Intelligent sort-timing prediction for image-activated cell sorting.

Area of Science:

• Biotechnology
• Artificial Intelligence
• Microfluidics

Background:

• Intelligent image-activated cell sorting (iIACS) integrates AI for high-throughput, image-based single-cell sorting.
• Accurate sort-timing prediction is critical in iIACS due to processing latency, which can be affected by cell flow speed fluctuations.

Purpose of the Study:

• To develop and demonstrate a machine learning technique to enhance sort-timing prediction accuracy in iIACS.
• To improve the sort event rate, yield, and purity of cell sorting.

Main Methods:

• Trained a machine learning algorithm using cell morphology, position, and flow speed to predict sort timing.
• Applied the algorithm to morphologically heterogeneous budding yeast cells within an iIACS system.

Main Results:

• The developed algorithm achieved 41.5% lower prediction error compared to methods relying solely on flow speed.
• This enhanced prediction accuracy is projected to increase the sort event rate of iIACS by approximately twofold.

Conclusions:

• Machine learning-based sort-timing prediction significantly improves iIACS performance.
• The proposed technique offers a pathway to more efficient and accurate AI-driven cell sorting.