2.1.

PAM System

A commercially available PAM equipment (PASONO-SKIN, Guangdong Photoacoustic Technology Co., Ltd., Foshan, China) was utilized in this study. The schematic diagram of the PAM system is shown in Fig. 1(a). The core integrated imaging probe is composed of single-mode fiber, fiber collimator, objective lens, high-frequency ringed ultrasonic transducer (central frequency: 35 MHz, bandwidth: 15 to 55 MHz, focal length: 8 mm), and coupling cup. When performing human skin imaging, the coupling cup needs to touch the skin with deionized water, and the two-dimensional (2D) scanning platform drives the integrated imaging probe to perform raster scanning at a speed of 10 mm/s. Considering that both melanin and hemoglobin have strong absorption at 532-nm wavelength, a 532-nm nanosecond pulsed laser was used as the excitation source of PA signals. A low-noise preamplifier (LNA-650, RFBAY) is used to amplify the photoacoustic signal, and then the amplified signal passes through a bandpass filter. Each laser pulse generated an A-line signal in the imaging depth direction, and B-scan tomography image composed of continuous A-line signals was obtained by 2D transverse scanning. For the spatial resolution of the imaging system, our previous experiments showed that the PAM system has a lateral resolution of about $7\mu \mathrm{m}$ and an axial resolution of about $50\mu \mathrm{m}$.^28,33

Fig. 1

Schematic diagram of the PAM system and skin layer segmentation. (a) Schematic diagram of the PAM system; FC1, fiber coupler; FC2, fiber collimator; SMF, single-mode fiber; OL, objective lens; UT, ultrasonic transducer; T1, T2, T3, trigger; DAQ, data acquisition; FPGA, field programmable gate array; AMP, amplifier. (b) A single B-scan PA image. (c) Segmented epidermal layer from (b). (d) Segmented dermal layer from (b). (e) PA lateral ($X\text{-}Y$ plane) MAP images of epidermal layer. (f) PA lateral ($X\text{-}Y$ plane) MAP images of dermal layer.

2.2.

Data Processing and Statistical Analysis

For 3D photoacoustic skin image reconstruction, each A-line signal is processed by median and filtering, then linearly projected to obtain a B-Scan image, and all B-Scan images are synthesized to display the 3D skin tissue structure. Normal 3D skin includes epidermis (stratum corneum and stratum basale), dermis, and subcutaneous tissue. These skin layers are distributed at different depths and have different morphological structures, which can be distinguished by depth information.^28,32,40 In the quantitative melasma classification, to accurately distinguish the distribution characteristics of melanin and blood vessels in different skin layers, a skin tissue stratification algorithm based on MATLAB (R2019b, MathWorks) was programmed to accurately segment skin tissue at different depths. The specific implementation steps were as follows: first, a threshold value is set to reduce the influence of the background noise in the imaging system. Then, the first extreme point of each A-line signal is extracted one by one to form the skin profile line [Fig. 1(b)]. A certain depth from the skin profile is defined as the segmented line in B-scan image, which is attempted based on different epidermal layer thicknesses. Finally, the PA data are intercepted according to the segmented line to obtain photoacoustic images of different skin layers, such as melanin in the epidermis [Figs. 1(c) and 1(e)] and blood vessels in the dermis [Figs. 1(d) and 1(f)]. Moreover, the parameters of melanin and blood vessels can be quantified based on the reconstructed 3D photoacoustic images. We converted the imaging results into grayscale images to analyze melanin in lesional and non-lesional skin. The mean diameter and density of blood vessels were analyzed by counting the vascular pixel values, where the vascular density was defined as the ratio of all vascular pixels in the given area to the total area pixels.

2.3.

Classification Method with PAM

The current classification of melasma is complex and overlapping, for example, some epidermal and mixed types of melasma are not only simply pigmentary abnormalities but also accompanied by vascular involvement. In this case, the imaging depth and specificity of PAM have advantages over conventional optical dermoscopy and confocal microscopy. With the characteristics of PAM, we first located the cortex where melanin was located, then analyzed whether its formation was related to blood vessels, finally subdivided melasma into epidermal M type, epidermal M+V type, mixed M type, and mixed M+V type. The steps of the proposed photoacoustic classification strategy are shown in Fig. 2. The imaging probe of PAM was used to successively aim at non-lesional skin and lesional skin of each patient. Since the skin thickness of each patient is different, each volunteer’s epidermis and dermis thickness needs to be identified from the 3D photoacoustic images. First, the skin layering algorithm was used to extract the first extreme point of each A-line signal in the patient’s B-scan image to form the skin contour. Second, set the skin contour line as the reference coordinate, and select a certain depth downward until the morphological features of blood vessels are found on the cross-sectional maximum amplitude projection (MAP) image. Finally, the depth $Z$ is defined as the dividing line between the epidermis and the dermis. After identifying the thickness of different skin layer, the next step was to locate the depth of pigmentation in lesional skin. If the pigment was distributed only in epidermis, it was epidermal type. If the pigment was distributed in both epidermis and dermis, it was mixed type. Finally, we analyzed the vascular morphology of dermis and compared the mean diameter and mean density of vessels in lesional skin and non-lesional skin. If the mean diameter and mean density of vessels in the lesional skin were greater than those in non-lesional skin, it was M+V type, otherwise were M type. Eventually, melasma was reclassified into epidermal M type, epidermal M+V type, mixed M type, and mixed M+V type.

