Through the use of a USAF 1951 reflective target, we were able to detect a resolution of approximately 150μm, corresponding to half the SD separation as expected. The device-generated image is shown in
along with a corresponding CCD camera image.
Fig. 3 Experiment measurement of the optical performance of the FSI system. Left: Image of the elements of group 7 of a USAF 1951 standard resolution target. Field of view: 1.2cm by 1.2cm Right: Image of the USAF target using Leica EZ4DMicroscope image of a (more ...)
We characterized the MEMS hyperspectral imaging system using both Polydimethylsiloxane (PDMS)-Quantum Dots (QDs) phantoms and biological samples such as porcine epithelium. A multilayer QD PDMS phantom sample was fabricated for evaluating the quality of depth sensitive and hyperspectral imaging simultaneously (
). Micro-contact printing (μCP) is used to pattern QDs on the surfaces of multi-stack PDMS thin layers. Each spin-coated PDMS thin-layer is 200μm thick, infused with titanium dioxide to simulate the scattering in tissue. The sample contained a 3x3 QD pattern array, each depth containing 3 different colors on it.
Schematic of the μCP fabricated QDs multilayer PDMS sample. Left: Isometric View. Right: Side View.
By actuating the MEMS micromirror scanning in a lissajous pattern, the lateral resolution and the field of view (FOV) of the real-time imaging system is experimentally determined as ~100μm and 1.2cm by 1.2cm, while rendering at a frame rate of 0.8 frames per second (fps). Using our real-time HSI system on the PDMS-QD phantom, we took 30 images of different wavelengths through the spectrum within 70 seconds (
). Therefore an acquisition speed of 0.43 wavelengths per second was obtained. The 100μm slit placed in front of the PMT guarantees a spectral resolution of 6nm.
MEMS HSI for PDMS-QD phantom imaging. (a) PDMS sample imaged under 12 selected wavelengths from the 30 total acquired wavelengths. (b) Normalized spectrum of 3 featured positions on the PDMS sample.
Using the images of 3 peak wavelengths for each color, the pseudocolor image was merged and rendered (
), which shows good preserving of features when comparing to the mosaic image from commercial microscope (). The resolution of the image is visually measured as roughly 100μm, suggesting that for a confocal arrangement (zero source-detector separation), the resolution is related to pinhole size.
Comparison of images acquired using MEMS HSI and Olympus microscope. (a) Pseudocolor image merged from 3 peak wavelength images. (b) Microscope image using Olympus BX51 microscope. Scale bars are 1mm.
For acquiring depth sectioned images; a 100μm pinhole is translated in the image plane as a collection aperture for light returning from the sample ().
Source-detector (SD) separation is radial distance on the sample surface between the source spot and the spot imaged by the collection aperture. This radial separation will be proportional to the mean sampling depth of the collected photons due to the diffuse nature light scattering in turbid samples. Offsetting the collection from the source spot on the sample creates a Source-Detector ‘σ’ separation. Equation (1)
describes the relationship of the fluorescence intensity comparison collected between the deeper and shallower layers:
Here I refer to the collected fluorescent light intensity from the quantum dots layers, and σ stands for different source-detector seperation values.
our different SD separations were performed, showing the sampling depth selectivity in terms of the ratios of fluorescence intensities between the shallowest and deepest layers. Each image was normalized to full dynamic range so that the trend could be seen qualitatively in the figure.
Fig. 7 (a)-(d): Depth sampled images for four different SD separations. Fluorescence intensities within selected regions of the image were calculated to obtain ratios of shallow to deep intensities. Red and yellow outlined regions contain shallow (200μm) (more ...)
implies the diminishing trend of the intensity ratios with increasing SD separation. This implies that fluorescence emanating from more deeply implanted quantum dots account for an increasing percentage of the overall collected light as SD separation increases, as indicated by previous studies [7
]. Here we have demonstrated that by increasing the SD separation from 0 to 400μm, the selectivity of the deepest layer increases by a factor of 1.83 times. While boosting the relative signal of deeper features, close inspection of will reveal that SD separation has an adverse effect on spatial resolution and signal to noise ratio.
Biological imaging potential was demonstrated using quantum dots with 2 different emission peaks placed under the surface of ex vivo porcine epithelium (
). The spectra of two QD implanted sites were measured. Image processing techniques including noise despeckling and contrast enhancing were used here to improve the image quality. Pseudo color images were acquired from hyperspectral images using the method of spectral unmixing. Using this method, we demonstrate our system’s capability of optically distinguishing between fluorophores. This shows the potential of using bio-conjugating agents for hyperspectral fluorescence imaging towards biomarker detection.
Fig. 8 Biological Sample of porcine epithelium with QDs placed underneath the surface, with SD separation being zero (essentially, the confocal configuration). (a) Camera image of sample, MEMS HSI scan area is delineated by the white box. (b) Emission spectrum (more ...)