TL;DR
This paper introduces FlatScope 2.0, a novel, thin, lensless microscope capable of high-resolution, wide-field in vivo fluorescence imaging. By combining optimized phase masks, computational reconstruction, and integrated illumination/filtering, it overcomes previous limitations of lensless microscopy in biological contexts, enabling unprecedented views of neural activity in freely behaving animals.
What Problem Does This Paper Address?
Understanding brain activity and its relation to behavior in freely moving animals is a major goal in neuroscience. Traditional microscopes offer high resolution or wide fields of view (FOV), but not both simultaneously in a compact form factor. Large, heavy lenses in conventional microscopes restrict animal movement, hindering studies of natural behavior. While lensless microscopy promises a thin, lightweight solution with high resolution and large FOV, it has faced challenges in in vivo applications, particularly with the low contrast of biological tissue and the integration of illumination and filtering into a compact device.
Methods
The core innovation of FlatScope 2.0 lies in its computational imaging approach. Instead of traditional lenses, it uses a phase mask to shape the light. This mask, designed with a "contour-based PSF" generated using Perlin noise and edge detection, is optimized to capture textural frequencies common in biological samples, improving reconstruction of dense, low-contrast tissues.
Key methodological advancements include:
- Optimized Phase Mask: A custom-designed phase mask generates a spatially sparse and high-contrast point spread function (PSF) that is robust to noise and preserves spatial frequency information.
- Computational Reconstruction: Advanced algorithms are employed to reconstruct images from the sensor measurements, effectively inverting the transfer function between the light field and the sensor. This includes Tikhonov regularization to mitigate noise amplification inherent in lensless imaging.
- Integrated Illumination and Filtering: A compact system for epifluorescence imaging is achieved using a fiber-optic cable with a microprism for illumination and a hybrid absorptive/interference filter integrated with the phase mask to block excitation light.
- 3D Imaging Capability: The PSF's depth-dependent nature allows for single-capture, three-dimensional imaging by using a series of depth-calibrated PSFs during post-processing.
Datasets
The paper evaluates FlatScope 2.0 using:
- USAF Resolution Targets: Used to quantify the spatial resolution of the reconstructed images.
- Fixed Biological Samples:
- Convallaria majalis (lily of the valley) stained with green fluorescent protein (GFP).
- Mouse brain slices expressing GCaMP6f.
- Live Animal Imaging:
- Hydra vulgaris expressing GFP and GCaMP7b.
- Mice expressing GCaMP6f in the motor cortex.
The paper does not introduce new benchmark datasets but rather demonstrates the system's performance on established biological samples and in vivo models.
Key Insights and Ideas
- Lensless Microscopy for In Vivo Biology is Achievable: The primary insight is that the challenges of in vivo lensless imaging, particularly low contrast and integration, can be overcome with careful design of the phase mask and reconstruction algorithms.
- Computational PSF Design is Crucial: The "contour-based PSF" is a significant innovation, enabling better reconstruction of complex biological scenes compared to simpler PSFs used in previous lensless systems.
- Thin, Lightweight Imaging for Unrestricted Behavior: FlatScope 2.0 offers a path towards studying neural activity in animals performing natural behaviors without the constraints of bulky microscopes.
- Single-Capture 3D Imaging: The inherent depth-dependent PSF of the lensless system unlocks efficient volumetric imaging without mechanical scanning, opening doors for high-speed 3D studies.
- Potential for Cellular-Resolution Calcium Imaging: The system demonstrates the ability to capture calcium dynamics at resolutions approaching that of individual cells, a critical capability for understanding neural circuits.
Key Results
- Spatial Resolution: Achieved better than 9 µm resolution on USAF targets, approaching the resolution needed for individual neurons.
- Field of View (FOV): Demonstrated a FOV of ~16 mm² in mice, which is approximately 30x larger than other cellular-resolution miniature microscopes.
- 3D Imaging Volume: Reconstructed a 4 mm³ volume of Hydra vulgaris from a single capture.
- In Vivo Calcium Imaging: Successfully captured in vivo calcium dynamics in mice motor cortex and Hydra vulgaris, showing high correlation (median correlation coefficient of 0.852) with conventional epifluorescence microscopy.
- Cell-Sized Signal Extraction: Demonstrated the ability to extract Ca²⁺ signals from regions as small as 150 µm² to 0.01 mm² using RPCA and k-means.
Limitations and Future Work
- Sub-cellular Resolution: While achieving cellular resolution, the authors acknowledge that further improvements are needed to reach sub-cellular resolution. This could be achieved by optimizing masks for closer scene-to-mask distances.
- FOV vs. Resolution Trade-off: Bringing the device closer to the sample to improve resolution leads to a decrease in FOV. Future work suggests arraying devices to maintain large scene coverage while keeping a thin profile.
- Light Collection Efficiency: Improvements in light collection efficiency are desired, which could be addressed by closer mask-to-sample distances and potentially using more sensitive sensors like sCMOS or SPADs.
- Integration of Light Source: Integrating the excitation light source directly into the device (e.g., using micro-LEDs or waveguides) is suggested to further improve form factor and illumination coverage.
- Imaging Through Scattering Media: While the system shows robustness to scattering, further advancements in computational methods could extract even more 3D information.
- Broader Imaging Modalities: The principles behind FlatScope 2.0 are not limited to fluorescence and could be extended to other modalities like brightfield, darkfield, and reflected-light microscopy for applications like endoscopy and point-of-care diagnostics.
- Imaging Tens of Thousands of Neurons: A significant future opportunity is the potential to image tens of thousands of neurons simultaneously in freely moving animals.
Summary updated