Fig. 2

Flow chart of melasma classification with PAM. 3D, three-dimensional; MAP, maximum amplitude projection; $Z$, selected sampling depth ($0\text{-}Z$) in Aline data, and the final determined $Z$ is the epidermal thickness. Y, yes; N, no; $D$, the depth when the area of melanin is 0; ${\mathrm{\Phi }}_{\mathrm{ls}}$, the average diameter of blood vessels in the lesional skin; ${\mathrm{\Phi }}_{\mathrm{ns}}$, the average diameter of blood vessels in the non-lesional skin; ${\rho }_{\mathrm{ls}}$, the average density of blood vessels in the lesional skin; ${\rho }_{\mathrm{ns}}$, the average density of blood vessels in the non-lesional skin; M + V type, melanized with vascularized type; M type, melanized type.

3.1.

Verify the Classification by PAM

First, before the human skin was imaged, the volunteers were first required to remove makeup and clean the entire face to rule out the influence of cosmetics on the imaging results. Experienced dermatologists classified the volunteers using conventional detection tools and then performed PAI. Imaging results of epidermal M type were shown in Fig. 3 (patient 1). The color contrast between lesional skin and non-lesional skin under Wood’s lamp was enhanced, and light brown reticular patches without vascular structure were visualized under the optical dermoscopy [Figs. 3(a)–3(h)], so the dermatologist diagnosed it as epidermal M type melasma. The $X\text{-}Z$ cross-sectional MAP images were shown in Figs. 3(i)–3(l), demonstrating that the melanin in the lesion sites showed strong optical absorption properties in the epidermis. The $X\text{-}Z$ cross-sectional MAP images were shown in Figs. 3(i1)–3(l2), indicating the absorption of laser by melanin and blood vessels. According to the defined boundary between the epidermis and dermis, for this volunteer, the subcutaneous $0\sim$ to $112.5\mu \mathrm{m}$ is the epidermis, and the depth below $112.5\mu \mathrm{m}$ is the dermis. It was also found that there was no significant difference and correlated relationship in the vascularity [Figs. 3(m)–3(p), $r=0.288$, $P>0.05$]. Based on above analysis, the type of melasma was epidermal M type, with melanin deposition in the subcutaneous depth of 0 to $112.5\mu \mathrm{m}$.

Fig. 3

Images and statistics of lesions and normal areas of epidermal M type (patient 1) under room light, wood’s lamp, optical dermoscopy, and PAM. (a) Lesions under room light; (b) lesions under wood’s lamp; (c) normal areas under room light; (d) normal areas under Wood’s light; (e), (f) lesions under optical dermoscopy; (g), (h) normal areas under optical dermoscopy; (i)–(l2) photoacoustic images; (m) PA amplitude of epidermal melanin; (n) mean diameter of dermal vessels; (o) mean density of dermal vessels; and (p) correlation statistics of the diameter and density of vessels in the lesions. L1, L2, imaging area of lesional skin; N1, N2, imaging area of non-lesional skin; RL, room light; WL, Wood’s lamp; OD, optical dermoscopy.

Imaging results of mixed M type was shown in Fig. 4 (patient 2). Differences in pigmentation of lesional and non-lesional skin under Wood’s lamp are highlighted, and more melanin accumulated at the lesional sites [Figs. 4(a)–4(d)]. As with the epidermal M type, dark brown reticular plaques were visible under optical dermoscopy [Figs. 4(e)–4(h)], but no vascular structures were revealed. This volunteer was diagnosed mixed M type by experienced dermatologists. According to the structural characteristics of the 3D photoacoustic images, it can be obtained that the thickness of the epidermis is $135\mu \mathrm{m}$, and the dermis is below $135\mu \mathrm{m}$ of the skin surface. Similarly, the PA signal in lesional skin was stronger than that in non-lesional skin [Fig. 4(m)]. Dense punctate melanin was shown in the lesional skin, which was distributed in the interstitial spaces of the blood vessels [Figs. 4(i2)–4(j2)]. Through fine segmentation, it can be analyzed that melanin deposits at the lesional skin to a depth of $165\mu \mathrm{m}$ below the skin surface [Figs. 4(i3)–4(j3)], reaching the dermis. Another factor, the morphology of blood vessels between lesioned skin and non-lesional skin had no significant difference and correlation [Figs. 4(n)–4(p), $r=0.228$, $P>0.05$]. Based on above analysis, the volunteer was diagnosed as mixed M type melasma, with melanin deposited at a depth of 0 to $165\mu \mathrm{m}$.

