Upconversion based mid-infrared hyperspectral imaging

Hyperspectral imaging in the mid-infrared spectral range is an emerging technology utilized for a multitude of applications, but its full potential is held back by the lack of sensitive mid-IR detectors. Nonlinear frequency upconversion offers a promising alternative to direct detection for room-temperature mid-IR spectroscopy and hyperspectral imaging.
In this work, upconversion based hyperspectral imaging has been demonstrated in the 1 to 10 µm spectral domain using different kinds of illumination sources narrowband/broadband, continuous wave/pulsed, such as globar, Quantum Cascade Laser, Optical Parametric Oscillator and Supercontinuum source. Nonlinear media such as Lithium Niobate (birefringent and quasi phase matched) and Silver Gallium Sulfide have been exploited for sum frequency generation i.e. upconversion.
Large field of view and broad spectral coverage have been achieved using phase-matching scanning techniques such as temperature and angle tuning of a nonlinear crystal. Postprocessing techniques have been developed for the construction of upconversion hyperspectral cubes based on both narrowband and broadband light sources.
A video-frame rate upconversion imaging system is realized using a standard CCD camera, in synchronism with the crystal rotation of an upconversion system. This system is capable of acquiring 64 kpixel upconverted mid-IR images in 2.5 ms, without the need for post-processing. This approach is generic in nature and constitutes a major simplification in realizing video-frame rate hyperspectral imaging in the mid-IR.
Based on this setup, a pilot study on oesophageal tissues samples from a tissue microarray, is presented, in the 3 to 4 µm wavelength range using computer assisted classification. Comparing the stained sections evaluated by a pathologist to those obtained by either FTIR or upconversion hyperspectral imaging based on machine learning shows great promise for future research pointing towards potential clinical translation.
In the last phase of the project, point-spread function engineering in the upconversion process is presented. Dark field imaging of static and dynamic near-infrared phase objects is demonstrated – resulting in edge-enhanced upconverted images in the visible spectral range. Through this scheme, the spiral phase filter operates at the pump wavelength rather than in the mid-IR signal wavelength which can potentially be broadband. This approach offers some advantages due to the lack of high-performance spiral phase filters and cameras operating at the signal wavelength (especially for mid-IR objects). Furthermore, upconversion dark field imaging is extended numerically and experimentally from collinear to non-collinear interactions between signal and pump.