- Medical physicst/AI researcher interested in application for radiation therapy
- Now working with physics residency
- Educated and trained in clinical research at Sungkyunkwan university and Samsung medical center
- Two bachelor degrees: Computer engineering, Radiological science
Sungkyunkwan University, Republic of Korea
Dept. Samsung Advanced Institute for Health Sciences & Technology (SAIHST)
Medical Physics Lab. (Major advisor: Prof. Youngyih Han, youngyih@skku.edu)
Yonsei University, Republic of Korea
Dept. Computer Engineering and Communication and Information Engineering.
Bachelor's degree (Major advisor: Prof. Jaekwon Kim)
Summer internship, Dept. Medical engineering
- Lymphedema measurement using KINECT volume reconstruction method [link]
Excellent Review Paper Award:
Review paper: Deep-learning in Radiation Oncology [link]
Oral presentation Award:
3D isocenters quality assurance in radiation treatment room using a motion capture camera system [link]
Oral presentation Award:
Two-dimensional time-resolved mirrorless scintillation detector for pretreatment QA [link]
Oral presentation Award:
Prediction of the patient respiratory signal using deep learning model: LSTM [link]
Young Investigation Award:
Development of a mirrorless compact scintillation detector
3rd prize:
Development of Gamma camera simulator using Arduino [link]
The project of liver segmentation has been studied and presented at SNU-TF (Seoul university-Tensorflow Korea).
The LiTs Competition data was used for training network. Backbone architecture is "U-net" [link]
The project of patient respiratory signal prediction using Long Short Term Memory (LSTM) was studied and presented at Korea society of medical physics (KSMP), Spring, 2018.
The respiratory signal was acquired from 4D CT data. Backbone architecture is LSTM.
In this study, LSTM method was compared with Multi-later perceptron and Decision Tree. [link]
Status: Under description for research paper
The project of "Deep learning application to patient-specific organs at risk auto-segmentation" was studied and presented at KOSRO (Korea society for radiation oncology), Fall, 2018.
Reduced field or cone down field techniques utilizing multiple CT scans are commonly used for adaptive radiation therapy (ART) in radiation oncology. In this process, not only the target but also the organs at risk (OAR) need to be manually delineated at each CT scans. In this study, a deep learning method is investigated to automatically segment the OARs in the cone down CTs without Bigdata [link]
Status: Under description for research paper
Accurate dose measurement of the penumbra region is essential to measure the dose distribution most closer to the true dose distribution using a scintillation detector. However, the in-house developed scintillation detector has shorter penumbra width. Therefore, the accumulated dose distribution measured by the scintillation detector contains errors.
The objective of this study is to recover true dose distribution from the measured dose distribution using a convolution neural network named PenumbraNet.
Status: Medical Physics (major revision)
In this project, deep-learning based super-resolution technique was applied to stereo portable gamma camera (SPGC) system.
The SPGC system could track the position of radiation source as 3D coordiates (x,y,z).
The developed super-resolution technique improved the tracking accuracy of SPGC system for in vivo proton range verification
Status: Journal of Korea Physics of Society (minor-revision)
In this project, we proposed a new method to perform pQA by predicting the delivered dose distribution in the water phantom using actual machine parameters of Dynalog file for all timesteps through the fluence to dose network (FDNet).
Status: Journal of Korea Physics of Society (submission)
The project of Proton nozzle modeling was carried out by J.Lee.
I supported this project for
Automatic 6-D robotic couch QA was performed using Visual Tracking System (VTS).
The VTS was consist of four Bonita B10 infra-red camera, which could tracking Infra-reflective (IR) markers as 3D coordinates.
For measuring IR marker in treatment room coordinate system, coordinate matching algorithm was applied between treatment room coordinate system and VTS coordinate system. [link] [paper]
3D isocenter coincidence QA was performed using VTS and in-house phantom named Eagle.
The isocenters which exist in radiation treatment room should be a single point, but it is hard to defined as a single point.
The aim of this study is to define the 3D position of each isocenter as 3D Cartesian coordinates. [link]
Status: Physica Medica (submission)
[1] Education seminar of Geant4
[2] Modeling of LINAC.
[3] Modeling of Proton therapy nozzle
[4] Modeling of Dual Gamma camera
[5] Modeling of Scintillator
[1] Education seminar of GATE [link] [code]
[2] Modeling of CBCT
[3] 3D dose calculation on CT image
[4] GPU calculation using GATE
-Sequential data
[1] Prediction of respiratory signal
[2] Prediction of actual position of MLC
-Medical image data
[3] Segmentation for Liver on CT img
[4] Segmentation for OAR
-Dose distribution data
[5] Deconvolution for 2D scintillation detector
[6] Dose prediction from fluence map
-2D radiographic image
[7] Super-resolution
[1] Design hexagonal collimator of gamma camera for monte-carlo simulation
[2] Design multileaf collimator for monte-carlo simulation