Photoacoustic microscopy (PAM) has many attributes that lend to in vivo dermatological imaging applications, such as cutaneous microvascular and pigmented lesion imaging. Particularly, these attributes include: high spatial resolution, nondestructive nature, relatively large penetration depth, and high sensitivity to oxy and deoxyhemoglobin (HbO2 and HbR) and melanin. Many reports have detailed the advantages of PAM for microvascular imaging and its potential impact in a variety of clinical applications. 1, 2, 3, 4, 5 Such capabilities have been exploited to provide different structural and functional information and analysis of oxygen saturation (sO2) (Ref. 5), vasomotion and vasodilation,6 and wound and burn healing. 1, 7 However, the majority of PAM achievements imaging microvascular networks have been limited to small animals or phantom experiments, with far fewer reports from in vivo human imaging studies. 8, 9, 10 Here, we demonstrate the feasibility of PAM to image microvascular networks in human skin in vivo. More specifically, cutaneous blood vessel images have been obtained with dark-field PAM11 from both the forearm and palm of a volunteer. We have utilized similar instrumentation as used and reported by Zhang, who also presented a photoacoustic image from a human palm.5 The results of Zhang served as a proof of principle for imaging human microvasculature in general.5 Here we identify key anatomical structures and differentiate palm and forearm skin from in vivo PAM images for the first time. In the acquired images, there are several types of identifiable structures, and significant vascular differences between skin from the palm (acral skin), and skin from the forearm (nonacral skin) are clearly visible, as would be expected for skin from these two regions of the body.
PAM is also well-suited for imaging pigmented lesions and potentially for differentiating benign lesions from melanoma. Currently, melanoma detection relies heavily on clinical suspicion. Typically, lesions are clinically evaluated for asymmetry, border irregularity, color, size (lateral dimensions), and change (evolution, lesion growth). If a lesion raises sufficient concern, it is biopsied and histologically analyzed, and a definitive diagnosis is rendered. There are a large number of pigmented lesions that may have some of the clinical characteristics of melanoma. Although it is not possible, nor clinically feasible, to excise every one of these lesions, many are nonetheless sampled and determined to be benign (i.e., unnecessary biopsies). Conversely, however, not all melanomas fully meet the clinical criteria, especially while in the earliest most curable stage, and may go undiagnosed.
For histologically confirmed melanoma, the lesion thickness, measured from the top of the epidermis, or Breslow's thickness, is the most important prognostic indicator determining propensity to metastases. 12, 13 This measurement is established histologically using an ocular micrometer within the microscope, and is the distance in millimeters from the top epidermal layer (granular layer) to the deep tumor margin within the dermis or subcutaneous tissue. This value is a measurement of depth of invasion and does not provide information regarding tumor size, volume, or other tumor characteristics. PAM has the potential to both match a current gold standard (Breslow thickness) and exceed it, providing three-dimensional information of the entire tumor.5 PAM images could equip physicians with many parameters to improve early diagnosis and guide treatment plans with new prognostic indicators. Here we demonstrate the capability of PAM to characterize a pigmented lesion in vivo. More specifically, a nevus (i.e., a benign tumor of melanocytes) was imaged prior to excision, and the in vivo, three-dimensional PAM data was compared with the histology of the excised lesion.
Photoacoustic (PA) images were acquired with a dark-field photoacoustic microscope assembled in the laboratory.14 The system schematic in Fig. 1 further describes the system. A single element, broadband, focused ultrasonic transducer with a 50 MHz central frequency and 70% nominal bandwidth was utilized in the PAM system. The transducer was purchased from Panametrics (V214-BB-RM), and the focusing lens was made in the lab. Optical energy was provided by an INNOSLAB IS Nd:YLF pump laser (Edgewave GmbH, Aachen, Germany) operating in its second harmonic at a wavelength of 523 nm, with a pulse width less than 10 ns and maximum repetition rate of 3 kHz. With a dye laser (Cobra, Sirah Laser-und Plasmatechnik, Germany), the wavelength was tuned within the ranges of 561–600 nm and 680–710 nm using the following dyes, Rodamine 6G, Pyrromethene 597, and LDS 698 (Exciton, Dayton, OH). The output pulse energy from the dye laser did not exceed 3 mJ, and pulse energy incident on the surface of the target was less than 1.5 mJ. The beam was weakly focused with an optical condenser to a diameter of 2 mm in free space at the focal point. The laser energy deposited on the surface was carefully monitored and never exceeded the ANSI safety limit of 20 mJ/cm2. The total time to collect an image ranged from approximately 5 to 10 min, depending on the scanning step size and field of view. All presented images were acquired with a 20 μm step size in each direction and a laser pulse repetition rate of approximately 500 Hz. All data were digitized and recorded using a commercial data acquisition card (NI PCI 5124, National Instruments Corporation, Austin, Texas) and PC.
