 Best call girl...Miss joya
 
Call Girls Service Chennai Jiya 7001305949 Independent Escort Service Chennai
Call Girls Service Chennai Jiya 7001305949 Independent Escort Service ChennaiCall Girls Service Chennai Jiya 7001305949 Independent Escort Service Chennai
Call Girls Service Chennai Jiya 7001305949 Independent Escort Service ChennaiNehru place Escorts
 
Call Girls Service Navi Mumbai Samaira 8617697112 Independent Escort Service ...
Call Girls Service Navi Mumbai Samaira 8617697112 Independent Escort Service ...Call Girls Service Navi Mumbai Samaira 8617697112 Independent Escort Service ...
Call Girls Service Navi Mumbai Samaira 8617697112 Independent Escort Service ...Call girls in Ahmedabad High profile
 
Russian Escorts Girls Nehru Place ZINATHI 🔝9711199012 ☪ 24/7 Call Girls Delhi
Russian Escorts Girls  Nehru Place ZINATHI 🔝9711199012 ☪ 24/7 Call Girls DelhiRussian Escorts Girls  Nehru Place ZINATHI 🔝9711199012 ☪ 24/7 Call Girls Delhi
Russian Escorts Girls Nehru Place ZINATHI 🔝9711199012 ☪ 24/7 Call Girls DelhiAlinaDevecerski
 
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...astropune
 
VIP Mumbai Call Girls Hiranandani Gardens Just Call 9920874524 with A/C Room ...
VIP Mumbai Call Girls Hiranandani Gardens Just Call 9920874524 with A/C Room ...VIP Mumbai Call Girls Hiranandani Gardens Just Call 9920874524 with A/C Room ...
VIP Mumbai Call Girls Hiranandani Gardens Just Call 9920874524 with A/C Room ...Garima Khatri
 
(👑VVIP ISHAAN ) Russian Call Girls Service Navi Mumbai🖕9920874524🖕Independent...
(👑VVIP ISHAAN ) Russian Call Girls Service Navi Mumbai🖕9920874524🖕Independent...(👑VVIP ISHAAN ) Russian Call Girls Service Navi Mumbai🖕9920874524🖕Independent...
(👑VVIP ISHAAN ) Russian Call Girls Service Navi Mumbai🖕9920874524🖕Independent...Taniya Sharma
 
College Call Girls Pune Mira 9907093804 Short 1500 Night 6000 Best call girls...
College Call Girls Pune Mira 9907093804 Short 1500 Night 6000 Best call girls...College Call Girls Pune Mira 9907093804 Short 1500 Night 6000 Best call girls...
College Call Girls Pune Mira 9907093804 Short 1500 Night 6000 Best call girls...Miss joya
 
Bangalore Call Girls Nelamangala Number 7001035870 Meetin With Bangalore Esc...
Bangalore Call Girls Nelamangala Number 7001035870  Meetin With Bangalore Esc...Bangalore Call Girls Nelamangala Number 7001035870  Meetin With Bangalore Esc...
Bangalore Call Girls Nelamangala Number 7001035870 Meetin With Bangalore Esc...narwatsonia7
 
Bangalore Call Girls Majestic 📞 9907093804 High Profile Service 100% Safe
Bangalore Call Girls Majestic 📞 9907093804 High Profile Service 100% SafeBangalore Call Girls Majestic 📞 9907093804 High Profile Service 100% Safe
Bangalore Call Girls Majestic 📞 9907093804 High Profile Service 100% Safenarwatsonia7
 
Call Girls Cuttack Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Cuttack Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Cuttack Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Cuttack Just Call 9907093804 Top Class Call Girl Service AvailableDipal Arora
 
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort ServicePremium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Servicevidya singh
 
Call Girls In Andheri East Call 9920874524 Book Hot And Sexy Girls
Call Girls In Andheri East Call 9920874524 Book Hot And Sexy GirlsCall Girls In Andheri East Call 9920874524 Book Hot And Sexy Girls
Call Girls In Andheri East Call 9920874524 Book Hot And Sexy Girlsnehamumbai
 

