Principal component analysis-based three-dimensional quantitative characterization of cracks on perforated embryo...

Area of Science:

• Biotechnology
• Materials Science
• Computational Biology

Background:

• Vaccine production via chicken embryo cultivation requires precise inoculation.
• Perforation-induced cracks in eggshells can compromise cultivation, necessitating accurate screening.
• Existing 2D methods for crack detection lack precision and fail to assess crack acceptability.

Purpose of the Study:

• To develop a novel global-local 3D quantitative characterization framework for evaluating cracks in perforated embryo eggs.
• To improve the accuracy and reliability of identifying and assessing the acceptability of cracks induced during egg perforation.
• To transition egg screening from simple crack presence detection to a more sophisticated crack acceptability evaluation.

Main Methods:

• Image processing and 3D reconstruction to generate global and local point clouds of cracks.
• Principal Component Analysis (PCA) for global quantitative characterization, extracting 7 geometric feature parameters.
• Random Forest algorithm for crack shape identification and local quantitative characterization of complex cracks.

Main Results:

• Achieved precise 3D reconstruction and quantitative characterization of perforation-induced cracks.
• Developed a crack identification model with 99.19% mean accuracy and 98.91% macro F1-Score.
• Generated quantitative parameters to support objective crack acceptability evaluation.

Conclusions:

• The proposed 3D framework enables accurate, quantitative assessment of crack acceptability in perforated embryo eggs.
• This methodology enhances the reliability of screening processes in vaccine production, reducing cultivation failures.
• Facilitates potential automated screening of embryo eggs based on defined crack acceptability criteria.