Results and Discussion
A small section of the lower palm at the base of the thumb of a volunteer was imaged at a laser wavelength of 584 nm. An area of 8×8 mm2 was scanned. The resulting images, depicted in Fig. 2, are typical of the images obtained from the palms of other volunteers in other imaging experiments. Figure 2b is a cross-sectional or a B-scan image. The bright pixels (bright red, for color images) in this image indicate strong PA signals generated from the surface of the skin and blood vessels below it. From this image, the boundaries of the epidermis, the stratum corneum, and the epidermal-dermal junction are easily distinguishable. Approximately 200–300 μm below the surface of the palm is the epidermal–dermal junction [Fig. 2b]. Large concentrations of smaller blood vessels, namely capillaries in the dermal papillae, are found in this region. Below this depth are larger vessels, presumably of the subpapillary plexus. It is well known that acral skin, like that on the palm or sole of the foot, has a much thicker epidermal layer than skin elsewhere on the body. From the PA B-scan image, the epidermal thickness is approximated as 300 μm; this measurement is within the range of reported values from studies that used different measurement techniques. 15, 16 A maximum amplitude projection (MAP) of the 3-D image toward the skin surface is shown in Fig. 2c, and the cross-sectional slice shown in Fig. 2b is indicated. To make the vessels more clearly visible, the PA signal from and above the epidermal–dermal junction and the unresolved smaller papillary capillary vessels, were removed during construction of the MAP image. The image acquisition time for this image was relatively fast (~10 min); however, faster imaging times will reduce complications such as mechanical movement of the volunteer and enhance the spectroscopic capabilities necessary for functional measurements.
Figure 3 shows representative images acquired with PAM from the forearm. Again, an 8×8 mm2 area was scanned. Also, during construction of the MAP image [Fig. 3c] signal from and above the epidermal–dermal interface was removed to clearly show the dermal vasculature. There are clear differences between the forearm and palm images, particularly visible in the B-scan images [Figs. 2b and 3b]. The epidermal layer on the forearm is not as clearly resolved as in the images of the palm and is indistinguishable from the epidermal–dermal junction. The epidermal layer is much thinner in nonacral skin, as on the forearm, and published reports on the epidermal thickness values for nonacral skin range from tens of microns up to <~100 μm. 15, 16 Both the stratum corneum and the cellular epidermis are thinner in nonacral skin; 17, 18, 19 moreover, the stratum corneum is more compact in acral skin as compared to the less dense basket weave configuration commonly found in nonacral skin. These differences in the epidermis, particularly in the stratum corneum, likely account for the differences in the surface signals from the palm and forearm in the PAM images. In skin with a relatively thick epidermis, such as the palm, signals generated from the epidermis itself and the dermal–epidermal junction are more discrete and easily distinguished on B-scan images. This is in contrast to skin with a thinner epidermis, in which the stratum corneum and epidermal–dermal junction are difficult to differentiate. Furthermore, the denser stratum corneum layer in the palm intrinsically absorbs more laser light and yields a stronger PA signal than the relatively uncompressed, basket woven layer in the forearm, enabling the top epidermal layer in the palm to be detected and resolved. Another difference can be seen in the epidermal–dermal interface. In forearm images, this interface has a much stronger and more diffuse signal. It is believed that the bulk of this signal is generated from the layer of capillary vessels in the dermal papillae that appears to be much denser than seen in the palm images. Microcirculation and microvascular structure are different in acral and nonacral skin. Approximately 75% of microcirculation in acral skin is through arteriovenous (AV) anastomoses, and only 25% through nutritional capillaries.20 This is in stark contrast with microcirculation through nonacral skin, which has far fewer (or no) AV shunts; here blood circulates primarily through nutritional capillaries. 20, 21 These structural and functional differences could account for the strong signal observed in the PA images of the epidermal–dermal interface in the forearm as compared to the weak signal in the palm. In essence, the greater number of perfused capillaries in the forearm generates a signal that is both stronger and broader or more diffuse than the more discrete signal from the palm images.