Recently uploaded (20)

Call Girls Service In Shyam Nagar Whatsapp 8445551418 Independent Escort Service
Call Girls Service In Shyam Nagar Whatsapp 8445551418 Independent Escort ServiceCall Girls Service In Shyam Nagar Whatsapp 8445551418 Independent Escort Service
Call Girls Service In Shyam Nagar Whatsapp 8445551418 Independent Escort Service
 
Artifacts in Nuclear Medicine with Identifying and resolving artifacts.
Artifacts in Nuclear Medicine with Identifying and resolving artifacts.Artifacts in Nuclear Medicine with Identifying and resolving artifacts.
Artifacts in Nuclear Medicine with Identifying and resolving artifacts.
 
Call Girls Colaba Mumbai ❤️ 9920874524 👈 Cash on Delivery
Call Girls Colaba Mumbai ❤️ 9920874524 👈 Cash on DeliveryCall Girls Colaba Mumbai ❤️ 9920874524 👈 Cash on Delivery
Call Girls Colaba Mumbai ❤️ 9920874524 👈 Cash on Delivery
 
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls AvailableVip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
Vip Call Girls Anna Salai Chennai 👉 8250192130 ❣️💯 Top Class Girls Available
 
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore EscortsCall Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
Call Girls Horamavu WhatsApp Number 7001035870 Meeting With Bangalore Escorts
 
Low Rate Call Girls Pune Esha 9907093804 Short 1500 Night 6000 Best call girl...
Low Rate Call Girls Pune Esha 9907093804 Short 1500 Night 6000 Best call girl...Low Rate Call Girls Pune Esha 9907093804 Short 1500 Night 6000 Best call girl...
Low Rate Call Girls Pune Esha 9907093804 Short 1500 Night 6000 Best call girl...
 
Call Girls Service Chennai Jiya 7001305949 Independent Escort Service Chennai
Call Girls Service Chennai Jiya 7001305949 Independent Escort Service ChennaiCall Girls Service Chennai Jiya 7001305949 Independent Escort Service Chennai
Call Girls Service Chennai Jiya 7001305949 Independent Escort Service Chennai
 
Call Girls Service Navi Mumbai Samaira 8617697112 Independent Escort Service ...
Call Girls Service Navi Mumbai Samaira 8617697112 Independent Escort Service ...Call Girls Service Navi Mumbai Samaira 8617697112 Independent Escort Service ...
Call Girls Service Navi Mumbai Samaira 8617697112 Independent Escort Service ...
 
Escort Service Call Girls In Sarita Vihar,, 99530°56974 Delhi NCR
Escort Service Call Girls In Sarita Vihar,, 99530°56974 Delhi NCREscort Service Call Girls In Sarita Vihar,, 99530°56974 Delhi NCR
Escort Service Call Girls In Sarita Vihar,, 99530°56974 Delhi NCR
 
Russian Call Girls in Delhi Tanvi ➡️ 9711199012 💋📞 Independent Escort Service...
Russian Call Girls in Delhi Tanvi ➡️ 9711199012 💋📞 Independent Escort Service...Russian Call Girls in Delhi Tanvi ➡️ 9711199012 💋📞 Independent Escort Service...
Russian Call Girls in Delhi Tanvi ➡️ 9711199012 💋📞 Independent Escort Service...
 
Russian Escorts Girls Nehru Place ZINATHI 🔝9711199012 ☪ 24/7 Call Girls Delhi
Russian Escorts Girls  Nehru Place ZINATHI 🔝9711199012 ☪ 24/7 Call Girls DelhiRussian Escorts Girls  Nehru Place ZINATHI 🔝9711199012 ☪ 24/7 Call Girls Delhi
Russian Escorts Girls Nehru Place ZINATHI 🔝9711199012 ☪ 24/7 Call Girls Delhi
 
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
♛VVIP Hyderabad Call Girls Chintalkunta🖕7001035870🖕Riya Kappor Top Call Girl ...
 
