2017 Fiscal Year Final Research Report
Methodological study of non-arbitrary quantification for fluorescent microscopy
| Project/Area Number |
16K12530
|
|---|
| Research Category |
Grant-in-Aid for Challenging Exploratory Research
|
| Allocation Type | Multi-year Fund |
|---|
| Research Field |
Life / Health / Medical informatics
|
|---|
| Research Institution | Tokyo Medical and Dental University |
|---|
Principal Investigator |
|
|---|
| Project Period (FY) |
2016-04-01 – 2018-03-31
|
| Keywords | バイオイメージインフォマティクス / 蛍光強度 / 要約統計量 / クラス分類 / 機械学習 / ベイズモデル |
|---|
| Outline of Final Research Achievements |
Non-arbitrary, non-biased measurement is an essential property for quantification of fluorescent microscopy images. Conventional methods of the image quantification has been heavily relied on human ability of cognition, and that means there are a plenty of room to errors and inconsistencies.
To eliminate these inconsistencies and to achieve the non-arbitrary quantification, I developed a non-supervised machine learning (ML) based method that utilized median and IQR (interquartile range) intensities of the fixed-size tiles. These image tiles were successfully classified by ML to few clusters. Selection of appropriate cluster can lead an extraction of particular group of tiles, which corresponds suitable set of biologically meaningful regions. The study indicated that ML-based classification combined with descriptive statistics values of tiling images, such as median and IQR, should be a non-arbitrary, non-biased quantification of the biological and medical images.
|
| Free Research Field |
発生生物学
|
