Non-invasive imaging of pathological scars using a portable handheld two-photon microscope

Abstract

Background:

Pathological scars are a disorder that can lead to various cosmetic, psychological, and functional problems, and no effective assessment methods are currently available. Assessment and treatment of pathological scars are based on cutaneous manifestations. A two-photon microscope (TPM) with the potential for real-time non-invasive assessment may help determine the under-surface pathophysiological conditions in vivo. This study used a portable handheld TPM to image epidermal cells and dermal collagen structures in pathological scars and normal skin in vivo to evaluate the effectiveness of treatment in scar patients.

Methods:

Fifteen patients with pathological scars and three healthy controls were recruited. Imaging was performed using a portable handheld TPM. Five indexes were extracted from two dimensional (2D) and three dimensional (3D) perspectives, including collagen depth, dermo-epidermal junction (DEJ) contour ratio, thickness, orientation, and occupation (proportion of collagen fibers in the field of view) of collagen. Two depth-dependent indexes were computed through the 3D second harmonic generation image and three morphology-related indexes from the 2D images. We assessed index differences between scar and normal skin and changes before and after treatment.

Results:

Pathological scars and normal skin differed markedly regarding the epidermal morphological structure and the spectral characteristics of collagen fibers. Five indexes were employed to distinguish between normal skin and scar tissue. Statistically significant differences were found in average depth (t = 9.917, P <0.001), thickness (t = 4.037, P <0.001), occupation (t = 2.169, P <0.050), orientation of collagen (t = 3.669, P <0.001), and the DEJ contour ratio (t = 5.105, P <0.001).

Conclusions:

Use of portable handheld TPM can distinguish collagen from skin tissues; thus, it is more suitable for scar imaging than reflectance confocal microscopy. Thus, a TPM may be an auxiliary tool for scar treatment selection and assessing treatment efficacy.

Introduction

Pathological scars may cause cosmetic problems, and they may result in functional complications such as contractures and subjective symptoms including pain and pruritus, which may severely affect the patients' quality of life. In developed countries, approximately 100 million new cases suffer from scar-related issues each year.^[1,2,3,4] Pathological scars include keloids, hypertrophic scars, and atrophic scars. These scars can be distinguished from normal skin by their abnormal color, increased thickness, uneven surface area, and poor functional quality, which is brought on by a loss of pliability and contraction.^[5] To date, the pathophysiological effects of scar tissue are unclear, and there is a high risk of recurrence. Despite some progress regarding treatment methods, a lack of timely and effective evaluation methods remains.^[6] Several methods to measure scar color, thickness, relief, pliability, and surface area have been tested for clinical application in recent years. For instance, tristimulus reflectance colorimetry and narrow-band spectrophotometry are used to evaluate color; the Dermascan C (Cortex Technology, Aalborg, Denmark) and the Tissue Ultrasound Palpation System (Biomedical Ultrasonic Solutions, Hong Kong, China) have been employed to assess thickness; Silflo silicon polymer (Flexico Developments Ltd., Hertfordshire, United Kingdom) and Phaseshift Rapid in vivo Measurement Of the Skin (PRIMOS; GFMesstechnik GmbH, Teltow, Germany) can be used to evaluate relief; and pliability was measured using a Cutometer Skin Elasticity Meter (Courage+Khazaka, Cologne, Germany), a Dermal Torque Meter (Dia-stron, Andover, United Kingdom), a Reviscometer RVM 600 (Courage + Khazaka), and a Durometer (Rex Gauge Company, Inc., Glenview, IL, USA). Further, various auxiliary measurement methods based on the planimetry principle were applied in previous studies and can be used to evaluate the surface area of scars.^[5] However, these measurement methods are associated with some complications, such as invasiveness, complicated operation, and difficult clinical application, and they cannot specifically reflect the pathophysiological state of pathological scars in vivo.^[7,8] Therefore, research is required for determining how to achieve objective non-invasive scar assessment in vivo.