VIP Mumbai Call Girls Hiranandani Gardens Just Call 9920874524 with A/C Room ...
VIP Mumbai Call Girls Hiranandani Gardens Just Call 9920874524 with A/C Room ...VIP Mumbai Call Girls Hiranandani Gardens Just Call 9920874524 with A/C Room ...
VIP Mumbai Call Girls Hiranandani Gardens Just Call 9920874524 with A/C Room ...
 
(👑VVIP ISHAAN ) Russian Call Girls Service Navi Mumbai🖕9920874524🖕Independent...
(👑VVIP ISHAAN ) Russian Call Girls Service Navi Mumbai🖕9920874524🖕Independent...(👑VVIP ISHAAN ) Russian Call Girls Service Navi Mumbai🖕9920874524🖕Independent...
(👑VVIP ISHAAN ) Russian Call Girls Service Navi Mumbai🖕9920874524🖕Independent...
 
College Call Girls Pune Mira 9907093804 Short 1500 Night 6000 Best call girls...
College Call Girls Pune Mira 9907093804 Short 1500 Night 6000 Best call girls...College Call Girls Pune Mira 9907093804 Short 1500 Night 6000 Best call girls...
College Call Girls Pune Mira 9907093804 Short 1500 Night 6000 Best call girls...
 
Bangalore Call Girls Nelamangala Number 7001035870 Meetin With Bangalore Esc...
Bangalore Call Girls Nelamangala Number 7001035870  Meetin With Bangalore Esc...Bangalore Call Girls Nelamangala Number 7001035870  Meetin With Bangalore Esc...
Bangalore Call Girls Nelamangala Number 7001035870 Meetin With Bangalore Esc...
 
Bangalore Call Girls Majestic 📞 9907093804 High Profile Service 100% Safe
Bangalore Call Girls Majestic 📞 9907093804 High Profile Service 100% SafeBangalore Call Girls Majestic 📞 9907093804 High Profile Service 100% Safe
Bangalore Call Girls Majestic 📞 9907093804 High Profile Service 100% Safe
 
Call Girls Cuttack Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Cuttack Just Call 9907093804 Top Class Call Girl Service AvailableCall Girls Cuttack Just Call 9907093804 Top Class Call Girl Service Available
Call Girls Cuttack Just Call 9907093804 Top Class Call Girl Service Available
 
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort ServicePremium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
Premium Call Girls Cottonpet Whatsapp 7001035870 Independent Escort Service
 
Call Girls In Andheri East Call 9920874524 Book Hot And Sexy Girls
Call Girls In Andheri East Call 9920874524 Book Hot And Sexy GirlsCall Girls In Andheri East Call 9920874524 Book Hot And Sexy Girls
Call Girls In Andheri East Call 9920874524 Book Hot And Sexy Girls
 

Error of Multileaf collimator prediction using recurrent neural network (LSTM)

