Page 190 - Vibrational Spectroscopic Imaging for Biomedical Applications

166 Cha pte r S i x

longer experiments. Because of this trade-off between fluorescence and
SNR, the laser excitation wavelength chosen should be based on the
specific application and goal.

6.4.2 Correction for Dark Current
Cameras containing a CCD sensor are often used as detectors in
chemical imaging. Because Raman scattering is a weak phenomenon,
a highly sensitive camera containing an electron-multiplying CCD
(EMCCD) has become a preference for widefield Raman imaging. For
all types of CCDs, the dark current attributed to the sensor must be
corrected and thus eliminated from the background.
In the absence of radiation, a constant charge accumulates within
the camera wells and on the chip that is the result of the random gen-
60
eration of electrons. This constant response, or dark current, adds to
the signal produced. To remove this response, a background image is
collected without any source of radiation. This image is then subtracted
from each image frame during acquisition. Alternatively, it may also be
subtracted from the image stack after data acquisition is complete.
6.4.3 Cosmic Filtering
One important step that is applied to widefield Raman images before
other preprocessing steps is a cosmic filter. Random cosmic events
often occur while collecting Raman data. These events are seen as
bright pixels while scrolling through the wavelength dimension of a
widefield Raman image. Applying a cosmic filter eliminates these
cosmic spikes. The cosmic filter is a median filter that considers each
pixel in the image and checks the nearby neighboring pixels to deter-
mine if it is representative of its surroundings. A standard deviation
and a window size are set to determine the range and acceptable pixel
intensity. An alternative method is to perform a threshold filter to the
data. Using this method, upper and lower threshold values are set.
Any pixels outside of this threshold value are replaced by the average
intensity of the moving window of a selected size of neighboring pixels.
This intensity is filtered at the center.

6.4.4 Instrument Response Correction
A major source of background interference in widefield Raman
images is signal contributed by the instrument, known as instrument
response. Every Raman imaging instrument contains complex optical
components. The signal attributed to these components is present in
the background and should be removed to reveal the subtle differences
in the Raman signal of biological samples. A sample with a known
spectral profile, such as calibrated light sources, blackbody radiators,
and white light, may be used to correct for instrument response.
Recently, the National Institute of Standards and Technology (NIST)
has tested and developed standard reference materials (SRM) with