| 研究課題/領域番号 |
23KF0296
|
|---|
| 研究種目 |
特別研究員奨励費
|
| 配分区分 | 基金 |
|---|
| 応募区分 | 外国 |
|---|
| 審査区分 |
小区分62010:生命、健康および医療情報学関連
|
|---|
| 研究機関 | 国立研究開発法人理化学研究所 |
|---|
研究代表者 |
Skibbe Henrik 国立研究開発法人理化学研究所, 脳神経科学研究センター, ユニットリーダー (00735764)
|
|---|
| 研究分担者 |
DECROOCQ MEGHANE 国立研究開発法人理化学研究所, 脳神経科学研究センター, 外国人特別研究員
|
|---|
| 研究期間 (年度) |
2023-11-15 – 2026-03-31
|
| 研究課題ステータス |
交付 (2023年度)
|
| 配分額 *注記 |
2,100千円 (直接経費: 2,100千円)
2025年度: 800千円 (直接経費: 800千円)
2024年度: 800千円 (直接経費: 800千円)
2023年度: 500千円 (直接経費: 500千円)
|
|---|
| キーワード | multi-scale cGAN / microscopy imaging / neuron tracing / generative AI |
|---|
| 研究開始時の研究の概要 |
Advancing neural morphology analysis through automated generation of realistic, annotated neural network images, enhancing deep learning applications in biomedical imaging.
|
| 研究実績の概要 |
We developed a new method leveraging conditional generative adversarial networks (cGANs) to generate diverse, high-resolution microscopy images for neuron tracing model training. The goal is to circumvent the limitations associated with the scarcity of annotated data for training machine learning models. The results were submitted as a full paper to the MIDL (Medical Imaging with Deep Learning) conference, where it has been accepted for presentation. The title of the paper is : "Multi-scale Stochastic Generation of Labelled Microscopy Images for Neuron Segmentation".
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
1: 当初の計画以上に進展している
理由
After only a few months into the project, we have already achieved a significant milestone with a paper accepted at a competitive, international conference. This early success demonstrates the effectiveness of our approach.
|
| 今後の研究の推進方策 |
We plan to advance our research by developing new tracing methods based on graph neural networks, which will incorporate global neuron topology into the learning process. Our goal is to seamlessly integrate these methods with our generative image synthesis algorithm, thereby creating a comprehensive pipeline for the automated analysis of neuron microscopy images. This integration is expected to significantly enhance the accuracy and efficiency of neuronal morphology analysis.
|