  • 1. MLC error prediction using recurrent neural network (LSTM) Wonjoong Cheon1), Kim Seong Jung2), Youngyih Han3), Hyebin Lee4), Byung Jun Min4*), Heerim Nam4) 1) Department of Health Sciences and Technology, SAIHST, Sungkyunkwan University, Seoul, 06351, Korea. 2) Department of Computer Engineering, Yonsei University, 03722, Seoul, Korea. 3) Department of Radiation Oncology, Samsung Medical Center, Sungkyunkwan University School of Medicine, Seoul, 06351, Korea. 4) Department of Radiation Oncology, Kangbuk Samsung Hospital, Sungkyunkwan University School of Medicine, Seoul, 03181, Korea. 1 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 2. Contents I. Introduction II. Material & Method 1) Convert Rtplan to Expected position of MLC 2) Prediction Actual position of MLC using artificial neural network 3) Statistical analysis III. Results IV.Conclusion 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 3. Ⅰ. Introduction 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 4. Ⅰ. Introduction Multi-Leaf Collimator 160 MLC™ (Siemens) Agility™ (Elekta) • Intensity modulated radiation therapy (IMRT) technique uses MLCs for modifying the beam fluence in the same treatment field in order to improve the conformity of prescribed dose distribution around the tumor region. (Mundt & Roeske, 2005) • The volumetric modulated arc therapy (VMAT) technology is a novel delivery method that is capable of producing highly conformal dose distributions through concomitant optimization of MLC shapes, dose rate, and gantry speed.(Schreibmann et al., 2009) Millennium™ MLC (Varian) Figure resource: www.elekta.com, www.varian.com, www.healthcare.siemens.com 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 5. Actual MLC position Expected MLC position Ⅰ. Introduction MLC log files Dicom RT Eclipse™ Treatment Planning System, Varian Dynalog file Trilogy® System Linear Accelerator, Varian • MLC leaf position uncertainties directly affect the dose delivered to tumor targets and sensitive structures in IMRT. (Mu et al., 2008) • Actual multileaf collimator (MLC) position data can be tracked throughout a treatment delivery in the form of MLC log files. (Kerns, Childress, & Kry, 2014) Mismatch (=Errors) Figure resource: www.varian.com 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 6.  Radiation treatment plan (RTplan) - plans may contain fractionation information, and define external beams and/or brachytherapy  Expected position - Position of each leaf calculated and predefined by RTplan file  Actual position - Position of each leaf conducted and written to dynalog file - Containing machine errors Ⅰ. Introduction RT workflow 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 RT workflow
  • 7. Ⅰ. Introduction MLC positional error • Most of the studies on MLC positional error — both random and systematic errors were analyzed — have reported that the random errors were insignificant, while systematic errors have shown significant effects on dose distributions, even with only 1 mm of positional errors applied for MLC positions. (Nithiyanantham et al., 2015) • MLC position and velocity errors are affected by Friction and Gravity. (Wijesooriya, K., et al., 2005, Lee, Jeong-Woo, et al.,) • Varian said that “The leaves have to fight gravity” Figure reference : http://medicalphysicsweb.org/cws/article/research/44068 1 2 3 4 5 6 56 57 58 59 60 2 3 4 Friction Gravity 2 3 4 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 8. Ⅰ. Introduction Aim of this study • Prediction multi-leaf collimator position and velocity errors before radiation treatment using radiation treatment plan (RTplan) file and artificial neural network (RNN:LSTM). 