DeepAD: An automated and interpretable machine learning pipeline for image analyses of biomarkers in Alzheimer's disease
Project/Area Number |
22K15658
|
Research Category |
Grant-in-Aid for Early-Career Scientists
|
Allocation Type | Multi-year Fund |
Review Section |
Basic Section 51030:Pathophysiologic neuroscience-related
|
Research Institution | Institute of Physical and Chemical Research |
Principal Investigator |
|
Project Period (FY) |
2022-04-01 – 2026-03-31
|
Project Status |
Granted (Fiscal Year 2022)
|
Budget Amount *help |
¥4,420,000 (Direct Cost: ¥3,400,000、Indirect Cost: ¥1,020,000)
Fiscal Year 2025: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2024: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Fiscal Year 2023: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2022: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
|
Keywords | alzheimer's disease / amyloid / deep learning / image processing / microscopy / 画像情報処理・画像認識 / 分子・細胞・神経生物学 / 認知症疾患 / グリア細胞 / コンピュータビジョン |
Outline of Research at the Start |
Our goal is to create an automated tool to conduct image analyses common in Alzheimer's disease research, specifically: quantification of pathological Aβ plaques and glial cells. Automating analyses will reduce human error and bias, thereby improving the reproducibility of Alzheimer's disease research. The tool will be developed in collaboration with neuroscientists to ensure that it is easily understood and useful.
|
Outline of Annual Research Achievements |
研究の目的 To goal of the project is to develop an image processing pipeline that uses computational tools, including deep learning, to automate image analyses of Alzheimer's disease microscopy images.
研究実施計画 For FY2022 and FY2023, the 2 goals are to use computational tools to automatically preprocess microscopy images of AD mouse brain, and to segment, quantify, and conduct morphological feature analysis of Ab plaques. I have developed image preprocessing pipelines to preprocess 2D AD images from the original collaborator, as well as 3D images from a new collaborator. I developed an automated workflow for processing gene expression images, which was accepted as a peer-reviewed conference paper; the computational tools developed for this work can be applied to the 若手 project. In collaboration with other lab members, we developed novel methods to conduct instant segmentation of sparse, punctate objects; this work was submitted as for peer-review at 2 other conferences.
|
Current Status of Research Progress |
Current Status of Research Progress
3: Progress in research has been slightly delayed.
Reason
As planned, I have focused on preprocessing and plaque processing steps. However, with the addition of data from an additional collaborator, it is evident that image profiles from different labs can vary greatly. For example, images' SNR, background staining, 2D vs 3D, what cells/objects are stained for, etc. Because the goal is to create a generalized processing pipeline for AD microscopy images, it is now obvious that more time should be allocated to develop ways to improve generalization.
I had planned for 1 publication per year, but current results are too specific to the datasets. Instead, in collaboration with other lab members, we have submitted 3 peer-reviewed conference papers to describe methods that we have been developing, which are applicable for this project.
|
Strategy for Future Research Activity |
Data from more collaborators should be solicited earlier to test the generalizability of the pipeline. This will be done at the JNS conference this year.
I had planned for 1 publication per year. However, it now seems more effective to plan for a publication that summarizes the entire pipeline at the end of the project. If appropriate, findings from each FY can be submitted as peer-reviewed conference proceeding papers, as was done this year.
|
Report
(1 results)
Research Products
(6 results)