In vivo results using photothermal tomography for imaging cutaneous blood vessels

Previous studies suggest that optimal port wine stain (PWS) laser treatment parameters require knowledge of skin characteristics such as blood vessel size, depth, and distribution. Effective and rapid imaging modalities are not widely available. In the present study, photothermal tomography (PTT) images of an in vivo hamster window model and human PWS skin were obtained and analyzed. Subtherapeutic laser light pulses at 585 and 600 nm were applied to skin surface and image sequences acquired with an infrared camera. A nonnegatively constrained conjugate gradient algorithm was used to reconstruct a PTT image of the initial temperature distribution immediately following pulsed laser irradiation. Vessel dimensions determined from PTT images of hamster window model skin compared well with those measured directly using video microscopy. PTT images of human PWS skin contained vessels with estimated diameters of 200-250 μm over a 250-320 μm depth range. Use of dual wavelength excitation (DWE) analysis allowed for imaging of shallow vessels.

trauma. Current therapy involves use of pulsed laser light combined with cryogen spray cooling.
The primary problem is that few (< 10%) treated patients achieve complete removal of their PWS after multiple treatments {1}. Results of previous studies [2-7J suggest that selection of optimal treatment parameters depends on knowledge of PWS skin characteristics such as depth of the epidermal basal layer, maximum epidermal temperature rise, depth of the most superficial blood vessels, and vessel size and distribution.
Photothermal tomography (PTT) [8] is a noninvasive technique designed to provide quantitative information on the aforementioned PWS skin characteristics. Briefly, a subtherapeutic laser pulse is applied to a specified PWS area, and the subsequent increase in infrared emission is detected with a camera. A sequence of infrared images is processed with an inversion algorithm to reconstruct an image of the 3-D temperature distribution immediately following pulsed laser irradiation.
The purpose of the present study is to analyze PTT images computed from data obtained from a representative in vivo animal model and from a PWS subject currently undergoing treatment at the Beckman Laser Institute and Medical Clinic.

MATERIALS AND METHODS
2.1.

PTT Theory
Telenkov et al. [8] previously provided a detailed description of PTT theory. Briefly, an infrared image sequence [R(x,y,t>O)] is acquired after pulsed laser irradiation. AR is related to the actual skin temperature distribution immediately following irradiation {zT(x,y,z,t=O)] by the following equation: AR(x',y',t>O)= JJJK(x'-x,y'-y,z,t)LtT(x,y,z,t=O)dxdydz (1) x,y,z where K is a kernel matrix describing heat diffusion dynamics and attenuation of blackbody radiation from the skin: K(x'-x,y'-y,z,t) = Kr(X'X,y'y,t) . C B2(T0) PIRJKz(z',t) e_/1RZ' dz' (2) where Kr and K are thermal pointspread functions in lateral and axial dimensions [9] ; C is a constant accounting for numerical aperture and losses of the collection optics, and detector specifics (e.g., size, sensitivity, integration time, and detection band); B'(T0) is the first derivative of Planck's law of radiation, evaluated at initial temperature T0; and Ji1 is the attenuation coefficient of skin to blackbody radiation. Since a band limited infrared camera detection system is used, we are concerned only with Jiia over the detection band. Majaron et al. [10] determined that use of a 4.5-5 im band justifies selection of a constant value (26.5 mm1).
In discrete form, Equation (1)  where x and y are components of AR due to blood and melanin absorption, respectively; and a and are constants accounting for differences in light absorption in blood and melanin, respectively, at the two wavelengths. Since melanin absorption is similar at 585 and 600 nm, 1. Since blood absorption is higher at 585 than at 600 nm, a < 1.
With appropriate selection of a and 3, according to principles outlined by Majaron et al. [12,13], PTT images of epidermal melanin and PWS blood vessels are obtained in separate reconstruction steps. Previous studies [12][13][14] of 1-D depth profiling of PWS skin using this dual wavelength excitation (DWE) approach have demonstrated its advantages over single wavelength excitation (SWE) (e.g. , use of 585 nm alone). In the present study, we focused only on using x to compute PTT images of blood vessels.

Laser Parameters
A commercial pulsed dye laser (ScleroPlusTM, Candela Corp., Wayland, MA) was used. This laser was capable of emitting 1.5 ms laser pulses at wavelengths of 585 and 600 nm. Subtherapeutic light doses of -3 J/cm2 were used in all experiments, with a laser spot diameter of -7 mm.

Infrared Detection System
An InSb focal plane array camera (Galileo, Raytheon, Dallas, TX) imaged time resolved changes in infrared emission after pulsed laser irradiation. The nominal detection band of the camera was 3-5 .tm. To obtain a uniform J.tIR value (see Section 2.1), a filter reduced the band to 4.5-5 .tm.
A unity magnification objective lens (f/2, 50 mm diameter) imaged a 1.92 x 1.92 mm area on to a 256 x 256 array of detector elements with dimensions of 30 j.tm x 30 j.tm. The high acquisition rates (700 or 1000 frames per second) required reduction of the number of active detector elements to a 64 x 64 array. Images were transferred to PC memory with a 12-bit A/D converter.

