Single Cell Tomography for Early Cancer Detection

1
Single Cell Tomography for Early Cancer
Detection Vivek Nandakumar, Laimonas Kelbauskas,
Roger Johnson, Deirdre Meldrum Arizona State
University, Tempe, Arizona
ABSTRACT Nuclear morphology is a proven biomarker
for early detection of deadly diseases such as
cancer. High resolution 3D cell imaging may
facilitate sensitive and specific early
detection. We perform optical tomographic
imaging on individual, hematoxylin-stained cells
to quantify variations in nuclear morphology in
several cell lines spanning the neoplastic
progression spectrum in esophageal cancer. Our
3D cell images are obtained by applying
principles of optical projection tomography to
obtain isotropic resolution of 350 nm. Each cell
image is generated with mathematical
reconstruction algorithms from 500 projection
images acquired over 360. Using this technique
we observe qualitative and quantitative
differences in nuclear morphology between
esophageal cell lines representing normal and
dysplastic cells. Our results validate the
superiority of 3D over 2D quantitative cytometry.

• SINGLE CELL TOMOGRAPHY USING Cell CT
• Isotropic resolution of 350 nm.
• Cell imaged in suspension and not on slide.
• 3D image reconstructed from projection images.
• 500 projections acquired over 360.
• Optical Projection Tomography2 used to acquire
  projection image.
• Imaging modality transmission (brightfield)
  mode, using 100x oil objective.
• Cells to be imaged are embedded in a thixotropic
  carrier gel that has
• same refractive index as glass.

RESULTS (Quantitative)
Morphological differences observed between
studied cell lines !

• MATERIALS AND METHODS
• Cells Human esophageal epithelial cell lines3 .
• EPC (normal) CP-C (early dysplastic) CP-D
  (late dysplastic)
• Stain Hematoxylin, an absorption dye that
  heavily stains the nucleus.
• Sample preparation
• Fix cells using CytoLyt.
• Stain using hematoxylin.
• Embed stained cells in gel.
• Image acquisition
• Image cells using cell CT.
• Automated 3D cytometry
• Compute nuclear morphometric features and
  analyze data.

• INTRODUCTION
• Changes in nuclear morphology are strong
  indicators of onset of malignancy.
• Computer Aided Detection (CADe) has emerged as a
  useful tool for early
• cancer diagnosis.
• CADe requires high resolution imagery and robust
  computer algorithms.

Histograms that illustrate quantitative
differences between some of the computed
morphological features of the studied cell lines.
One hundred 3D cell images of each cell type were
used for the analysis. A total of forty features
were computed using fully automated 3D image
processing techniques. Significant variation is
observed in the computed morphological features.
1
2
3
Cells before and after staining with Hematoxylin.
Brightfield image acquired at 50x magnification.
4

• CONCLUSIONS
• Utility of single cell tomographic imaging
  demonstrated for early cancer detection..
• Qualitative and quantitative differences
  observed in nuclear morphology between cell
  lines.
• Morphological factors such as cell size, nuclear
  size, nucleus-cytoplasm ratio, nucleolar
• margination toward the nuclear membrane and
  total DNA content were elevated in abnormal
• cells.
• The DNA is observed to be more clumpy in
  abnormal cells.
• Variations were observed in texture of nuclear
  surface and chromatin.

Normal nucleus
Cancerous nucleus
Zink et al. 1
RESULTS (Qualitative)
Morphological differences observed between
studied cell lines !
DRAWBACKS OF CURRENT APPROACHES

• DISCUSSION
• Current CT capability limited to structural
  imaging of fixed cells.
• Use of Eosin to stain the cell cytoplasm is
  under evaluation.
• Incorporation of fluorescence CT modality
  planned to facilitate functional imaging .
• Significance of cell cycle in morphological
  changes to be assessed.
• More robust algorithms planned to improve
  accuracy of 3D nuclear morphometry.
• A quantitative approach to determine the most
  discriminatory features is currently under study.

3D imagery (confocal microscopy)
2D imagery

• Axial (along z-axis) resolution is poorer than
  lateral resolution (along x-,y-) by a factor of
  at least 2.
• Loss of morphological detail due to imaging on
  glass slide.

• Orientation Dependence

• REFERENCES
• D. Zink, A.H. Fischer, and J.A. Nickerson, Nature
  Reviews Cancer 4, 677 (2004).
• M. Fauver, E.J. Seibel, J.R. Rahn et al., Optics
  Express 13 (11), 4210 (2005).
• M. C. Palanca-Wessels, M. T. Barrett, P. C.
  Galipeau et al., Gastroenterology 114 (2), 295
  (1998).

• Focal Plane Dependence

ACKNOWLEDGEMENTS This research is supported by
NCI Center for the Convergence of Physical
Science and Cancer Biology, grant number
U54CA143862.
Rendered confocal micrograph of a mammary
epithelial cell stained with syto-9 dye. Image
acquired at 100x magnification.
Volume renderings of EPC (top), CP-C (middle) and
CP-D (bottom) esophageal epithelial cells. Left
images show the nuclear surface, middle images
illustrate the nuclear interior, and right images
depict a slab through the volume.