These imaging experiments demonstrate the potential of PAM to characterize the structure and functioning of human cutaneous microvascular networks in vivo. The ability to accurately image microvasculature in human skin in vivo can aid the study and understanding of microcirculation, in turn aiding in the diagnosis and monitoring of many cardiovascular diseases, including diabetes, renal disease, hypertension, peripheral vascular disease, 22, 23, 24, 25, 26 as well as vascular lesions, such as hemangiomas and angiogenesis of tumors.
A common blue nevus in the forearm of a volunteer is shown in Fig. 4a. It was photoacoustically imaged and subsequently removed with a 3 mm punch biopsy, then histologically analyzed. Prior to excision, a 6 mm × 4 mm area encompassing the nevus was imaged with PAM at two different laser excitation wavelengths, 570 and 700 nm. The first wavelength, 570 nm, was chosen because it is an isosbestic wavelength for HbO2 and HbR, and the signal at that wavelength can be used as a measure of total hemoglobin. At the longer excitation wavelength, 700 nm, the absorption of melanin, HbR, and HbO2 is significantly weaker than at 570 nm, which enables deeper penetration of the laser light into the nevus. In the PA image acquired at 570 nm laser excitation wavelength, the surrounding microvasculature is clearly displayed, along with the nevus, in the MAP image [Fig. 4c]. In contrast, only the nevus and unresolved signal from some capillary vessels in the epidermal–dermal junction are visible in the PA images acquired at 700 nm [Fig. 4d]. The B-scan in Fig. 4b is a representative cross-section (along the scanning direction) of the nevus imaged with 570 nm light and shows the depth of the nevus and surrounding vessels.
Several measurements can be derived from the PA images that cannot be obtained with dermoscopy or photography, including the lesion's depth, thickness, and volume. To validate these measures, sections of the nevus were histologically analyzed. For comparison, it was necessary to select the correct “slice” of the 3-D data that corresponded to the same approximate section of the nevus in the histology section. To match histology with the PA data, the nevus was manually marked in photos from serial histology sections. The spacing between histology sections varied between 50 and 80 μm. A montage was made of the labeled photos and a MAP image of the nevus was formed from the histology data as shown in Fig. 5a. Based on the shape of the nevus and other key features, namely a prominent hair follicle disrupting the nevus at the edge, the histology and PA MAP images were co-registered, and “slices” of the PA data that most closely matched with the corresponding histology sample were selected [Figs. 5b, 5c, 5d]. Depth and thickness measurements were taken from the same position in the PA and histology data. PA data collected using 570 nm laser excitation light yielded the following measurements: 270 μm nevus thickness [Fig. 5e], 140 μm depth from the surface to the top of the nevus (410 μm total depth from the surface) [Fig. 5e], and a 2.66 mm width. The PA data collected with 700 nm laser light produced similar results: 262 μm nevus thickness [Fig. 5f], 135 μm depth from the surface to the top of the nevus (397 μm total depth from the surface) [Fig. 5f], and a 2.64 mm width. The edges of the nevus were defined as the points where the signal first exceeded two standard deviations from the average noise value. In both depth and lateral dimensions, differences in spatial measurements from both wavelengths were within the resolution of the system (15 μm axial and 45 μm lateral resolutions). These results are consistent, at least qualitatively, with the histology, which shows the nevus to be approximately 450 μm thick and located 150 μm deep below the surface (600 μm total depth from the surface), with a width of 2.18 mm. The measurements from the PA data are justifiably different from histology because of tissue distortion during excision, formalin fixation, and dehydration/paraffin embedding that result in tissue shrinkage artifact. We thus anticipated that the histological measurement of lesion width would be less. Taking into account that the histology section was generated from a 3 mm punch biopsy, the width of the nevus could be scaled to the entire tissue slice, yielding a more accurate measurement. From this calibration procedure, the width of the nevus was determined to be 2.7 mm, which agrees well with PA measurements. The smaller PAM values for depth (thickness and depth) compared to histology measurements are also to be expected from distortion of the tissue along the depth direction from excision, which relaxes the lateral tension of the removed tissue and causes the tissue to collapse in the lateral directions and elongate in the depth direction.
