MicrAnt: Towards Regression Task Oriented Annotation Tool for Microscopic Images

• Published in: Combinatorial Image Analysis Vol. 12148; pp. 209 - 218
• Main Authors: Jirik, Miroslav, Moulisova, Vladimira, Schindler, Claudia, Cervenkova, Lenka, Palek, Richard, Rosendorf, Jachym, Arlt, Janine, Bolek, Lukas, Dejmek, Jiri, Dahmen, Uta, Jirikova, Kamila, Gruber, Ivan, Liska, Vaclav, Zelezny, Milos
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer International Publishing AG 2020; Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Annotation; Decellularization; Liver; Microscopy; Scaffold
• ISBN: 9783030510015; 3030510018
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-030-51002-2_15

Annotating a dataset for training a Supervised Machine Learning algorithm is time and annotator’s attention intensive. Our goal was to create a tool that would enable us to create annotations of the dataset with minimal demands on expert’s time. Inspired by applications such as Tinder, we have created an annotation tool for describing microscopic images. A graphical user interface is used to select from a couple of images the one with the higher value of the examined parameter. Two experiments were performed. The first compares the speed of annotation of our application with the commonly used tool for processing microscopic images. In the second experiment, the texture description was compared with the annotations from MicrAnt application and commonly used application. The results showed that the processing time using our application is 3 times lower and the Spearman coefficient increases by 0.05 than using a commonly used application. In an experiment, we have shown that the annotations processed using our application increase the correlation of the studied parameter and texture descriptors compared with manual annotations .