Automatic stage identification of Drosophila egg chamber based on DAPI images

Dongyu Jia, Qiuping Xu, Qian Xie, Washington Mio, Wu‐Min DengScientific 2016 · DOI

Abstract

The Drosophila egg chamber, whose development is divided into 14 stages, is a well-established model for developmental biology. However, visual stage determination can be a tedious, subjective and time-consuming task prone to errors. Our study presents an objective, reliable and repeatable automated method for quantifying cell features and classifying egg chamber stages based on DAPI images. The proposed approach is composed of two steps: 1) a feature extraction step and 2) a statistical modeling step. The egg chamber features used are egg chamber size, oocyte size, egg chamber ratio and distribution of follicle cells. Methods for determining the on-site of the polytene stage and centripetal migration are also discussed. The statistical model uses linear and ordinal regression to explore the stage-feature relationships and classify egg chamber stages. Combined with machine learning, our method has great potential to enable discovery of hidden developmental mechanisms.

Introduction: The Challenge of Manual Staging

In developmental biology, the Drosophila egg chamber is a cornerstone model for studying how cells grow, move, and differentiate. Its development is traditionally divided into 14 distinct stages. However, identifying these stages is a major bottleneck: it requires extensive expert training, is highly subjective, and is prone to human error—especially during "inter-stage" transitions where features are subtle.

Furthermore, scientists often rely on specific protein biomarkers to confirm a stage. If a researcher is studying a mutation that disrupts those very biomarkers, they lose their "biological clock." There is a clear need for a standardized, objective method to stage these samples based on morphology alone.

Methodology: An Automated Pipeline

Researchers have developed a two-step automated "toolbox" implemented in MATLAB to replace subjective manual staging with an objective, repeatable method.

1. Feature Extraction via Computer Vision

The first step involves extracting quantitative data from DAPI-stained images:

  • Segmentation: The pipeline uses Otsu’s method for initial thresholding and the Chan-Vese algorithm to precisely define the boundaries of egg chamber nuclei.
  • Morphometrics: Principal Component Analysis (PCA) determines the chamber's orientation and elongation ratio.
  • Biological Milestones: The code automatically detects specific developmental milestones, such as "blob-like" chromosomes in Stage 4 (using a watershed segmentation algorithm) and the migration of follicle cells toward the center in Stage 10B.
  • Uniformity: To distinguish between stages 8 and 9, the tool calculates the "$\Delta$-distance," a statistical measure of how uniformly follicle cells are distributed around the chamber.
Overview of Drosophila egg chamber developmental stages (1-14) with corresponding DAPI images and morphological identification criteria.

2. Statistical Modeling

The second step uses the extracted features for classification:

  • Linear Regression: Maps the relationship between developmental stage and physical features (like size).
  • Ordinal Regression: Employed for the actual classification. Unlike standard classification, ordinal regression recognizes the natural "growth order" of the data (e.g., Stage 3 must follow Stage 2), which significantly improves accuracy.

Key Insights and Results

The study utilized a dataset of 172 confocal microscopy images of DAPI-stained Drosophila egg chambers, pre-labeled by skilled biologists to provide a ground truth.

  • The Exponential Growth Rule: The study found a remarkably strong linear relationship ($R^2 = 0.95$) between the developmental stage and the logarithm of the egg chamber size, confirming that egg chambers grow at a constant exponential rate throughout most of oogenesis.
  • Staging Without Biomarkers: The tool successfully identified Delta mutant egg chambers as Stage 8 based purely on geometry, even though traditional biomarkers like the protein Broad were suppressed.
  • Settling Biological Debates: By precisely staging chambers and checking for the mitotic marker PH3, the researchers confirmed that mitosis ceases at Stage 6, marking the transition to the "endocycle."

Limitations and Future Work

The current iteration of the tool analyzes 2D cross-sections. Future work aims to incorporate full 3D "z-stacks" to improve accuracy and automate the initial image cropping phase for high-throughput batch processing. While the study focused on Stages 2 through 12, the authors note that Stages 1, 13, and 14 are generally the easiest for humans to identify manually due to their unique morphological extremes.

By providing an objective, repeatable method for quantifying cell features, this pipeline has the potential to standardize results across laboratories and enable the discovery of previously hidden developmental mechanisms.

Summary updated

Abstract

The Drosophila egg chamber, whose development is divided into 14 stages, is a well-established model for developmental biology. However, visual stage determination can be a tedious, subjective and time-consuming task prone to errors. Our study presents an objective, reliable and repeatable automated method for quantifying cell features and classifying egg chamber stages based on DAPI images. The proposed approach is composed of two steps: 1) a feature extraction step and 2) a statistical modeling step. The egg chamber features used are egg chamber size, oocyte size, egg chamber ratio and distribution of follicle cells. Methods for determining the on-site of the polytene stage and centripetal migration are also discussed. The statistical model uses linear and ordinal regression to explore the stage-feature relationships and classify egg chamber stages. Combined with machine learning, our method has great potential to enable discovery of hidden developmental mechanisms.
BibTeX
@inproceedings{jia2016automatic,
  author = {Dongyu Jia and Qiuping Xu and Qian Xie and Washington Mio and Wu‐Min Deng},
  title = {{Automatic stage identification of Drosophila egg chamber based on DAPI images}},
  booktitle = {Scientific},
  year = {2016},
  doi = {10.1038/srep18850},
}
Figure