Remote non-invasive stereoscopic imaging of blood vessels: first in-vivo results of a new multispectral contrast enhancement technology

Abstract

We describe a contactless optical technique selectively enhancing superficial blood vessels below variously pigmented intact human skin by combining images in different spectral bands. Two CMOS-cameras, with apochromatic lenses and dual-band LED-arrays, simultaneously streamed Left (L) and Right (R) image data to a dual-processor PC. Both cameras captured color images within the visible range (VIS, 400-780 nm) and grey-scale images within the near infrared range (NIR, 910-920 nm) by sequentially switching between LED-array emission bands. Image-size-settings of 1280 x 1024 for VIS & 640 x 512 for NIR produced 12 cycles/s (1 cycle = 1 VIS L&R-pair + 1 NIR L&R-pair). Decreasing image-size-settings (640 x 512 for VIS and 320 x 256 for NIR) increased camera-speed to 25 cycles/s. Contrasts from below the tissue surface were algorithmically distinguished from surface shadows, reflections, etc. Thus blood vessels were selectively enhanced and back-projected into the stereoscopic VIS-color-image using either a 3D-display or conventional shutter glasses. As a first usability reconnaissance we applied this custom-built mobile stereoscopic camera for several clinical settings:* blood withdrawal;* vein inspection in dark skin;* vein detection through iodide;* varicose vein and nevi pigmentosum inspection. Our technique improves blood vessel visualization compared to the naked eye, and supports depth perception.

Figure 1.

Experimental setup. Two CMOS-cameras, with apochromatic lenses and dual-band LED-arrays, simultaneously stream Left (L) and Right (R) image data to a dual processor PC. Both cameras captured color images within the visible range (VIS, 400–780 nm) and grey-scale images within the near infrared range (NIR, 910–920 nm) by sequentially switching between LED-array emission bands.

Figure 2.

Schematic diagram of image aquisition. The sequentially acquired alternating VIS and NIR raw image frames form 3D-matrices for the Left and Right channel. The NIR image size is smaller than the VIS image size to increase framerate while maintaining an overview of the imaged area. Enlarged details illustrate the NIR transparency of the Bayer pattern RGB-filters which are applied to obtain a VIS color image. The time domain axis is expressed in image cycles. Along this image cycle axis, the control signals for NIR and VIS LEDs (synchronized with respectively NIR and VIS camera exposures) are visualized.

Figure 3.

Schematic diagram of image processing. The images captured within the visible range (VIS) and the images captured within the near infrared range (NIR) are combined, which reveals blood vessel patterns below the skin. The left and middle column focus on edge-enhancement and suppression of superficial artifacts, the right column serves to fill-in the blood vessel lumen. For raw VIS & NIR images as well as processed results see figures 6, 7, 8 and 9.

Figure 4.

Relative intensity distribution of the different spectral bands. Dimensionless normalized distribution of the intensities for the 4 aquired red (I_R), Green (I_G), Blue (I_B) and near infrared (I_NIR) individual spectral bands expressed in ratio to the normalized intensity (I_VIS) of the composed RGB-image. The four data clouds (R, G and B labeled by their natural colors and NIR labeled as pink) show a generally marked separation, but especially for a number of pixels within the red and infrared the data clouds partly overlap. This indicates regions where the VIS contrast potentially is superior to the NIR contrast. By comparing an adjustable threshold with the calculated ratio of (I_R/I_VIS)/(I_NIR/I_VIS) it can be decided which spectral band provides superior contrast for the pertaining pixel and thus whether or not it is used for enhanced backprojection.

Figure 5.

Principle of applied autostereoscopic LCD-monitor (reprinted with permission from Sharp). In 2D mode, only one camera-channel is displayed (either from the L or R camera) and the parallax barrier is not actuated. Both eyes of an observer therefore receive the same image and a conventional flat image with full resolution is seen. In 3D mode, both camera channels are displayed (L&R) and the parallax barrier is actuated. The left eye and right eye of an observer now receive different images, and a stereoscopic (in-depth image) with halve resolution is seen.

Figure 6.

Routine blood withdrawal. Image pairs showing unprocessed images for VIS (a) and NIR (b) as well as the result after application of the new image processing method (c). Note the forked shadow (which is not effected by the enhancement algorithm), the clearly visualized subcutaneous bleeding and the improved visibility of the needle tip.

Figure 7.

Influence of skin pigmentation. A dark skin color (a) has no significance for the applied NIR wavelength of 920 nm. Blood vessels provide good contrasts (b) and the resulting enhanced image (c) offers an improved visualization of the vasculature.

Figure 8.

Vein detection through iodide. Within the visible range, the superficial vasculature is only vaguely discernable (a). After filling the Petri-dish with a 3 mm thick layer of iodide solution, this fully blocks out the tissue view within the visual range (b), whereas a clear view of the vasculature remains possible at the applied NIR wavelength of 920 nm (fig 8c).

Figure 9.

The VIS image does not contain much information about the underlying vascular pattern (a). The NIR image, however, clearly shows what’s hiding beneath the surface (b). Note that, when building the enhanced image, the nevus which is present in the VIS image can freely either be suppressed as a surface contrast (c) or be kept visible (d).