Computational Microscopy Methods for High-Throughput 3D Imaging and Analysis

dc.contributor.advisor

Horstmeyer, Roarke

dc.contributor.author

Kim, Kanghyun

dc.date.accessioned

2025-07-02T19:03:03Z

dc.date.available

2025-07-02T19:03:03Z

dc.date.issued

2025

dc.department

Biomedical Engineering

dc.description.abstract

This dissertation presents advancements in multi-camera array microscopy (MCAM) technology for high-throughput, large-field-of-view imaging applications across diverse domains. The MCAM leverages a densely packed array of micro-cameras to simultaneously capture unique sample areas, producing composite images with pixel counts significantly exceeding those of conventional microscopy systems. Experimental results are showcased for three configurations: (1) 3D imaging of objects across a 100 × 135 mm² field-of-view at 20 μm resolution (0.15 gigapixels per snapshot), (2) video capture over an 83 × 123 mm² field-of-view at 0.48 gigapixels per frame, and (3) high-resolution (2 μm) imaging of large histopathology specimens yielding 9.8 gigapixel composites.

MCAM system with the third configuration demonstrates particular utility in clinical cytopathology, where challenges in digitizing large, thick specimens require both high resolution and 3D imaging capabilities. A specialized Multi-Camera Array Scanner (MCAS) configuration is described, capable of scanning 54 × 72 mm² fields-of-view at 1.2 and 0.6 μm resolution. By capturing 624 megapixels per snapshot, this system offers rapid imaging compared to conventional whole-slide scanners. Applications include digitizing cytology slides in minutes and employing machine learning models for adenocarcinoma detection and slide-level classification, achieving up to 0.969 AUC performance.

Further, the MCAM technology is applied to organoid imaging, addressing the limitations of traditional microscopes like the EVOS system for large-scale experiments. The system enables high-throughput imaging of up to 48 organoids in a 96-well plate, capturing both morphological data and fluorescent GFP signals for longitudinal analyses. This capability enhances our understanding of organoid development and response to experimental conditions, such as morphogen exposure and genetic marker tracking.

Collectively, these results highlight the transformative potential of MCAM for wide-field, high-resolution imaging in biomedical research and clinical diagnostics.

dc.identifier.uri

https://hdl.handle.net/10161/32647

dc.rights.uri

https://creativecommons.org/licenses/by-nc-nd/4.0/

dc.subject

Optics

dc.subject

Medical imaging

dc.title

Computational Microscopy Methods for High-Throughput 3D Imaging and Analysis

dc.type

Dissertation

duke.embargo.months

11

duke.embargo.release

2026-06-07T16:47:00Z

Files

Collections