Two-photon microscopy has become a powerful method to image living tissue without fixation, sectioning, or the use of exogenous dyes or stains. Elastin and collagen can efficiently generate two-photon excited fluorescence (TPEF) and second harmonic generation (SHG) signals, respectively, and thus TPEF and SHG imaging techniques have been applied in imaging microstructures and characterizing the pathological features in diseased skin tissue such as in collagen-related diseases, inflammatory diseases, skin cancer, and aging.^{[9,10,11,12,13,14,15,16,17]} Previous ex vivo studies based on TPEF and SHG imaging techniques have shown that the morphology and content of collagen and elastin differed between normal and scarred skin.^{[10,18,19,20]} Two-photon microscope (TPM) has been used for quantitative analysis of elastin and collagen, and it was suggested as an effective approach for in vivo scar diagnosis. Visualizing the dynamic architecture of normal skin and keloids may inform physicians regarding the evaluation of treatment efficacy.

Here, we used a portable handheld TPM to perform in vivo imaging in patients with scars. Through image analysis, the scar-related indexes (dermo-epidermal junction [DEJ] contour ratio and collagen depth, thickness, occupation, and orientation) were calculated. The extracted indexes were used to diagnose scars and evaluate their physiological state. By further observing the changes in skin lesions before and after treatment, the availability of the indexes was verified. Here, we successfully used a portable handheld TPM to perform non-invasive in vivo imaging of scar patients and to evaluate the outcome in patients.

Methods

Study design and patients

This morphological and observational study was carried out at the China-Japan Friendship Hospital (Beijing, China) from April 1 to September 1, 2022. All patients were recruited from the outpatient department or the ward of dermatology department (China-Japan Friendship Hospital) and were diagnosed and classified by two experienced associate chief physicians or chief physicians.

The following inclusion criteria applied: (1) a pathological scar was clinically diagnosed (including keloid, hypertrophic scar, and atrophic scar), and the lesion area was larger than the field-of-view (FOV); (2) the skin lesions were dry and clean without ulceration or erosion; (3) patients for observation were untreated or were treated previously but not within the previous two years; and (4) patients or their guardians provided fully informed consent for the purpose and procedures of the study and signed an informed consent form. The exclusion criteria were as follows: (1) patients in the process of treatment; (2) ulceration, erosion, crusting, or other conditions affecting image acquisition from the skin lesions; (3) skin lesions that cannot be imaged, such as genitalia, palms, and soles; and (4) the patient's physical condition not meeting the imaging requirements. In addition to the separately recruited healthy controls, healthy skin on the opposite side of the patient's body or approximately 5 cm around the lesion was used as a healthy control.

All patients were informed of the objectives and methods of the study and provided written informed consent before the commencement of the study. The study was approved by the Research Ethics Committee of China-Japan Friendship Hospital (No. QX2022-021-02) and followed the ethical guidelines of the 1975 Declaration of Helsinki.

Experimental setup

The portable handheld TPM [Figure 1A–C] used in this study was based on a self-made 80 MHz Er-doped femtosecond fiber laser operating at 1560 nm. Connected with a frequency doubling module, the 1560 fs-laser functioned as an excitation source at 780 nm. The output power under the objective was approximately 50 mW, which was sufficiently low for human skin imaging. The excited photon was detected using a dual channel detection module with two photomultiplier tubes, H10770PA-40 (Hamamatsu Photonics, Hamamatsu, Shizuoka Prefecture, Japan). The TPEF channel utilized a 420–580 nm bandpass filter (Edmund Optics Inc., Barrington, NJ, USA) to extract autofluorescence signals, and SHG utilized a 375–400 nm bandpass filter (Edmund) to extract SHG signals. The working head of this portal handheld TPM comprised a two dimensional (2D) MEMS scanner (Mirrorcle Technologies Inc., Richmond, CA, USA) and homemade high numerical aperture (NA) focusing optics (NA = 0.9), and contributes to an FOV of 150 × 150 μm.

Figure 1.

Portable handheld TPM. (A) Overview of portable handheld TPM used in this study. (B) Probe and bracket of the portable handheld TPM. (C) Usage of a portable handheld TPM. TPM: Two-photon microscope.

