영상기반 딥러닝 의료 분야 응용 (kist 김영준) - 2017 대한의료영상학회 발표
TRANSCRIPT
영상기반 딥러닝 의료응용
2017.10.27
한국과학기술연구원 의공학연구소
김 영 준
Contents 연구소개
딥러닝 의료응용 사례
딥러닝 배우기
딥러닝 의료응용 (KIST 연구)
2
3
3D 의료영상 처리 S/W
4
▶ 3D 가상 수술
▶ 3D 환자모델링
▶ 3D 형상 모델 처리기술
3D Patient Modeling
5
3D Surgical Planning
6
7
Contents 연구소개
딥러닝 의료응용 사례
딥러닝 배우기
딥러닝 의료응용 (KIST 연구)
8
Google's AlphaGo Zero destroys humans all on its own
9
10
AlphaGo Zero
11
Google, 안과전문의수준의AI 기술 (JAMA, 2016.11)
12
Google, 안과전문의수준의AI 기술 (JAMA, 2016.11)
13
14
15
Deep Learning 기반 피부 질환 진단 (Nature, 2017.2)
Stanford Univ. / 21명 피부과 전문의와 진단 성능 동등함 증명
16
G. Litjens, et al. A Survey on Deep Learning in
Medical Image Analysis, arXiv, 2017
18
19
X.Wang et al. ChestX-ray8: Hospital-scale Chest X-ray Database and
Benchmarks on Weakly-Supervised Classification and Localization of
Common Thorax Diseases, CVPR, 2017
https://chuckyee.github.io/cardiac-segmentation/
D. Nie, et al. Medical Image Synthesis with Context-Aware
GAN (Generative Adversarial Networks), 2016
21
Z. Li, et al. Deep Learning based Radiomics (DLR) and its
usage in noninvasive IDH1 prediction for low grade glioma,
Scientific Reports, 2017
22
C.M. Deniz, et al. Segmentation of the Proximal Femur from MR
Images using Deep Convolutional Neural Networks, arXiv, 2017
23
24
25
[‘VUNO’ & 서울아산병원의 딥러닝 적용 폐질환 진단 소프트웨어]
[AI기반 의료영상 진단기업 ‘루닛’, 세계 100대 AI 기업 선정]
[AI기반 의료영상 진단기업 ‘Lunit’, 세계 100대 AI startup 선정]
Bone Age 측정
26
MGH
VUNO
NIH, 100,000장의 Chest X-ray 영상 데이터셋 공개
The dataset of scans is from more than 30,000 patients,
including many with advanced lung disease.
27
https://nihcc.app.box.com/v/ChestXray-NIHCC
Grand Challenges in Biomedical Image Analysis
https://grand-challenge.org/All_Challenges/
28
The Cancer Imaging Archive (TCIA) collections
http://www.cancerimagingarchive.net/
29
Contents 연구소개
딥러닝 의료응용 사례
딥러닝 배우기
딥러닝 의료응용 (KIST 연구)
30
31
Deep Learning Online Lecture
1. Coursera / Deep Learning Specialization (유료)https://www.coursera.org/specializations/deep-learning
2. Udacity / Deep Learninghttps://classroom.udacity.com/courses/ud730
3. Udacity / Deep Learning Nanodegree Foundation (유료)https://classroom.udacity.com/nanodegrees/nd101/syllabus/core-curriculum
4. 모두를 위한 딥러닝 – 기본적인 머신러닝과 딥러닝 강좌https://www.inflearn.com/course/기본적인-머신러닝-딥러닝-강좌/
5. CS231n: Convolutional Neural Networks for Visual Recognitionhttp://cs231n.stanford.edu/
Deep Learning Online Lecture
1. Coursera / Deep Learning Specializationhttps://www.coursera.org/specializations/deep-learning
34
Jupyter Notebook
http://jupyter.org/
The Jupyter Notebook is an open-source web application that allows you to
create and share documents that contain live code, equations, visualizations and
explanatory text. Uses include: data cleaning and transformation, numerical
simulation, statistical modeling, machine learning and much more.
Python
http://i-
systems.github.io/HSE545/machine%20learning%20all/Workshop/KSME/00_
basic_python.html
딥러닝 개요
38
https://www.slideshare.net/yongho/ss-79607172
Deep Learning Papers Reading Roadmap
https://github.com/songrotek/Deep-Learning-Papers-Reading-Roadmap
39
Contents 연구소개
딥러닝 의료응용 사례
딥러닝 배우기
딥러닝 의료응용 (KIST 연구)
40
AI기반 3D 의료용 S/W 기술
41
임상빅데이터
환자 3D 데이터
AI 진단
Classification
AI 치료계획
치료방법추천
AI 치료분석
수술결과분석
3D CNN 기반 Rotator Cuff Tear 진단
42
회전근개 MRI (fat suppression)
정상 / 파열 두 가지 클래스로 분류
정상 데이터 710명 (N=~2000), 파열 환자 1138명 (N=~2000)
데이터 프로세싱 및 진단 S/W
Dicom 파일 loading orientation 정보로 자동 정렬
Proximal humerus의 volume crop & interpolation (64x64x64)
Preprocessing & saving of labeled data
3D CNN 기반 Rotator Cuff Tear 진단
43
학습 Voxception-ResNet 네트워크 사용, epoch 110
Train set 3573개 (None-RCT : 1749, RCT : 1824) , Test set 200개
결과 Test set에서 약 95퍼센트 정확도 진단
학습 진행에 따른 정확도 및 히트맵 변화Voxception-ResNet
3D CNN 기반 Rotator Cuff Tear 진단
44
AI 기반 수술 계획 S/W
45
3D 수술계획 자동수립
Ex) 양악수술 자동 3D수술계획 추천, 랜드마크 자동 입력 및 분석
임상빅데이터
환자 3D 데이터
AI 진단
Classification
AI 치료계획
치료방법추천
AI 치료분석
수술결과분석
Deep Q Learning 기반 좌표계 정합
46
† 2017 ACDDE 국제학회, Best Paper Award
48
Chapter4. DICOM Viewer 제작 (고급 응용 프로그램 예제)
4-1 DICOM Viewer 소개 -124
4-2 프로젝트 생성 및 환경 설정 -128
4-3 4분할 윈도우 구성 -144
4-4 VTK Window 초기화 -154
4-5 DICOM 파일 읽기 168
4-6 Volume 데이터 읽기 및 렌더링 -206
부 록1 VTK 설치법 252
2 GDCM 설치법 -259
3 주요 DICOM 태그 -267
4 기타 VTK 프로그래밍 팁 268
Summary
딥러닝 의료응용 사례
딥러닝 배우기
딥러닝 의료응용 (KIST 연구)
Thank you!
