Acceleration of Magnetic Resonance Imaging by Machine Learning
Project/Area Number |
17K00308
|
Research Category |
Grant-in-Aid for Scientific Research (C)
|
Allocation Type | Multi-year Fund |
Section | 一般 |
Research Field |
Intelligent informatics
|
Research Institution | Nagoya University (2019) Kyushu University (2017-2018) |
Principal Investigator |
|
Co-Investigator(Kenkyū-buntansha) |
實松 豊 九州大学, システム情報科学研究院, 准教授 (60336063)
久原 重英 杏林大学, 保健学部, 教授 (60781234)
|
Project Period (FY) |
2017-04-01 – 2020-03-31
|
Project Status |
Completed (Fiscal Year 2019)
|
Budget Amount *help |
¥4,550,000 (Direct Cost: ¥3,500,000、Indirect Cost: ¥1,050,000)
Fiscal Year 2019: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2018: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2017: ¥1,430,000 (Direct Cost: ¥1,100,000、Indirect Cost: ¥330,000)
|
Keywords | MRI / 深層学習 / 多重解像度 / ディープニューラルネットワーク / 脳動脈瘤 / U-net / 超解像 / 空間結合圧縮センシング / 検査時間短縮 / 機械学習 |
Outline of Final Research Achievements |
Shortening the shooting time of Magnetic Resonance Imaging (MRI) causes the deterioration of MRI image quality since only less information can be observed. We proposed a new deep neural network (DNN) in order to restore the image quality of the deteriorated image. The existing DNN for super-resolution is insufficient for restoring the image quality of the MRI images. We developed a new model of DNN by introducing the idea of multi-resolution. We prepared a lot of Brain MRI images of healthy persons and cerebral aneurysm patients. The proposed DNN was trained by using these data for a sufficiently long time. For the MRI images taken at 5x speed, the proposed DNN showed high performance in terms of the peak signal to noise ratio. The restored image attained good evaluation result of medical doctor’s interpretation.
|
Academic Significance and Societal Importance of the Research Achievements |
MRI撮影では患者は30分から1時間程度MRI装置の中で静止する必要があり精神的肉体的負担がかかる.また医療現場にかかる時間的負荷も大きく,MRI撮影の高速化が求められている.我々は,深層学習によるMRI高速化の研究にいち早く着手した.我々の提案したDNNは十分な画質を保ちつつ,約5倍の高速撮影が可能である.本手法は特許出願済みであり,企業ライセンシングを目指している.深層学習は,MRIの高速化のみならず,画像から病変を検出する医療診断支援の実用化が期待されている.本研究によって得られた成果と知見は,我々が現在研究している病変検出に役立っている.
|
Report
(4 results)
Research Products
(6 results)
-
-
-
-
-
-
[Patent(Industrial Property Rights)] 特許権2019
Inventor(s)
竹内純一,實松豊,川喜田雅則,北崎自然,久原重英
Industrial Property Rights Holder
竹内純一,實松豊,川喜田雅則,北崎自然,久原重英
Industrial Property Rights Type
特許
Industrial Property Number
2019-040207
Filing Date
2019
Related Report