A stitched image was captured using a homemade multimodal multiphoton pathology microscopy platform. Each grid of the stitched image was excited using a 760-nm laser (Insight X3+, Spectra-Physics, Irvine, CA, USA) with an FOV of 200 × 200 μm or 300 × 300 μm. The filters for the AF and SHG channels were FESH0600 (Thorlabs, Newton, NJ, USA) and FF01-389/38-25 (IDEX Health and Science, Rochester, NY, USA). The scanner utilized on our platform was a Model 6215H Optical Scanner (Cambridge Technology, Lexington, MA, USA). To capture a stitched image, the sample was moved accurately by displacement using an MLS203 Stage device (Thorlabs, Newton, NJ, USA).

Image acquisition and analysis

Image acquisition methods

Our image acquisition pipeline is demonstrated as follows. (a) Ensure a comfortable environment and body position. (b) Clean the skin with 75% alcohol by volume or 0.9% saline, keeping the surface of the skin slightly moist. (c) Lock the imaging position (skin lesion): the intersection of the long axis and short axis of the lesion (intersection point); the midpoint of the line connecting the intersection point and the edge of the long axis; and the midpoint of the line connecting the intersection point and the minor axis edge. (d) Lock the imaging position (healthy): the same part of the trunk on the opposite side of the lesion or 5 cm around the lesion.

We took about thrice as many samples in the lesional area as in healthy areas. During each imaging, the stratum corneum was manually positioned as the starting point (0 μm), and 90 consecutive layers were scanned from the starting point; each layer was 2 μm apart and was scanned 40 times. The total imaging depth was 180 μm, resulting in 3600 images per observation.

Analysis

DEJ contour and depth of collagen

Computation of DEJ surface was performed as described previously,^[21] and we calculated this index using a Python 3.8 script (https://www.python.org/) as follows. For a three dimensional (3D) SHG image stack, a global OTSU^[22] was performed to identify a suitable threshold. Then, pixel values exceeding the threshold were defined as collagen areas. The 3D stack was viewed as several columns along the z-axis. For each column, the collagen area was searched from top to bottom. To exclude noise data, the original stack was pre-processed with a 3 × 3 Gaussian filter and a 3 × 3 median filter. Additionally, only five consecutive collagen pixels can be considered "true" collagen areas. After all columns were examined, a depth map that indicated the surface of collagen was generated. In some corner cases, no collagen was found throughout the column. This point was specifically labeled in depth map and was not used in the following computation. A triangular mesh was generated using a depth map, and the area of the DEJ contour (curved surface) was computed by thinking of the surface as triangles joined together. Areas of the triangles were computed using Heron's formula:

$$S=\sqrt[]{\mathrm{s}\left(\mathrm{s}-\mathrm{a}\right)\left(\mathrm{s}-\mathrm{b}\right)\left(\mathrm{s}-\mathrm{c}\right)}$$

Where a, b, and c were the lengths of sides of the triangles and s is the semi-perimeter of the triangles. As described previously,^[23] the DEJ contour ratio was normalized by the area of the flat plane. Depth of collagen was defined as the average depth of all valid data points.

Thickness index

The thickness index was calculated using a Python 3.8 script as follows. A 2D image was first normalized to 0–1 with the minimum value mapped to 0 and the maximum value mapped to 1. Each normalized image was then processed using 2D fast Fourier transform (FFT), and the logarithm of the amplitude component was used. Then, a 2D FFT map was generated where low-frequency signals were in the middle, and high-frequency signals were in the periphery. A collagen image should have more low-frequency signals when collagens appear thicker and more high-frequency signals when collagens appear thinner. To quantify this phenomenon, a line graph was first extracted from the FFT map. Data points in the 2D FFT map were mapped to the line graph by averaging points with the same distance to the center. Curves in a range of 1/μm to 0.022/μm were intercepted, corresponding to a physical thickness of 1–45 μm, and were fitted according to the exponential function y = Ae ^-Bx + C, where B was the morphological characteristic of collagen thickness.

Occupation index

The occupation index was calculated as follows:

Occupation index = Area of collagen/Area of FOV

The area of collagen was manually marked using Cellpose software^[24] from GitHub repository, https://github.com/MouseLand/cellpose. In the "npy" file generated by Cellpose software, each pixel was labeled with an integer, where zero indicated a blank area and other values represented collagen. The area of collagen was defined by counting the number of zeros, and the area of FOV was identified by counting the number of pixels. All counting tasks in the "npy" files were performed using Python 3.8.

