PNNL-13501 - Pacific Northwest National Laboratory

In the optical microscopy image, the right side of the bead
and the top right side of the oocyte are poorly defined.
This is due to the presence of three small oocytes in the
optical path and below the optical microscopy plane,
which produce shadows in the plane of interest. The high
optical microscopy intensity at the oocyte boundary is
believed to arise from the mitochondria in the surrounding
follicle cell layer, while the interior is optically opaque.
Additional features include stained connective tissue near
the oocyte and a fluorescing layer along the inner wall of
the sample tube. We conclude that excellent image
registration of both images is obtained. Figure 2 shows
similar confocal and magnetic resonance images obtained
on a smaller stage-2 (0.38-mm-diameter) transparent
oocyte. In contrast to results obtained on the larger
oocyte, the confocal image clearly shows the
mitochondrial cloud, whereas in the water magnetic
resonance image, only an inside layer of enhanced water
intensity is detected. The results clearly illustrate that
combined microscopy provides significantly more
information than obtained with each of the techniques
individually. The magnetic resonance images provide
detailed information about the intra-cellular structure of
the larger opaque oocyte that is not observed with optical
microscopy. On the other hand, in the smaller transparent
oocytes, the high-resolution optical images can be used to
complement the relatively low-resolution magnetic
resonance images.
Figure 2. Application of confocal image data to enhance the
resolution and contrast of magnetic resonance images at
object boundaries. (a) magnetic resonance water image of
the stage-2 oocyte; (b) optical microscopy image of the same
oocyte, and the contour plot, obtained from this image; (c)
magnetic resonance image with enhanced resolution and
contrast near the contour boundaries. Enhancement was
achieved by overlaying the image (a) and the contour plot (b)
to distinguish magnetic resonance image pixels containing
the boundaries between the cell and the surrounding
medium. Next, the average magnetic resonance intensities in
these pixels were redistributed using the confocal
boundaries. The scale bar shown in (b) is 0.2 mm in length.
First Optical Microscopy/Magnetic Resonance
Microscopy Application: Sharpening of the Magnetic
Resonance Images
The a priori knowledge provided by the confocal image
can be used to improve the boundary resolution and the
contrast in the magnetic resonance images. This is
illustrated in Figure 2 as well. By overlaying the highresolution
optical contour plot, given in Figure 2b, with
the relatively low-resolution magnetic resonance image
shown in Figure 2a, image pixels can be identified in the
magnetic resonance image containing the boundary
between the oocyte and the surrounding medium. Then
the average intensity in each of these pixels can be
redistributed into each compartment inside these pixels.
Figure 2c shows the resulting magnetic resonance image.
It follows that both the boundary resolution and contrast
are significantly enhanced. Hence integrated optical
microscopy/magnetic resonance microscopy can be used
to produce images in which the optical spatial resolution
is combined with the magnetic resonance contrast, which
could be important for a variety of applications, including
the diagnosis of diseased cells.
Statistical Analysis
The research and development of algorithms to analyze
the combined measurement sets from the magnetic
resonance and optical microscopes focused on three
problems: 1) image and spectrum enhancement, 2) image
registration (alignment of data spaces), and 3) spectral
unmixing/spectral sharpening. The first problem involves
the elimination or minimization of measuring artifacts.
The second problem is due to differences in physical
alignment, resolution, and measuring mode of the two
distinct instruments. The third problem is about
capitalizing on the complementary, but distinct,
information provided by two distinct instruments
simultaneously measuring the same sample. Solutions to
these problems using existing software such as Image-J
by NIH and IBM’s OpenDX melded with custom
components are being explored. At this time, we have
identified the necessary operations (deconvolution, image
warping, spectral unmixing); we have derived preliminary
algorithms, and have developed prototype subroutines.
We have not, as yet, developed a refined and integrated
analysis package.
Biosciences and Biotechnology 55