HyperSpectral Skin Imaging

1
HyperSpectral Skin Imaging Tianchen Shi, Prof.
Charles A. DiMarzio Department of Electrical and
Computer Engineering, Northeastern University,
Boston, MA 02115 This work was supported in
part by CenSSIS, the Center for Subsurface
Sensing and Imaging Systems, under the
Engineering Research Centers Program of the
National Science Foundation ( Award Number
EEC-9986821),

• CenSISS Value Added
• Current Status
• Achieved fast imaging ability to obtain
  reasonable skin chromospheres concentrations, and
  oxygen saturation.
• Obtained blood vessels general geometry
  information average diameter and depth of blood
  vessels.
• Future Plans
• Extend to a spatially resolved concentration
  mapping.
• Explore blood vessels diameter and depth
  distributions
• Obtain better light source and detection arrays
  for high accuracy measurements,and increase the
  accuracy and stability of the information
  extraction procedures.
• Take more in-vivo data for different skin
  subjects.

• Abstract
• Chromosphere monitoring is an important aspect
  in most of skin studies, especially for
  hemoglobin in human skin, which is closely
  related to development and diagnosis of various
  skin diseases. However, recent studies reveals
  that hemoglobin concentration obtained by
  assuming homogeneous distribution may deviate
  from its true value. On the other hand, diameter
  and distribution of blood vessels in skin are
  also crucial in monitoring of blood pressure and
  body temperature. Among various optical skin
  imaging techniques, HyperSpectral Imaging is a
  fast, stable method to provide spatial and
  spectral information of skin tissue. In this
  poster, we present a Monte Carlo assisted
  HyperSpectral imaging method to meet both of the
  challenges mentioned above. Reflection spectrum
  measurements are used to determine absolute
  concentration of chromospheres with aid from a
  Monte Carlo model, and blood vessel information
  can be further derived by a correction model for
  the inhomogeneity in skin layers.
• State of the Art
• Izumi Nishidate, Muroran Institute of
  Technology, Japan 4. The group used a multiple
  regression analysis assisted by a homogeneous
  Monte Carlo model to extract chromospheres
  concentration for point based total reflection
  measurement.
• W. Verkruysse, Department of Radiation Oncology,
  University hospital Rotterdam, Rotterdam, The
  Netherlands 3. The group was using a correction
  factor obtained by a single blood vessel
  assumption for a fiber based skin measurement.
• Significance and Challenges
• Linear multiple regression based HyperSpectral
  imaging may provide possibility of fast imaging
  over an area of skin.
• Pre-computed inhomogeneous Monte Carlo model may
  allow to obtain general information of blood
  vessels at superficial skin layer.
• The method may need very high accuracy and
  stability of the Monte Carlo model to solve the
  non-linear and inter dependent relationship
  between chromosphere concentration and obtained
  multiple regression coefficients.
• Technical Approaches
• Orthogonal polarization detection with
  incoherent HyperSpectral imaging for in-vivo
  human skin measurement.
• Monte Carlo simulation for inhomogeneous layered
  skin model with very carefully selected
  parameters.
• Fast 2-D imaging with linear multiple regression
  analysis and non-linear conversion by
  pre-computed Monte Carlo results.

Figure 2, Experiment Setup Layout
Figure 1, Two-layer skin model randomly
distributed blood vessels with diffusively
incident photon packets.

• Skin Model
• Modified Lambert-Beers Law
• where A is absorption, S is scattering term, ?
  is molar concentration of prevailing
  chromospheres (melanin, hemoglobin), and L is
  mean free transport path length.
• Monte Carlo assisted Multiple Regression
  Analysis
• where am is linear multiple regression
  coefficients, a is a vector consisting of the
  combination of am for the first and the third
  order, and b is a conversion vector nonlinearly
  relating Cm and am , obtained with an
  inhomogeneous Monte Carlo skin model.
• Inhomogeneity in skin model (blood vessels)

Figure 3, In-vivo human skin absorption data cube
Figure 4, Extracted oxygen saturation map,
average chromospheres volume concentration
melanin (4.30), total hemoglobin(0.32), average
blood vessels diameter (101um) and depth (164um)
Figure 5, Comparison of in-vivo spectrum and
Monte Carlo simulation with extracted parameters