A major concern with PA data pertains to the accuracy of the measured thickness of the lesion. From a single PA measurement, it is not necessarily guaranteed that the entire thickness of the lesion has been recorded because the PA data will not inherently indicate if light has fully penetrated the lesion. However, by imaging with two different wavelengths, it is possible to confirm that the entire thickness of the lesion was accurately recorded, if the measurements at both wavelengths are consistent. The primary absorber in a pigmented lesion is melanin and the optical absorption of melanin decays exponentially with increasing optical wavelength. Thus if one wavelength is sufficiently longer, the absorption coefficient of melanin can be an order of magnitude (or more) lower at that wavelength, which results in much deeper light penetration. The PA signal is proportional to the product of the absorption coefficient, μa, and the local laser fluence, which can be assumed to decay exponentially with μa inside a specific target (i.e., nevus, blood vessel, etc.). For example, using two wavelengths, λ1 and λ2, if μa λ1 = 5 μa λ2, at a depth of d = 1/μa λ1 into the target, the PA signal from λ1 will be approximately 11 times greater than the PA signal from λ2, or PA(d)λ1 ~ 11PA(d)λ2. So, if the absorption coefficient is significantly different and the initial laser fluences are roughly equivalent at the two wavelengths, similar thickness measurements will confirm that light has fully penetrated the lesion at both wavelengths and both measurements are accurate. If the thickness measurements are different, it is guaranteed that the shorter measure is inaccurate and it remains unknown if the larger value is correct or not. However, this scenario does not represent a fundamental limitation of photoacoustic imaging because utilizing longer excitation wavelengths and/or increasing the illumination intensity will enable an accurate lesion thickness measurement. The data presented here show similar thickness values using both 570 and 700 nm laser light, which validates the accuracy of the measurement. Furthermore, a blood vessel is clearly evident directly below the nevus in the PA image collected with a 570 nm laser wavelength (1). The presence and detection of this blood vessel confirms that a sufficient number of photons penetrated the nevus, enough to generate the PA signal from the blood vessel. Consequently, if the nevus extended to the depth of the blood vessel seen beneath it, then a detectable PA signal should have been generated from the nevus itself at this depth. The absence of a PA signal from the nevus at the depth of this vessel suggests that the lower boundary of the nevus was detected.10.1117/1.3528661.1
With better characterization and analysis of pigmented lesions, there is great potential to improve the diagnosis and assessment of clinically atypical nevi and melanoma. Using clinical criteria that include asymmetry, border irregularity, color variation, diameter, and evolution, early melanomas that do not exhibit these criteria are missed. Conversely, many benign lesions (atypical nevi) may be needlessly biopsied. The use of dermoscopy has aided clinical assessment; however, this tool is useful only in the evaluation of surface characteristics and is not effectively employed by all clinicians. With PAM, additional characteristics of pigmented lesions may be noninvasively evaluated, including vasculature, lesion architecture, and 3-D symmetry. In suspected melanoma, measurements including lesion thickness and tumor volume may also be assessed in vivo. Once multiple pigmented lesions have been imaged, analyzed, and compared with histologic diagnoses, we anticipate that algorithms will be devised using PAM parameters to aid in the diagnosis of melanoma. This work is currently ongoing. Such information could be helpful not only in diagnosis of melanoma, but potentially in planning definitive surgery.
In conclusion, PAM has been used to collect in vivo images of the microvasculature and a pigmented lesion. Images of the microvasculature clearly show the structure of the vessels and the differences between vascular networks in acral and nonacral skin, and they demonstrate potential for numerous functional measurements. A benign nevus was photoacoustically imaged, biopsied, and histologically analyzed. Thickness and depth measurements of the nevus were consistent in both PA images, acquired using different laser wavelengths, and agreed well with the histology measurements, including the effect of tissue distortion during sample preparation. These results exhibit the capabilities and potential of PAM in dermatological applications.
Patients: All of the experiments were conducted in accordance with the human studies protocols approved by the Institutional Review Board at Washington University in St. Louis. The Declaration of Helsinki protocols were followed and all volunteers freely submitted their written informed consent.
This research was sponsored by National Institutes of Health Grant Nos. 5 T32 AR07284, R01 EB000712, R01 EB008085, R01 CA134539, and U54 CA136398. L.W. acknowledges financial interest in Endra, Inc., which, however, did not support this work.