Deep Learning for Microfluidic-Assisted Caenorhabditis elegans Multi-Parameter Identification Using YOLOv7.

Area of Science:

• * Biomedical research and genetics
• * Development of automated biological sorting systems

Background:

• * Caenorhabditis elegans (C. elegans) is a key model organism for human disease and genetics research due to its optical transparency.
• * Manual sorting of C. elegans populations is inefficient and labor-intensive for large-scale experiments.
• * Microfluidic platforms offer automated C. elegans sorting but require advanced identification for multi-parameter analysis.

Purpose of the Study:

• * To develop an automated deep learning model for accurate C. elegans identification and multi-parameter phenotyping.
• * To enhance the efficiency and precision of C. elegans sorting in microfluidic systems.

Main Methods:

• * Development of a deep learning model utilizing You Only Look Once (YOLO)v7 for C. elegans detection and recognition.
• * Training the model on a dataset of 3931 annotated C. elegans within microfluidic chips.
• * Evaluation of model performance against YOLOv5 and Faster R-CNN using mean average precision (mAP@0.5).

Main Results:

• * The YOLOv7 model achieved a 99.56% mAP@0.5 for automated C. elegans identification on the training set.
• * Demonstrated strong generalization with a 94.21% mAP@0.5 on an external validation set.
• * The model accurately identified multiple C. elegans phenotypes, including size, movement speed, and fluorescence.

Conclusions:

• * The developed deep learning model significantly improves the accuracy and efficiency of automated C. elegans identification.
• * This technology enhances multi-parameter sorting capabilities within microfluidic platforms.
• * The model has the potential to accelerate automated and integrated systems for biological research.