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Treatment Planning Computer RT plan Batch Folder LinacView Computer Logfile Batch Folder Linac Console RT Plans Log files RT Plans Convert RTplan to Expected position of MLC MLC position expected Prediction Actual position of MLC Artificial Neural NetworkStatistical analysis Probability error function 1 2 3 Is this error acceptable? No Warning MLC error prediction RT workflow If I operate the this linear accelerator based on my RTplan file, will it work safely?
  • 9. Ⅱ. Material & Method 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 10. Ⅱ. Material & Method StepⅠ  Convert Radiation treatment plan to Expected position of MLC • Computer Language: Matlab • Parameters: Monitor unit per second, Dose rate, Gantry angle Step Ⅱ  Prediction Actual position of MLC using Artificial neural network • Computer Language: Python • Type of neural net: RNN: LSTM • Deep learning framework: Tensorflow (Google) • Parameters: MLC position (expected, actual), Gantry angle, collimator angle, Beam on/off sign Step Ⅲ  Statistical analysis :Generate Probability error function • Preprocess: Central limit theorem • Model: Single Gaussian distribution (mean, std) • Computer Language: Matlab
  • 11. Radiation treatment plan file Dynalog file (expected position) Ⅱ. Material & Method Convert RTplan to Expected position using Matlab 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 12. Ⅱ. Material & Method StepⅠ  Convert Radiation treatment plan to Expected position of MLC • Computer Language: Matlab • Parameters: Monitor unit, Dose rate Step Ⅱ  Prediction Actual position of MLC using Artificial neural network • Computer Language: Python • Type of neural net: RNN: LSTM • Deep learning framework: Tensorflow • Parameters: MLC position (expected, actual), Gantry angle, collimator angle, Beam on/off sign Step Ⅲ  Statistical analysis :Generate Probability error function • Preprocess: Central limit theorem • Model: Single Gaussian (mean, std) • Computer Language: Matlab
  • 13. • RNN’s ability to anticipate also makes them capable of surprising creativity. • You can ask them to predict which are the most likely next notes in a melody, then randomly pick one of these notes and play it. Then ask the net for the next most likely notes.  How it is work ? 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Ⅱ. Material & Method Prediction Actual position using Artificial neural network
  • 14.  Learning for prediction of actual position of MLC Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60 1480Dynalog 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Ⅱ. Material & Method Prediction Actual position using Artificial neural network  Window size: 7 time stamp  Gantry angle: Ga(t)  Collimator angle: Ca(t)  Beam on/off signal: B(t)  Leaf expected position: leafPosi(t)  Leaf actual position: Actual position of 2nd leaf Time: t=8 Ga(t)Ca(t) B(t) leafPosi(t)
  • 15.  Learning for prediction of actual position of MLC Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60 1480Dynalog 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Ⅱ. Material & Method Prediction Actual position using Artificial neural network  Window size: 7 time stamp  Gantry angle: Ga(t)  Collimator angle: Ca(t)  Beam on/off signal: B(t)  Leaf expected position: leafPosi(t)  Leaf actual position: Actual position of 2nd leaf Time: t=9 Ga(t)Ca(t) B(t) leafPosi(t)
