A human erythrocytes hologram dataset for learning-based model training.

Area of Science:

• Biomedical Optics
• Computational Imaging
• Microscopy

Background:

• Digital Holographic Microscopy (DHM) enables quantitative phase imaging.
• Accurate phase reconstruction is crucial for biological sample analysis.
• Existing datasets may not fully support advanced learning-based reconstruction methods.

Purpose of the Study:

• To present a comprehensive dataset of human red blood cell (RBC) holograms and their reconstructed phase maps.
• To facilitate the development and validation of learning-based models for aberration-free phase imaging.
• To support research in telecentric and non-telecentric DHM systems.

Main Methods:

• Experimental recording of holograms using an off-axis telecentric DHM system.
• Utilizing a 40 × /0.65NA microscope objective and a 200 mm tube lens.
• Augmenting an initial dataset of 300 holograms to 36,864 instances.

Main Results:

• A paired dataset of raw holograms and reconstructed phase maps for human RBCs is established.
• The dataset supports training and testing of end-to-end models for quantitative phase imaging.
• The DHM system ensures accurate quantitative phase maps through its off-axis, telecentric, and diffraction-limited design.

Conclusions:

• The presented dataset is a valuable resource for advancing digital holographic microscopy and computational imaging.
• It enables robust investigation of learning-based models for reconstructing phase images from DHM data.
• This work promotes progress in quantitative phase imaging techniques for biological applications.