Samplify: a versatile tool for image-based segmentation and annotation of seed abortion phenotypes.

Area of Science:

• Plant Science
• Computational Biology
• Agricultural Technology

Background:

• Automated seed phenotyping is crucial for research and agriculture.
• Arabidopsis thaliana seed phenotyping is challenging due to small seed size and high plant output.
• Manual methods are time-consuming, labor-intensive, and susceptible to user bias.

Purpose of the Study:

• To develop a scalable, automated pipeline for seed segmentation and classification.
• To address limitations of manual seed counting and analysis in Arabidopsis.
• To enable accurate quantification of seed abortion in plant studies.

Main Methods:

• Integration of classical image processing with Meta's Segment Anything Model for segmentation.
• Development of a random forest classifier using seed shape features for classification.
• Application of the pipeline to quantify seed abortion in Arabidopsis interploidy crosses (triploid block).

Main Results:

• Samplify effectively segments Arabidopsis seeds, including those in dense clusters where traditional methods fail.
• Automated classification of seeds into normal, partially collapsed, and fully collapsed categories.
• High segmentation and classification reliability demonstrated across multiple datasets.

Conclusions:

• Samplify offers a valuable, automated solution for seed segmentation and classification in plant research.
• The pipeline significantly reduces manual annotation workload and user bias.
• This tool enhances the efficiency and accuracy of phenotyping studies, particularly for Arabidopsis.