  • 16.  Learning for prediction of actual position of MLC Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60 1480Dynalog 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Ⅱ. Material & Method Prediction Actual position using Artificial neural network  Window size: 7 time stamp  Gantry angle: Ga(t)  Collimator angle: Ca(t)  Beam on/off signal: B(t)  Leaf expected position: leafPosi(t)  Leaf actual position: Actual position of 2nd leaf Time: t=10 Ga(t)Ca(t) B(t) leafPosi(t)
  • 17.  Learning for prediction of actual position of MLC Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60 1480Dynalog 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Ⅱ. Material & Method Prediction Actual position using Artificial neural network  Window size: 7 time stamp  Gantry angle: Ga(t)  Collimator angle: Ca(t)  Beam on/off signal: B(t)  Leaf expected position: leafPosi(t)  Leaf actual position: Actual position of 2nd leaf Time: t=11 Ga(t)Ca(t) B(t) leafPosi(t)
  • 18.  Learning for prediction of actual position of MLC Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60 1480Dynalog 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Ⅱ. Material & Method Prediction Actual position using Artificial neural network  Window size: 7 time stamp  Gantry angle: Ga(t)  Collimator angle: Ca(t)  Beam on/off signal: B(t)  Leaf expected position: leafPosi(t)  Leaf actual position: Actual position of 2nd leaf Time: t=12 Ga(t)Ca(t) B(t) leafPosi(t)
  • 19.  Learning for prediction of actual position of MLC Header leaf 1 leaf 2 leaf 3 leaf 4 leaf 60 1480Dynalog 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Ⅱ. Material & Method Prediction Actual position using Artificial neural network  Window size: 7 time stamp  Gantry angle: Ga(t)  Collimator angle: Ca(t)  Beam on/off signal: B(t)  Leaf expected position: leafPosi(t)  Leaf actual position: Actual position of 2nd leaf Time: t=13 Ga(t)Ca(t) B(t) leafPosi(t)
  • 20. FC 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 𝑖𝑛𝑝𝑢𝑡𝐺𝑎𝑡𝑒(𝑡) = 𝜎(𝑊𝑥𝑖 ∙ 𝑥(𝑡) + 𝑊ℎ𝑖𝑇 ∙ ℎ(𝑡 − 1) + 𝑏𝑖) 𝑓𝑜𝑟𝑔𝑜𝑡𝐺𝑎𝑡𝑒(𝑡) = 𝜎(𝑊𝑥𝑓 ∙ 𝑥(𝑡) + 𝑊ℎ𝑓𝑇 ∙ ℎ(𝑡 − 1) + 𝑏𝑓) 𝑜𝑢𝑡𝑝𝑢𝑡𝐺𝑎𝑡𝑒(𝑡) = 𝜎(𝑊𝑥𝑜 ∙ 𝑥(𝑡) + 𝑊ℎ𝑜𝑇 ∙ ℎ(𝑡 − 1) + 𝑏𝑜) • LSTM structure • MLCLogNet structure 𝐿𝑜𝑛𝑔 𝑆ℎ𝑜𝑟𝑡 𝑡𝑒𝑟𝑚 𝑚𝑒𝑚𝑜𝑟𝑦 3 𝑐𝑒𝑙𝑙 𝐹𝑢𝑙𝑙𝑦 𝑐𝑜𝑛𝑛𝑒𝑐𝑡𝑒𝑑 𝑙𝑎𝑦𝑒𝑟 ℎ𝑖𝑑𝑑𝑒𝑛 dim = 20 o𝑢𝑡𝑝𝑢𝑡 dim = 1 Ⅱ. Material & Method Prediction Actual position using Artificial neural network  Learning for prediction of actual position of MLC Structure of learning frame
  • 21.  Validation of MLCLog Net Log 1 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Log 2 Log 3 Log 4 Log 175 Log 176 Log 178 Log 1 Log 2 Log 3 Training log set Test log set Ⅱ. Material & Method Prediction Actual position using Artificial neural network Log 30
  • 22. Ⅱ. Material & Method StepⅠ  Convert Radiation treatment plan to Expected position of MLC • Computer Language: Matlab • Parameters: Monitor unit, Dose rate Step Ⅱ  Prediction Actual position of MLC using deep learning • Computer Language: Python • Type of neural net: RNN: LSTM • Deep learning framework: Tensorflow • Parameters: MLC position (expected, actual), Gantry angle, collimator angle, Beam on/off sign Step Ⅲ  Statistical analysis :Generate Probability error function • Preprocess: Central limit theorem • Model: Single Gaussian (mean, std) • Computer Language: Matlab
