2022 Fiscal Year Final Research Report
Feasibility study of lung nodule screening using pulmonary MRI and development of computer-assisted detection software
| Project/Area Number |
20K08045
|
|---|
| 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 |
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
越野 沙織 順天堂大学, 医学部, 非常勤助手 (50801552)
|
|---|
| Project Period (FY) |
2020-04-01 – 2023-03-31
|
| Keywords | 肺MRI / 肺がん / スクリーニング / 検診 / CAD / AI / 胸水 |
|---|
| Outline of Final Research Achievements |
Pulmonary MRI with ultrashort echo time (UTE) is expected to be an alternative to low-dose CT for lung screening. The feasibility of breath-hold (BH) and respiratory triggered (RT) scans with UTE was investigated. Both results of qualitative and quantitative assessment suggested that a solid nodule was more clearly depicted than a ground-glass nodule. BH scan in pulmonary MRI should be more suitable for lung screening than RT scan.
We prepared a lung and bronchus region dataset and a lung nodule dataset to develop computer-assisted detection (CADe) software for lung nodule screening. We have developed an automatic extraction method of both lung areas and bronchi regions simultaneously using deep learning, and accurate extraction were achieved. Now, we are developing (CADe) software for lung nodule screening.
Chest MRI was highly sensitive in detecting pleural fluid, which was considered mainly physiological. It was relevant to other clinical factors.
|
| Free Research Field |
MRI
|
| Academic Significance and Societal Importance of the Research Achievements |
本研究により，肺MRIによる肺結節スクリーニングは実現可能であることが示された．呼吸停止下撮像が第一選択であり，時間に余裕があれば呼吸同期撮像を追加するのが望ましい．現状では，すりガラス状結節は充実性結節より検出が難しく，検出能の向上のために撮像技術の進歩などによる改善が望まれる．
本研究で開発した手法により，肺MRIでの肺野および気管支領域の自動抽出が可能となった．肺結節スクリーニングのためのコンピュータ支援検出ソフトウェアの開発を進めた．初期のソフトウェアでは期待していたほどの肺結節検出能が得られなかったが，学習症例数を追加することで性能改善が期待される．
|
