This document provides an overview of research topics in artificial intelligence and deep learning for medical image analysis and radiation treatment. It first introduces artificial intelligence and deep learning, including how artificial neural networks use deep learning to perform tasks by learning from large amounts of data. It then discusses specific deep learning techniques like convolutional neural networks. The document concludes by describing several of the author's research topics, including using deep learning for chest X-ray analysis, low-dose CT reconstruction, segmentation of organs from CT and MR images for radiation treatment planning, and isotope identification from gamma-ray spectra.
7. Deep Learning
Artificial Intelligence, Machine Learning and Deep Learning
7
1.
https://www.gettingsmart.com/2017/03/the-technologies-reshaping-life-and-livelihood/
8. Deep Learning
Artificial Neural Networks (인공 신경망)
8
1.
cs231n: Convolutional Neural Networks for Visual Recognition
The number of hidden layers
= Depth of the network
How can we learn the optimal 𝝎𝒊,𝒋,𝒌?
37. Research Topics
Chest X-ray
• Lunit Insight : https://www.youtube.com/watch?v=ZkWBVyNuE3A (1:20 - )
• Diagnosis of major lung diseases (식약처 의료기기 허가)
37
Medical Images
2.
기흉 폐결핵 폐암
https://insight.lunit.io/#examples
38. Research Topics
X-ray
• VUNO-MED BoneAge
• Software for measuring age of bones (식약처 의료기기 허가)
38
Medical Images
2.
https://www.youtube.com/watch?v=cI9nCVL40LM
39. Research Topics
Low-Dose CT
39
Medical Images
2.
Kang, et al. A Deep Convolutional Neural Network using Directional Wavelets for Low-dose X-ray CT Reconstruction
40. Research Topics
Low-Dose CT
40
Medical Images
2.
Kang, et al. A Deep Convolutional Neural Network using Directional Wavelets for Low-dose X-ray CT Reconstruction
41. Research Topics
Low-Dose CT
41
Medical Images
2.
Kang, et al. A Deep Convolutional Neural Network using Directional Wavelets for Low-dose X-ray CT Reconstruction
42. Research Topics
CT
• Liver and liver tumor segmentation
42
Medical Images
2.
Li, et al. H-DenseUNet: Hybrid densely connected UNet for liver and liver tumor segmentation from CT volumes
43. Research Topics
CT
• Liver and liver tumor segmentation
43
Medical Images
2.
Li, et al. H-DenseUNet: Hybrid densely connected UNet for liver and liver tumor segmentation from CT volumes
44. Research Topics
CT
• Liver and liver tumor segmentation
44
Medical Images
2.
Li, et al. H-DenseUNet: Hybrid densely connected UNet for liver and liver tumor segmentation from CT volumes
45. Research Topics
CT
• Liver and liver tumor segmentation
45
Medical Images
2.
Li, et al. H-DenseUNet: Hybrid densely connected UNet for liver and liver tumor segmentation from CT volumes
𝐷𝑆𝐶 𝐴, 𝐵 =
2 × |𝐴 ∩ 𝐵|
𝐴 + |𝐵|
46. Research Topics
Segmentation of the prostate and OAR in CT
46
Radiation Treatment
2.
Samaneh, et al. Segmentation of the prostate and organs at risk in male pelvic CT images using deep learning
bladder
prostate
rectum
47. Research Topics
Segmentation of the prostate and OAR in CT
• Prostate segmentation results
47
Radiation Treatment
2.
Samaneh, et al. Segmentation of the prostate and organs at risk in male pelvic CT images using deep learning
48. Research Topics
Segmentation of the prostate and OAR in CT
• Bladder, rectum segmentation results
48
Radiation Treatment
2.
Samaneh, et al. Segmentation of the prostate and organs at risk in male pelvic CT images using deep learning
50. Research Topics
MR-only Radiotherapy (MR-LINAC)
• Pseudo CT generation from MR images
50
Radiation Treatment
2.
Matteo, et al. Fast synthetic CT generation with deep learning for general pelvis MR-only Radiotherapy
51. Research Topics
MR-only Radiotherapy (MR-LINAC)
• Difference of dose calculation results from real CT and pseudo CT : -3 ~ 3 %
51
Radiation Treatment
2.
Matteo, et al. Fast synthetic CT generation with deep learning for general pelvis MR-only Radiotherapy
52. Research Topics
Isotope identification
• Input : Gamma-ray spectra (Contains up to 5 random isotopes among 33 isotopes in total)
• Target : Relative count contributions from each radioisotope
52
Radiation Detection
2.
M. Kamuda, et al. Automated Isotope Identification Algorithm Using Artificial Neural Networks