2022 Fiscal Year Final Research Report
Automatic white matter fiber bundle depiction using machine learning and construction of image support system for brain surgery patients
| Project/Area Number |
20K08016
|
|---|
| Research Category |
Grant-in-Aid for Scientific Research (C)
|
| Allocation Type | Multi-year Fund |
|---|
| Section | 一般 |
|---|
| Review Section |
Basic Section 52040:Radiological sciences-related
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
Suzuki Yuichi 東京大学, 医学部附属病院, 副診療放射線技師長 (70420221)
|
|---|
| Project Period (FY) |
2020-04-01 – 2023-03-31
|
| Keywords | MRI / 拡散強調画像 / 人工知能 / 脳疾患 |
|---|
| Outline of Final Research Achievements |
We investigated the effects of image quality deterioration and lack of data (examination interruption) in SMS (time reduction) technology used in diffusion weighted imaging on the visualization results of brain white matter automatic extraction software (TractSeg). It was found that the imaging time can be greatly shortened. We also adapted TractSeg to patients with cerebral arteriovenous malformations and evaluated its visualization ability.
We constructed a learning model (generation AI) for diffusion-weighted image generation that has not been imaged by deep learning DWI data and verified the accuracy of AI. This AI has the potential to cut scan time in half. Additionally, in the verification process, we optimized the application order of the motion probing gradient (MPG) related to the diffusion information acquisition direction during DWI data acquisition.
|
| Free Research Field |
放射線学
|
| Academic Significance and Societal Importance of the Research Achievements |
脳内の白質情報を可視化できる拡散強調画像の臨床応用が盛んであるが、生体内の複雑な情報を可視化するためには、多くのデータ収集が必要であり、比例して撮像時間が長くなり被検者への負担が増え、臨床現場での応用にも限界があった。今回既存のAI技術や自作したAIを用いることで、取得データ数を減らす（撮像時間を短縮する）場合でも、得られる結果が従来とほとんど変わることなく得られることがわかった。
これにより、被検者（患者）の負担を減少できるため社会的意義は大きい。また撮像時間が短縮できるため、臨床で従来より容易に応用できる環境となった。多くの疾患や病態の解明つながっていくことで学術的意義も大きいと言える。
|
