Rapid 3D imaging at cellular resolution for digital cytopathology with a multi-camera array scanner (MCAS)

Area of Science:

• Biomedical Engineering
• Computational Pathology
• Digital Pathology

Background:

• Optical microscopy is standard for cytopathology diagnosis.
• Whole slide scanners are slow, expensive, and not widely available.
• Cytology specimens require 3D capture due to their large area and thickness.

Purpose of the Study:

• Introduce a novel parallelized microscope for rapid, wide-field 3D scanning of thick specimens.
• Develop machine learning software for automated analysis of digitized cytology slides.
• Enhance diagnostic efficiency and accessibility in cytopathology.

Main Methods:

• Developed a Multi-Camera Array Scanner (MCAS) with 48 micro-cameras for parallel imaging.
• Achieved wide fields-of-view (54 × 72 mm²) at high resolutions (1.2 and 0.6 μm).
• Implemented machine learning models for adenocarcinoma detection and lung smear classification.

Main Results:

• Digitized entire cytology samples in 3D within minutes.
• MCAS captures 624 megapixels per snapshot, significantly faster than conventional scanners.
• Demonstrated a 0.73 recall for adenocarcinoma detection and 0.969 AUC for slide-level classification.

Conclusions:

• The MCAS system offers a faster, more efficient alternative to conventional whole-slide scanners.
• Machine learning integration aids pathologists in diagnosing complex cytology specimens.
• This technology has the potential to improve accessibility and accuracy in cytopathology.