Increasing a microscope's effective field of view via overlapped imaging and machine learning

Area of Science:

• Microscopy
• Biomedical Imaging
• Computational Pathology

Background:

• Automated analysis of morphological features is vital for biomedical research and disease diagnosis.
• Convolutional neural networks (CNNs) enhance cell counting accuracy but are limited by microscope throughput.
• Limited space-bandwidth product (SBP) of conventional microscopes hinders overall processing speed.

Purpose of the Study:

• To develop a multi-lens microscopic imaging system for high-efficiency automated specimen analysis.
• To overcome the throughput limitations of conventional microscopes in cell and feature counting.
• To demonstrate the system's capability for accurate detection and counting of diagnostically relevant features.

Main Methods:

• Implemented a multi-lens imaging system overlapping multiple fields of view onto a single sensor.
• Co-designed analysis software for processing overlapped image data.
• Validated the system through simulation and experimental analysis, including white blood cell and malaria parasite counting.

Main Results:

• Achieved accurate detection of diagnostically relevant features using overlapped imaging and analysis software.
• Demonstrated a multi-fold increase in detection and processing throughput.
• Maintained minimal reduction in accuracy compared to conventional methods.

Conclusions:

• The developed multi-lens microscopic imaging system significantly enhances automated specimen analysis throughput.
• Overlapped imaging combined with co-designed software offers a viable solution for high-efficiency biomedical research and diagnostics.
• This approach provides a practical method for accelerating cell counting and feature detection in microscopy.