Hamster Window Model
The hamster dorsal skin flap window chamber model [15] has been used to investigation laser light interactions with microcirculation [4]. A Syrian Golden hamster was anesthetized with a 4:3 ratio mixture of ketamine and xylazine (0.1 mL/100 g body mass). The back and flanks were shaved and excess hair removed with a depilatory cream. Dorsal skin was lifted and sutured to a clamp. One complete layer of skin was removed to expose the underlying dermis and fascia of the opposing layer. A surgically implanted aluminum chamber had apertures to provide simultaneous access to the epidermal and dermal sides of the skin flap.
A color video image of the window was obtained using a CCD camera (Sony, Japan) connected to a stereo microscope (Wild, Switzerland) . The animal was then positioned so the window was at the focal plane of the infrared camera. The laser handpiece was positioned perpendicular to the optical axis of the camera. A microprism in front of the camera directed laser light towards the window. A pulse of 585 nm light was applied to the window and 700 infrared images were acquired at 700 frames per second. A second image sequence was obtained after application of a 600 nm pulse to the same location. A blackbody calibration source (BB7O1, Omega Engineering, Stamford, CT) was used to obtain infrared emission images at various temperatures over a range 20-96°C.

PWS Subject
Infrared image sequences were acquired from an adult subject with a PWS on the left upper extremity. A region on the back of the hand was selected for imaging. Since the stratum corneum has substantially different thermal properties from epidermis and dermis, it was removed using the tape stripping technique [16]. Pads soaked with ethanol were used to remove residual adhesive and tissue fragments, and the stripped area was rehydrated with isotonic saline soaked gauze. The experimental protocol was similar to that used with the hamster window model (Section 2.5), except that 1000 images were acquired at 1000 frames per second. Instruments, Austin, TX) was used to process infrared image sequences. Each sequence was calibrated with the blackbody images. Background temperature was subtracted from each image to obtain '585 and AR sequences. DWE principles outlined by Majaron et al. [12,13] were used to determine values of a and . Using Equation (4), the following was derived to compute the component of AR585 and iR due solely to hemoglobin absorption: x = '585 fl '6OO (5) 1-afl' The iterative inversion algorithm was applied to convert  After pulsed laser irradiation of the window at 585 and 600 nm, infrared image sequences were acquired at 700 frames per second. Since infrared images were acquired from the opposing epidermal side of the window, the video and infrared images are mirror images of each other. Thus, the infrared images in Figure 2 were "flipped" laterally to correspond to the geometry shown in Figure 1. In the 585 nm image sequence (Figure 2), focal points of heating were evident in early images (Figure 2b). At 100 ms after laser irradiation, a high emission region along the right edge was evident (Figure 2c). At 250 ms, this region was broader, and a second region of increased emission was evident in the lower right corner (Figure 2d, arrow).
In depth resolved en face cross sections ( Figure 2) of a DWE PTT image reconstructed from the infrared image sequence, the vertical vessel was evident at a depth of 165 .tm (Figure 3a). Note that a full scale contrast stretch was applied to each section to use the entire 8 bit grayscale dynamic range for image display. From this section, the vessel diameter was estimated to be -300 im. At a depth of 255 J.tm, features of a single -200 im diameter angled vessel are evident (arrow, Figure 3b).
En face cross sections of a SWE PTT image reconstructed from an infrared image sequence taken from PWS skin did not reveal features of subsurface blood vessels (Figure 4(a-d)). In the surface section (Figure 4a), epidermal melanin heating was evident. However, in deeper sections ( Figures   4(b-d)), no vessels were evident. In sections of a DWE PTT image reconstructed from image sequences acquired at the same location, three blood vessels are evident at depths of 255-315 jim ( Figure 5). Vessel diameters were estimated at 200-250 jim. In the surface section (Figure 5a), no evidence of epidermal heating was evident. In (c), increased infrared emission due to blood vessel heating was evident. In (d), features of a second region of heating (arrow) were evident.   Best-case lateral spatial resolution of our system (-30 j.tm) is insufficient to identify correctly the presence of two distinct vessels. In reality, spatial resolution of imaging techniques similar to PTT is known to degrade with increasing depth [19] . In PTT images, blood vessel isotherms are not cylindrical in shape, but have local regions of constriction and dilation. Unfortunately, determination of an average vessel diameter is not a trivial task. Sophisticated image processing techniques are required to determine vessel cross sectional vessel at multiple lateral positions. A low pass filter based template matching scheme [21] will be investigated as a potential tool for outlining vessel boundaries and determining more precisely vessel dimensions in PTT images.
In the present study, 100 iterations of the inversion algorithm were performed to compute PTT images. In prior studies [8,17,20], 30-75 iterations were used. As a first step, we used 100 iterations based on preliminary 3-D DWE imaging results [18]. Theoretical modeling and hamster window studies will be used to investigate the role of iteration number on PTT image accuracy.
We are involved in an ongoing clinical study to acquire infrared image sequences from PWS subjects. In the present study, two PTT image examples (Figures 4 and 5) of the same site are presented. Telenkov et al. [8] demonstrated that SWE PTT can image PWS skin. In the present study, the reconstructed SWE PTT image (Figures 4 and 6a) did not provide information on blood vessel characteristics. However, with DWE analysis, the reconstructed PTT image ( Figures  5 and 6b) showed the presence of three blood vessels. From 1-D DWE depth profiling analysis ( not shown here) of the same site, the average epidermal basal layer depth was -200 j.tm. Since the most superficial vessel (255 m depth) was in close proximity to the basal layer, the SWE approach was limited in its ability to differentiate heating of superficial vessels from that of the lower epidermal regions [12]. Use of the DWE approach (Equation 5) allows us to image selectively the vascular component of PWS skin, as evidenced by absence of epidermal heating in the superficial cross section (Figure 5a). To validate these measurements, future studies are planned to correlate vessel diameters and depths determined from PTT images with those identified using optical Doppler tomography [22] and histological analysis of biopsies [23].

CONCLUSIONS
Blood vessel dimensions and lateral position determined from PTT images of hamster window model skin compared well with those measured directly from microscopic video images. PTT images of human PWS skin contained vessels with estimated diameters of 200-250 tm over a 250-320 tm depth range. Use of dual wavelength excitation (DWE) analysis allowed for imaging of shallow vessels.