  • 23. Ⅱ. Material & Method Statistical analysis:Generate probability error function 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 𝑷𝒐𝒔𝒊𝒕𝒊𝒐𝒏𝑬𝒓𝒓𝒐𝒓 𝒏, 𝒕 = 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑𝑃𝑜𝑠𝑖𝑡𝑖𝑜𝑛 𝑛, 𝑡 − 𝐴𝑐𝑡𝑢𝑎𝑙𝑃𝑜𝑠𝑖𝑡𝑖𝑜𝑛(𝑛, 𝑡) 𝑽𝒆𝒍𝒐𝒄𝒊𝒕𝒚𝑬𝒓𝒓𝒐𝒓 𝒏, 𝒕 = 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑛, 𝑡 − 𝐴𝑐𝑡𝑢𝑎𝑙𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦(𝑛, 𝑡) 𝐴𝑐𝑡𝑢𝑐𝑎𝑙𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑛, 𝑡 = )−𝐴𝑃 𝑡 + 2, 𝑛 + 8 ∗ 𝐴𝑃 𝑡 + 1, 𝑛 − 8 ∗ 𝐴𝑃 𝑡 − 1, 𝑛 + 𝐴𝑃(𝑡 − 2, 𝑛 𝑑𝑒𝑙𝑡𝑎 𝑡 ∗ 12 𝐸𝑥𝑝𝑒𝑐𝑡𝑒𝑑𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑛, 𝑡 = )−𝐸𝑃 𝑡 + 2, 𝑛 + 8 ∗ 𝐸𝑃 𝑡 + 1, 𝑛 − 8 ∗ 𝐸𝑃 𝑡 − 1, 𝑛 + 𝐸𝑃(𝑡 − 2, 𝑛 𝑑𝑒𝑙𝑡𝑎 𝑡 ∗ 12 Central limit theorem Central limit theorem [mm] [mm/s]
  • 24. Ⅲ. Results 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 25. StepⅠ  Convert Radiation treatment plan to Expected position of MLC Step Ⅱ  Prediction Actual position of MLC using Artificial neural network Step Ⅲ  Statistical analysis :Generate Probability error function Ⅲ. Results
  • 26. Footnote: Validation of calculated Expected position from RTplan Footnote: Comparison calculated expected position and expected position extracted from dynalog file (30 th leaf of MLC) Ⅲ. Results Convert RTplan to Expected position using Matlab 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 27. Plot the cost value of l2 loss function Prediction actual position of MLC using recurrent neural network (LSTM) Ⅲ. Results Prediction Actual position of MLC using Artificial neural network
  • 28. Plot the cost value of l2 loss function Prediction actual position of MLC using recurrent neural network (LSTM) Ⅲ. Results Prediction Actual position of MLC using Artificial neural network
  • 29. Ⅲ. Results Statistical analysis :Generate Probability error function [mm] [mm/s]
  • 30. Ⅰ. Introduction Aim of this study • Prediction multi-leaf collimator position and velocity errors before radiation treatment using radiation treatment plan (RTplan) file and artificial neural network (RNN:LSTM). 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Treatment Planning Computer RT plan Batch Folder RT PlansRT Plans Convert RTplan to Expected position of MLC MLC position expected Prediction Actual position of MLC Artificial Neural NetworkStatistical analysis Probability error function 1 2 3 Is this error acceptable? No Warning MLC error prediction [mm] [mm]
  • 31. Ⅳ. Conclusion 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 32. Ⅳ. Conclusion • Multi-leaf collimators have had a major impact on the development of radiation therapy such as IMRT, VMAT and IMPT. • Small defects of the multi-leaf collimator can have a large effect on the results of radiation therapy. • We build workflow for predicting mechanical errors of MLC before radiation treatment. • Consequently, our team developed a method of prediction multi-leaf collimator position and velocity errors before radiation treatment using radiation treatment plan (RTplan) file and artificial neural network (RNN:LSTM). 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 33. V. Further study • Actual dose prediction using recurrent neural network (LSTM) • Expected dose distribution • Actual dose distribution 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐
  • 34. Thank you for your attention. 2017/11/10 대한방사선수술물리연구회 제15차 학술대회 (2017.11.10) @ 서울산업진흥원 컨텐 Acknowledgement: This work was supported by the National Research Foundation of Korea(NRF) grant funded by the Korea government (MSIP) (No. NRF-2017R1C1B2011257)
  • 35. ERROR RANGE 1 Max acc 49.49 Max acc 55.65 With BCGWithout BCG
  • 36. ERROR RANGE 2 Max acc :84.28 Max acc :84.74
  • 37. Max acc :96.33 ERROR RANGE 3 Max acc :97.19
  • 38. Max acc :99.45 ERROR RANGE 4 Max acc :99.60
  • 39. Max acc : 1.00 ERROR RANGE 5 Max acc : 1.00

