| Project/Area Number |
07555127
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (B)
|
| Allocation Type | Single-year Grants |
|---|
| Section | 試験 |
|---|
| Research Field |
計測・制御工学
|
|---|
| Research Institution | Tokyo Institute of Technology |
|---|
Principal Investigator |
KOSUGI Yukio Tokyo Institute of Technology, Interdisciplinary Graduate School of Science and Engineering, Associate Professor, 大学院・総合理工学研究科, 助教授 (30108237)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
KAMEYAMA Keisuke Tokyo Institute of Technology, Interdisciplinary Graduate School of Science and, 大学院・総合理工学研究科, 助手 (40242309)
NISHIKAWA Junichi University of Tokyo, School of Medicine, Associate Professor, 医学部, 助教授 (00010322)
|
|---|
| Project Period (FY) |
1995 – 1996
|
| Project Status |
Completed (Fiscal Year 1996)
|
| Budget Amount *help |
¥8,000,000 (Direct Cost: ¥8,000,000)
Fiscal Year 1996: ¥600,000 (Direct Cost: ¥600,000)
Fiscal Year 1995: ¥7,400,000 (Direct Cost: ¥7,400,000)
|
|---|
| Keywords | PET / image fusion / neural networks / inverse problems / ill-posedness / MR images / dynamic regularization / blood-flow distribution / 血流分布 / Position Emission Tomography / Brain / MR images / Neural Networks / Partial Voiume Effect / Network Inversion / Fusion |
|---|
| Research Abstract |
Positron Emission Tomography (PET) will give important information in diagnosing intra-cranial metabolic disease as well as in investigating the functional mechanism of the brain, when we can get clear images.
Amongst variety of methods so far proposed, the MR-based deconvolution process might be one of the most important key-techniques in improving the PET imaging quality. In this research, we investigated the inverse problem of the deconvolving process within the framework of the network inversion technique, with respect to the neural network model for the partial volume effect arising in the positron emission tomography. To stabilize the inverse solution, we incorporate such a priori knowledge as the histological information given by MR images and the smoothness in the distribution of the blood-flow, which can be realized by the Tikhonov's regularization.
In particular, in this research we proposed a "dynamic regularization technique" in which the regularizing parameter should be changed according to each stage of the ongoing iterative optimization procedure. We developed the method and proved the effectiveness of the dynamic regularization in use with respect to a simple inverse problem, followed by the actual application to the brain PET image restoration process, with successful results demonstrated.
|