Fig. 4

Images and statistics of lesions and normal areas of mixed M type (patient 2) under room light, Wood’s lamp, optical dermoscopy, and PAM. (a) Lesions under room light; (b) lesions under Wood’s lamp; (c) normal areas under room light; (d) normal areas under Wood’s light; (e), (f) lesions under optical dermoscopy; (g), (h) normal areas under optical dermoscopy; (i)–(l3) photoacoustic image; (m) PA amplitude of melanin; (n) mean diameter of dermal vessels; (o) mean density of dermal vessels; and (p) correlation statistics of the diameter and density of vessels in the lesions.

Imaging results of epidermal M+V type was shown in Fig. 5 (patient 3). The pigmentation of lesional skin under Wood’s lamp is more obvious than that of non-lesional skin [Figs. 5(a)–5(d)]. Brown reticular plaques and reddish capillaries can be detected under optical dermoscopy [Figs. 5(e)–5(h)], and this volunteer was diagnosed as epidermal M+V type. According to the PA results, the depth of 0 to $165\mu \mathrm{m}$ under the skin surface is the epidermis, and the depth below $165\mu \mathrm{m}$ is the dermis. Further morphological observation found that melanin was only distributed in the epidermis. Figures 5(i2)–5(l2) provided the vascular morphology of the dermis, showing that blood vessels in lesioned skin are larger and denser than those in non-lesions skin. The statistical results further demonstrated that the PA signal, mean vascular diameter, and density were significantly stronger in lesional skin [Figs. 5(m)–5(o)]. Importantly, vascular diameter and density in lesional skin were statistically significant and strongly correlated [Fig. 5(p), $r=0.934$, $P<0.05$]. Based on the photoacoustic analysis results, the volunteer was diagnosed as epidermal M+V type melasma, and melanin was only deposited in the epidermal layer at a depth of 0 to $165\mu \mathrm{m}$.

Fig. 5

Images and statistics of lesions and normal areas of epidermal M+V type (patient 3) under room light, Wood’s lamp, optical dermoscopy, and PAM. (a) Lesions under room light; (b) lesions under Wood’s lamp; (c) normal areas under room light; (d) normal areas under Wood’s light; (e), (f) lesions under optical dermoscopy; (g), (h) normal areas under optical dermoscopy; (i)–(l2) photoacoustic image; (m) PA amplitude of epidermal melanin; (n) mean diameter of dermal vessels; (o) mean density of dermal vessels; and (p) correlation statistics of the diameter and density of vessels in the lesions.

3.2.

Reclassification of Indeterminate Melasma by PAM

There are some clinical cases of melasma that are difficult to determine its type by conventional Wood’s lamp and optical dermoscopy [Figs. 6(a)–6(h)], and dermatologists generally classify these cases as indeterminate type. Figure 6 provides the imaging results of indeterminate melasma (patient 4). First, the 3D PA images were segmented to obtain the epidermis with a depth of 0 to $112.5\mu \mathrm{m}$ [Figs. 6(i1)–6(l1)] and the dermis below $112.5\mu \mathrm{m}$ in depth [Figs. 6(i2)–6(l2)]. By observing the structural characteristics of the dermis, there is no accumulation of melanin in the interstitial space of blood vessels. Figures 6(m)–6(o) show that the PA signal, mean vascular diameter, and density in lesional skin are significantly stronger than those in non-lesional skin, and the vascular diameter and density in lesional skin are statistically significant and strongly correlated [Fig. 6(p), $r=0.959$, $P<0.05$]. Based on the above analysis, it can be concluded that the indeterminate melasma was epidermal M+V type.

Fig. 6

Images and statistics of lesions and normal areas of indeterminate type (patient 4) under room light, Wood’s lamp, optical dermoscopy, and PAM. (a) Lesions under room light; (b) lesions under Wood’s lamp; (c) normal areas under room light; (d) normal areas under Wood’s light; (e), (f) lesions under optical dermoscopy; (g), (h) normal areas under optical dermoscopy; (i)–(l2) photoacoustic image; (m) PA amplitude of epidermal melanin; (n) mean diameter of dermal vessels; (o) mean density of dermal vessels; and (p) correlation statistics of the diameter and density of vessels in the lesions.