Orientation index

The orientation index was used according to a previous study^[25] to describe collagen orientation with our original code in Python 3.8. We first normalized our 2D image by mapping the minimum value to 0 and the maximum value to 1. Each normalized image was processed with 2D FFT, and the logarithm of the amplitude component was used. Then, a 2D FFT was performed, and all data points were transformed to their logarithm. Thresholding^[26] was used to binarize the 2D FFT image. Then, the outer contour of the strong-signal area was marked and fitted by an ellipse model on the binarized image. This orientation index was defined as one minus the ratio of the length of the minor axe to that of the major axe.

Statistical analyses

Statistical Product and Service Solutions version 21.0 (IBM Corp., Armonk, NY, USA) was used for statistical analyses. Continuous data were expressed as mean ± standard deviation. All variables were normally distributed, and independent-sample t-tests were used. Statistical significance is reported at P <0.05.

Results

General information of the studied population

Fifteen patients were included (six males and nine females; age range 15–65 years; the duration of disease ranged from 1 month to 360 months; 11 patients had keloids, one had two hypertrophic scars from surgery and acne, two had hypertrophic acne scars, and one had an atrophic scald scar; five of the 15 patients had biopsies), as were three healthy controls (two males and one female; age range 26–59 years) [Supplementary Table 1, http://links.lww.com/CM9/B570].

High-resolution 2D and 3D images of healthy and scarred skin

In total, 368 data sets were obtained (each data set contained 90 images, and each image was superimposed by 40 consecutive scans). After excluding data with poor image quality that could not be analyzed, a total of 198 data were retained for analysis.

Normal skin

We used a self-built multimodal multiphoton pathology microscope to obtain the same FOV picture as the reflectance confocal microscopy (RCM) as much as possible and to compare the advantages and disadvantages of the RCM and TPM as accurately as possible. RCM and self-built multimodal multiphoton pathology microscopy images of in vivo normal skin were shown in Figure 2. Figures 2A,2B showed in vivo RCM images obtained from the spinous layer of skin. The cells were arranged in a honeycomb shape, and the local refraction signal was high, showing a medium to low refraction signal, which may be related to the amount of pigment in the cells. Figures 2C,2D showed the DEJ of normal skin under RCM. The ring structure with high refraction was the stratum basale, which is composed of basal cells. The structure of the dermal fibers was unclear. Figures 2E,2F showed the spinous layer cells under multiphoton pathology microscopy. Compared with the RCM, the typical honeycomb structure as observed previously^[27,28,29] can be seen; the signal intensity of the nucleus was low, and that of the cytoplasm was higher than that of the nucleus. DEJ under multiphoton pathology microscopy was shown in Figures 2G,2H. The green-colored area represented the epidermal cell structure, and the high collagen content in the DEJ gave rise to strong second harmonic signals (red). Figure 2H showed elastic fibers as an irregular, punctate, green-colored area between the red-colored collagen in the superficial dermis. Comparing RCM and multi-photon pathology microscope images, the multi-photon pathology microscope produced clear images of epidermal cells. Collagen fibers reflected by the second harmonic signal enabled the multi-photon pathology microscope to provide more collagen-related information than the RCM and distinguished elastic fibers from collagen fibers. In contrast, RCM was unable to distinguish between collagen and elastin fibers.

Figure 2.

Comparison of multiphoton pathology microscopy and RCM healthy skin imaging. (A) Spinous layer cells under RCM; (B) partial enlarged view of A; (C) DEJ under RCM; (D) partial enlarged view of C; (E) spinous layer cells under multiphoton pathology microscopy; (F) partial enlarged view of E; (G) DEJ under multiphoton pathology microscopy; and (H) partial enlarged view of G. Scale bars indicate 200 μm in A, C, E, and G, 100 μm in B, 65 μm in F and 85 μm in D, and H. In all figures, TPEF signals are displayed in green, and SHG signals are displayed in red. DEJ: Dermo-epidermal junction; RCM: Reflectance confocal microscopy; SHG: Second harmonic generation; TPEF: Two-photon excited fluorescence.

