Skip to content

Image Metadata Annotation

sk edited this page Sep 9, 2019 · 3 revisions

Image Metadata Annotation

Members:

State the problem you worked on

We developed an ontology “Microscopy ontology” to explain image data and their metadata, especially electron microscopy images. Using our developed ontology, we described the metadata for the images including each image and the combined large image (Kobayashi et al, IJSWIS, 2018.). Further, we published the images and their metadata via the RIKEN platform (Kume et al, JIST2017.: http://clst.multimodal.riken.jp/RIKENImageDB/).

The quantification analysis and its comparison of imaging data are very important. We expect to analyze the metadata of images and their quantified data as an ontology-based open resource. Currently there are mainly two issues: (1) We manually add the metadata to each image. It requires time and effort. (2) Only metadata for each image is annotated. It is still difficult to biologically annotate the morphological structures and shapes in the image. It is due to no metadata description for ROI &/or object-shaped region and the supporting system of metadata annotation for the images using machine learning.

Give the state-of-the art/plan

We discuss the metadata description (phenotypes/morphology) for ROI &/or masked region. We try the development of the metadata annotation supporting system for the morphological structures in the images using the machine learning, especially deep learning.

Describe what you have done/results starting with The working group created...

We introduced the metadata description of object-shaped ROI and the relation with the phenotype data and analyzed data. Then, we performed the nucleus segmentation in mouse kidney images obtained by the electron microscopy imaging and the accuracy results for test dataset were more than 90%.

Write a conclusion

We could describe the metadata of object-shaped ROI related to the morphological structures using the annotation-supporting system. These metadata description indicated the metadata of object-shaped ROI related to the morphological structures using the annotation-supporting system and then the phenotype data and the analyzed data including the particle information such as area and roundary shape and so on. It leads to analyze the shape properties and theri comparison of morphological structures by gathering the metadata.

Write up any future work

We are going to publish these metadata in the RIKEN metadatabase and create the OWL/RDF data.

Figure


Clone this wiki locally