Research - Polarimetry

Polarimetric imaging of skin cancer borders

True borders of certain skin cancers are hard to detect by the human eye. For this reason we are developing techniques such as polarized light imaging that enhance skin cancer contrast facilitating the doctor assessment before Mohs surgery procedures. We are also studing the skin microroughness, in order to improve tissue imaging techniques and as a way to measure skin texture.

We have built a BRDF layout that has no moving parts and can reconstruct the full backscattered Stokes vector using liquid crystal retarders. The new Hemispherical layout is shown the figure below. The light source elements are tricolor LED, this allows for an incoherent illumination where the wavelength can be modified programmatically. The LED wavelengths are centered at 472, 525 and, 672 nm respectively. A dichroic sheet polarizer (POL) in front of each LED polarizes the incident beam at 45 degrees to the scattering plane. In our previous work we used a 90 degrees incident polarization.

Hemispherical Layout

Illumination tube

For every azimuth angle &Phi_s and incident scattering angle &theta_i we measure the Stokes vector of the back-reflected light. Two incident scattering angle are available, at &theta_i = 24 degrees six different incident azimuth are available and at &theta_s = 49 degrees we have ten different illuminators. The detector is a 12 bit CCD camera located at &theta_s = 0 and &Phi_s = 180. Two liquid crystals variable retarders LC1 and LC2 (retardance &delta₁ and &delta₂ respectively) oriented at &theta₁ and &theta₂ with respect to the polarizer POL modulate the backreflected polarized light before reaching the camera.

Results

Some polarimetric images obtained with this device are shown in the figures below

Figure 3. S0 for images obtained using red illumination for &thetai=24 degrees and &Phis=0 degrees.

Figure 4. Ipol for images obtained using red illumination for &thetai=24 degrees and &Phis=0 degrees.

Figure 5 and 6 show images obtained with the same red illumination but for &theta_i=49 and &Phi_i= 114 degrees. At these angles the data seemed to deviate the most from the rough surface model tending towards the two source model. Indeed figure 4 and 6 are very different. In figure 4 skin pores are very visible while in figure 6 these structures have completely disappeared while other structures are now present in the image.

Figure 5. S0 for images obtained using red illumination for &thetai=49 degrees and &Phis=114 degrees.

Figure 6. Ipol for images obtained using red illumination for &thetai=49 degrees and &Phis=114 degrees.