The capability of acquiring hyperspectral information in low light conditions is potentially important for a variety of applications, ranging from remote sensing to biomedical fluorescence imaging. Particularly interesting is its use in optical analysis of biological samples in which the light level should be kept low to prevent tissue damage. For this purpose a low-light hyperspectral camera has been developed for the 0.4 to 0.9 µm spectral range. The camera is based on an electron-multiplying CCD (EMCCD) detector which effectively suppresses readout noise, and approaches the fundamental photon noise limit. It has been designed with close-up optics to image an area on the order of centimeters. Fluorescence images of skin samples illustrate the camera performance. Results show that low-light hyperspectral imaging has a potential for biomedical applications.Index Terms-biomedical imaging, fluorescence imaging, hyperspectral imaging, low-light imaging.
I TRODUCTIOHyperspectral imaging, also referred to as imaging spectroscopy, collects spatially resolved spectral information from the imaged scene. It has traditionally been applied to remote sensing and military surveillance. However, many new applications have emerged in areas like medicine, chemistry, geology, food science and industrial processes [1][2][3]. In order to acquire useful data, the noise floor of the device must be lower than the signal level generated by the photon flux. This requirement can be met with relative ease in daytime remote sensing, and also in laboratory or industrial applications where bright artificial illumination sources can be used to get an adequate signal-to-noise ratio (SNR).However, there are other potential applications with inherently low light environments, such as dusk-to-dawn remote sensing or laboratory imaging of light-degradable samples. Hyperspectral fluorescence imaging of living tissue or other biological samples is an example of the latter. Typically, fluorescent signals are several orders of magnitude weaker than the excitation sources used, and the excitation is limited by the sample damage threshold. It has been demonstrated that with a sensitive camera, it is possible to obtain spectral information from poorly illuminated scenes with photon noise limited SNR [4]. Such low-light performance can be achieved using an electron-multiplying CCD array (EMCCD), where the internal gain can be used to lift weak signals above the readout noise.Here we report on a new EMCCD-based hyperspectral camera design for low-light imaging. The camera has been tailored for a range of biomedical applications with varying light levels and object distances. With a scene pixel size of 25 µm, the resolution is lower than in microscopy, but the setup is capable of imaging objects several centimeters in size.In biomedical applications, this permits imaging of relatively large tissue samples, thus providing useful information complementary to that obtained from microscopy-based methods [5]. The overall performance of the camera was tested by fl...