TL;DR
This paper introduces Back-illumination Interference Tomography (BIT), a novel microscopy technique that uses a demagnified, on-axis light source to generate high-contrast, label-free images of thick biological tissues. BIT achieves this by creating partially coherent back-illumination, enabling visualization of subtle refractive index differences, and shows promise for high-speed, 3D imaging of unprocessed tissues.
What Problem Does This Paper Address?
Imaging biological tissues, especially thick, unprocessed ones, without labels or stains is a significant challenge in biomedical research. Existing techniques like Optical Coherence Tomography (OCT) and confocal microscopy offer depth-sectioning, while phase contrast microscopy excels at detecting subtle refractive index variations. However, these methods often have limitations: OCT can be complex, confocal microscopy can struggle with thick scattering samples, and traditional phase contrast is best suited for thin specimens to avoid multiple scattering artifacts. Oblique Back-Illumination Microscopy (OBM) offers a simpler approach for turbid media but can be computationally intensive and requires multiple triggered sources, limiting its speed. There's a need for a technique that combines the label-free, high-contrast capabilities of phase contrast with the ability to image thick, scattering tissues at high speeds.
Methods
The core innovation is the Back-illumination Interference Tomography (BIT) technique. It employs an incoherent LED source that is demagnified and imaged below the focal plane of a high numerical aperture (NA) objective lens. This creates a small, semi-coherent source of backscattered light. Crucially, this backscattered light undergoes a phase inversion within the objective's narrow depth of field. This phase shift, combined with the partially coherent illumination, generates interference contrast that is sensitive to subtle refractive index differences in weakly scattering objects at the focal plane.
The optical setup involves a condensing lens to demagnify the LED, a beamsplitter to direct the light, and a high-NA objective. The backscattered light is then collected by the objective and imaged onto a CMOS sensor via a tube lens. The coherence factor ($\gamma$) of the illumination is critical and is defined by the ratio of the illumination NA to the objective NA ($\gamma = NA_{illum} / NA_{obj}$). A $\gamma < 1$ is necessary for phase contrast.
The paper also details:
- Scattering Particle Measurement and Simulation: To validate the contrast mechanism, TiO2 particles in a PDMS phantom were used. Experimental axial scans were compared to simulations based on a partially coherent optical transfer function model.
- Microfluidic Chambers: Used for in-vitro imaging of blood cells, allowing precise control over the coherence factor by adjusting the LED source position.
- In Vivo Imaging: The BIT microscope was inverted and mounted on an ophthalmic slit lamp for imaging blood cells in human tongue capillaries.
- Ex Vivo Tissue Imaging: The system was adapted for imaging bulk human tissue specimens, comparing BIT images to adjacent H&E stained sections.
Datasets
The paper does not explicitly mention training datasets for machine learning models. Instead, it utilizes:
- TiO2 particles in PDMS phantom: A custom-made scattering phantom to mimic tissue optical properties for initial characterization and simulation validation.
- Microfluidic chambers with blood cells: Used for in-vitro imaging of red blood cells (RBCs) and white blood cells (WBCs) under controlled conditions.
- Human blood cells in vivo: Images acquired from the ventral surface of a human tongue.
- Ex vivo human tissue specimens: Adjacent regions of pancreas and duodenum tissues were imaged using BIT and compared to conventional H&E stained sections.
Key Insights and Ideas
- Novel Contrast Mechanism: BIT leverages a unique partially coherent, on-axis back-illumination strategy to generate interference contrast from subtle refractive index variations, offering a different contrast source than conventional brightfield or even oblique back-illumination methods.
- Sensitivity to Weak Scatterers: The technique is particularly effective at visualizing weakly scattering objects, such as blood cells, which are challenging to image with high contrast in thick tissues using other label-free methods.
- Epi-mode for Thick Tissues: The epi-illumination geometry makes BIT suitable for imaging bulk, turbid tissues where transmission-based methods would be hindered by scattering.
- High-Speed Imaging Potential: The simplicity of the setup and the use of a standard LED source, combined with the fact that imaging speed is limited only by the sensor frame rate, suggests potential for very high-speed 3D imaging.
- Label-Free Histopathology: BIT's ability to image unprocessed, unstained bulk tissue with high contrast, comparable to H&E stained sections, opens doors for faster and potentially more cost-effective histopathology workflows and intraoperative assessment.
Key Results
- Scattering Particle Validation: Experimental results for TiO2 particles showed good agreement with simulations, demonstrating the expected axial intensity inversion behavior characteristic of partially coherent illumination.
- In Vitro Blood Cell Imaging: BIT successfully imaged RBCs and WBCs in microfluidic chambers, showing enhanced contrast for cell membranes and structures. Contrast was observed to drop off significantly when the coherence factor ($\gamma$) exceeded 0.1.
- In Vivo Blood Cell Imaging: High-speed (200 fps) videos of RBCs, WBCs, and platelets flowing through human tongue capillaries were acquired, demonstrating the technique's ability to resolve these components with high contrast in vivo.
- Ex Vivo Tissue Imaging: BIT provided high-resolution images of unstained bulk pancreas and duodenum tissues, resolving nuclei and tissue structures with contrast comparable to adjacent H&E stained sections, highlighting its potential for label-free histopathology.
Limitations and Future Work
- Depth Penetration: The authors acknowledge that spatial coherence will degrade and background scattered light will increase for deeper objects, limiting the effective imaging depth. Further research is needed to understand and optimize imaging quality and optical sectioning as a function of depth.
- Scattering Properties: The scattering properties of the medium are fundamental to the BIT mechanism, and optimization of contrast as a function of sample optical properties requires further investigation.
- Quantitative Phase Imaging: While BIT provides contrast, developing quantitative phase imaging capabilities with BIT, possibly by combining images from different positions, is an area for future work.
- AI Integration: The potential to train unsupervised generative AI models to convert BIT images into virtual H&E stained images is highlighted as a promising future direction, building on existing work in OBM.
- Combination with OBM: The authors suggest that combining BIT and OBM could further improve the capture of highly detailed cellular images.
Summary updated