Abstract
In vitro inner ear organoids gene therapy has advanced medical development, yet extracting inner ear primary cells (IEP) remains complex and needs to scarify mice. IEP cells contain multiple cell types, including progenitor cells, which can differentiate into hair-progenitor or neuro-progenitor cells. Accurate cell detection is essential to understand cell interactions and optimize culture conditions, ultimately reducing the sacrifice of mice. Current commercial software counts only one cell type at a time, often making errors in aggregated regions and taking at least five minutes per count. This paper proposes the Inner Ear Multiple Primary Cell Type Detection System (IEP-CDS), which addresses data limitations through IEP augmentation, overcomes detection errors in aggregated regions using preprocessing methods for training YOLO models, and achieves an F1-score greater than 20% compared to commercial software, completing counts in under one second. Furthermore, IEP-CDS provides rare IEP image data with expert labels. Overall, IEP-CDS effectively improves the efficiency of IEP detection in developing cell therapy for inner ear organoids.