50
Contact Information
Youngjun Kim, Ph.D.
- Center for Bionics, Korea Institute of Science and Technology
- E-mail: [email protected]
- Tel.: 02-958-5606
- C.P.: 010-5234-5378
Journal Paper Publication (2016~)1. S.W. Chung, …, Y. Kim, "Serial Changes in 3-Dimensional Supraspinatus Muscle Volume following Rotator Cuff Repair", The American Journal of Sports Medicine,
Vol. 45, No. 10, Aug. 2017
2. Y.D. Choi*, Y. Kim*, E. Park, “Patient-Specific Augmentation Rhinoplasty Using a Three-Dimensional Simulation Program and Three-Dimensional Printing”,
Aesthetic Surgery Journal, DOI: 10.1093/asj/sjx046, May, 2017
3. S. Kim, D. Lee, S. Park, K. Oh, S.W. Chung, Y. Kim, "Automatic segmentation of supraspinatus from MRI by internal shape fitting and autocorrection", Computer
Methods and Programs in Biomedicine, Vol. 140, pp. 165-174, Mar. 2017
4. Y. Kim, B.H. Lee, K. Mekuria, H. Cho, S. Park, J.H. Wang, D. Lee, "Registration accuracy enhancement of a surgical navigation system for anterior cruciate ligament
reconstruction: A phantom and cadaveric study", The Knee, S0968-0160(16)30248-4, Feb. 2017
5. C. Kyu Lee, Y. Kim, N. Lee, B. Kim, D.Y. Kim, S. Yi, "Feasibility study of utilization of action camera, GoPro Hero 4, Google glass and Panasonic HX-A100 in Spine
surgery", Spine, Vol. 42, No. 4, pp. 275-280, Feb. 2017
6. Q.C. Nguyen, Y. Kim, H. Kwon, "Optimization of layout and path planning of surgical robotic system", Int’l Journal of Control, Automation and Systems, Vol. 15, No.
1, pp. 375-384, Jan. 2017
7. Q.C. Nguyen*, Y. Kim*, S. Park, H. Kwon, "End-effector path planning and collision avoidance for robot-assisted surgical system", Int’l Journal of Precision
Engineering and Manufacturing, *These authors contributed equally to this work, Vol. 17, No. 12, Dec. 2016
8. B.H. Lee, D.H. Kum, I.J. Rhyu, Y. Kim, H. Cho, J.H. Wang, "Clinical advantages of image-free navigation system using surface-based registration in anatomical
anterior cruciate ligament reconstruction", Knee Surgery, Sports Traumatology, Arthroscopy, Vol. 24, No. 11, pp. 3556-3564, Nov. 2016
9. J.G. Seo*, S.M. Kim, J.M. Shin, Y. Kim*, B.H. Lee, "Safety of simultaneous bilateral total knee arthroplasty using an extramedullary referencing system: results from
2098 consecutive patients", Archives of Orthopaedic and Trauma Surgery, Vol. 136, No. 11, pp. 1615-1621, Nov. 2016
10. Y. Kim, Y.H. Na, L. Xing, R. Lee, S. Park, "Automatic deformable surface registration for medical applications by radial basis function-based robust point-matching",
Computers in Biology and Medicine, Vol. 77, No. 1, pp. 173-181, Oct. 2016
11. S.H. Park, S.W. Moon, B.H. Lee, S. Park, Y. Kim, D. Lee, S. Lim, J.H. Wang, "Arthroscopically blind anatomical anterior cruciate ligament reconstruction using only
navigation guidance: a cadaveric study", The Knee, DOI: 10.1016/j.knee.2016.02.020, Jul. 2016 (Epub ahead of print)
12. Y. Kim, W. Kim, D. Lee, "3D Inspection by Registration of CT and Dual X-ray Images", Journal of Int’l Society for Simulation Surgery, Vol. 3, No. 1, pp. 16-21, Jun.
2016
13. J.P. Yoon, S.W. Chung, J. Kim, H.S. Kim, H.J. Lee, W.J. Jeong, K.S. Oh, D.O. Lee, A. Seo, Y. Kim, "Intra-articular injection, subacromial injection, and
hydrodilatation for primary frozen shoulder: a randomized clinical trial", Journal of Shoulder and Elbow Surgery, Vol. 25, No. 3, pp. 376-383, Mar. 2016
14. J. Charton, L. Kim, Y. Kim, "Boolean operations by a robust, exact, and simple method between two colliding shells". Journal of Advanced Mechanical Design,
Systems, and Manufacturing, (Accepted)
15. E. Shim, Y. Kim, D. Lee, B.H. Lee, S. Woo, K. Lee, “2D-3D registration for 3D analysis of lower limb alignment in a weight-bearing condition”, Applied Mathematics –
Journal of Chinese Universities (Accepted)
51