Editor's Notes

  1. 다엽콜리메이터, Multileaf collimator는 beam fluence에 변화를 시간에 따라 주면서dose conformity를 향상시켰습니다. Dose rate가 한 치료 내에서 변할 뿐만아니라 모든 각도를 사용하는 VMAT 방법의 경우에는 높은 dose conformit를 위해서 MLC 또한 정교한 움직임은 필수적입니다.
  2. 치료의 정확성 향상을 위해서 치료복잡도가 증가함에 따라 MLC leaf의 position uncertainty 또한 증가하게 되었습니다. 치료계획의 MLC 위치와 실제 장비에서 운영된 MLC의 위치를 확인하기 위해 가속기에서 움직인 파라미터들은 로그 형태로 기록됩니다. 가속기를 판매하고 있는 Varian 사의 경우에는 Dynalog라고 부르고 있고, 2000년대 초반에 Dynalog를 분석하는 연구가 많이 이뤄져왔습니다.
  3. 발표를 진행함에 있어서 자주 언급되는 단어들을 간단하게 정리해 보았습니다. Radiation treatment plan file, Rtplan 파일은 fraction , external beams 등 치료에 관련 된 정보를 가지고 있습니다. 이중 하나가 MLC의 segment 집합입니다. 생성된 Rtplan은 Linac consol을 통해서 장비로 들어가고, 장비는 치료를 수행하게 됩니다. 시간에 다른 장비 운영에 대한 기록을 담고 있는 Dynalog 파일이 장비에 의해서 생성됩니다. Dynalog 파일 내의 Expected positio은 Rtplan파일로 부터 치료수행 시간동안에 동작되길 바랬던 각 파라미터의 정보들이 들어있습니다. 반대로 Actual positio의 경우에는 치료수행 동안에 실제적으로 동작된 각 파라미터의 정보들이 들어있습니다. 즉 Machine error가 포함된 값을 가지고 있습니다.
  4. MLC에 발생할 수 있는 에러의 경우, Random error, systemetic error 그리고, 오른쪽에 2번 leaf을 기준으로 보았을 때, 근접된 MLC의 운동방향이 반대반향일 때 Friction에 의해서 영향을 받을 수 있고, Gantry angle이 90, 270도일 경우에 Gravity에 의해서 실제 위치는 영향을 받을 수 있습니다.
  5. 그래서 저희 연구팀은 MLC의 위치에러를 예측할 수 있는 방법에 대한 연구를 진행하였습니다. 기존의 Dynalog분석은 pre-treatment QA과정에서 장비운영 후 수행하는 사후 분석이지만, 둘간의 관계성이 명확한 Rtplan을 MLC expected position으로 변환하고, Expected positoin을 machine error가 포함된 actual position으로 변환 할 수 있는 deep learning network을 사용하여 에러가 포함 된 정보를 만들고 분석을 통해 사전에 예측하는 방법입니다.
  6. Rtplan과 expected position 사이의 관계는 명확합니다. 움직일 것으로 예상하는 계산값이기 때문입니다. 명확한 둘 사이의 관계를 Matlab으로 구현하여 변환을 시도하였습니다.
  7. 계산 된 expected position 을 인풋으로 machine error 를 포함한 actual positio을 예측하는 neural network 을 학습시켰습니다. 먼저 사용된 neural network에 대해서 설명드리겠습니다.
  8. Recurrent neural network은 예측을 하는 문제에 좋은 성능을 내고 있는 기본 네트워크 골격입니다. 작동방식은 4개의 음표가 주어졌을때, 초록색 음표를 예측하고, 예측 된 음표를 포함하여 다음 음표를 예측하는 방식으로 동작합니다. 해당 컨셉을 MLC에 적용해 보도록 하겠습니다.
  9. Dynalog를 그래픽컬 하게 시각화 하였습니다. Dynalog의 Header에서 gantry angle, collimator angle, beam on and off signal 과 7개의 time stamp 동안 Leaf들의 expected positoin을 이용하여 Actual position을 예측하게 됩니다.
  10. 그래서 저희 연구팀은 MLC의 위치에러를 예측할 수 있는 방법에 대한 연구를 진행하였습니다. 기존의 Dynalog분석은 pre-treatment QA과정을 통한 사후 분석이지만, 둘간의 관계성이 명확한 Rtplan을 MLC expected position으로 변환하고, Expected positoin을 machine error가 포함된 actual position으로 변환 할 수 있는 deep learning network을 사용하여 에러가 포함 된 정보를 만들고 분석을 통해 예측하는 방법입니다.