Slice-Inference-Assisted Lightweight Small Object Detection Model for Holographic Digital Immunoassay Quantification

Area of Science:

• Biosensing
• Deep Learning
• Holographic Imaging

Background:

• Portable, cost-effective detection is vital for food safety, environmental monitoring, and clinical diagnostics.
• Existing methods like ELISA can be limited in sensitivity and detection range for certain analytes.

Purpose of the Study:

• To develop a portable, sensitive, and cost-effective holographic biosensor for chloramphenicol quantification in food samples.
• To integrate a sliced inference-assisted lightweight small object detection model (SIALSO) for enhanced detection accuracy and efficiency.

Main Methods:

• Lens-free holographic imaging system combined with a lightweight deep learning model (SIALSO).
• SIALSO model utilizes sliced inference to improve small object detection and reduce computational complexity.
• Digital immunoassay for quantifying chloramphenicol using microsphere probes.

Main Results:

• SIALSO biosensor demonstrated a linear detection range of 50 pg/mL to 100 ng/mL (R² = 0.986).
• Achieved higher sensitivity and broader detection range compared to ELISA.
• SIALSO model reduced computational parameters by 29% versus YOLOv5s, with 98.2% precision and 95.7% recall.

Conclusions:

• The SIALSO holographic biosensor provides a robust platform for sensitive and efficient food safety analysis.
• This technology lays the groundwork for developing advanced portable detection devices for various monitoring applications.
• The integration of holographic imaging and deep learning offers a promising approach for digital immunoassays.