In vivo 3D portable handheld TPM features of healthy and scarred skin

The 3D perspective of healthy skin was shown in Figure 3A and Supplementary Videos 1 [http://links.lww.com/CM9/B571] and 2 [http://links.lww.com/CM9/B682]. The green-colored area was the epidermal layer, and the red-colored area was the dermal collagen structure. The green-colored area in the red-colored area represented elastic fibers in the superficial dermis. Figure 3B–F showed the structures of normal skin at 20 μm (the transition stage from granular layer to spinous layer), 40 μm (DEJ), 60 μm (the interwoven area of elastic fibers and collagen fibers in the superficial dermis), 100 μm (deep superficial dermis, obvious collagen fibers can be seen), and 120 μm (the collagen signal almost disappeared), respectively. The 3D perspective of pathological scars was shown in Figure 3G and Supplementary Videos 3 [http://links.lww.com/CM9/B573] and 4 [http://links.lww.com/CM9/B574]. Figure 3H–L showed a scar at 20 μm (stratum corneum to stratum granulosum area: cells with different shapes and sizes, larger nuclei, different cell spacing, and strong TPEF signal around nuclei), 40 μm (near stratum granulosum, cells with relatively uniform morphology), 60 μm (stratum spinous: cells arranged relatively neatly), 100 μm (lack of obvious basal cell structure,^[19] unclear DEJ, epidermis in green-colored area, collagen fibers in red-colored area, barely visible elastic fibers), and 120 μm (dense thickened collagen fibers with occasional elastic fibers), respectively. In general, the epidermis of scars was thickened, the density of epidermal cells was reduced, cell morphology was changed, the DEJ disappeared, elastic fibers in the superficial dermis were reduced, collagen fibers were thickened, the density was increased, and the depth of collagen was increased.

Figure 3.

Healthy and scarred 3D skin slices from 20 μm to 120 μm. (A) Healthy 3D skin perspective; (B) stratum spinosum; (C) DEJ; (D) superficial dermis; (E, F) dermal collagen fibers; (G) scarred 3D skin perspective; (H) stratum granulosum; (I, J) stratum spinosum; (K, L) dermal collagen fibers. Scale bar indicates 20 μm in 2D slices. In all figures, TPEF signals are displayed in green, and SHG signals are displayed in red. DEJ: Dermo-epidermal junction; SHG: Second harmonic generation; TPEF: Two-photon excited fluorescence; 2D: Two dimensional; 3D: Three dimensional.

Changes of DEJ contour and depth of collagen

The signal from the SHG channel could specifically distinguish collagen fibers from cells in the epidermis. As most collagen fibers appeared only in the dermis layer, the surface contour of the junction of the epidermis could be well distinguished from a 3D perspective. We used this feature to analyze the average depth and surface shape of the epidermal junction. The average depth of collagen, which also represented the average depth of the epidermis, was computed by averaging the surface of collagen. Table 1 illustrated that collagen fibers in healthy skin appeared at a shallower depth (approximately 60 μm) in comparison to patients' skin, where they were observed at a deeper level (approximately 110 μm). The DEJ contour ratio suggested that skin lesions showed more undulating folds.

Table 1.

Index comparison of depth, DEJ contour ratio, thickness, occupation, and orientation index on normal and disease cases at 10 μm under collagen.

Index name	Normal	Scar	t-value	P-value
Depth index	67.200 ± 16.000	101.5 ± 26.100	9.917	<0.001
DEJ contour ratio	1.501 ± 0.369	1.881 ± 0.548	5.105	<0.001
Thickness index	0.067 ± 0.029	0.103 ± 0.069	4.037	<0.001
Orientation index	0.705 ± 0.031	0.721 ± 0.028	3.669	<0.001
Occupation index	0.741 ± 0.220	0.814 ± 0.224	2.169	<0.050

Data are presented as mean ± standard deviation. DEJ: Dermo-epidermal junction. DEJ contour ratio: Degree of wrinkling of the junction; Depth index: Average depth when collagen first occurs in field of view; Occupation index: Proportion of collagen fibers in field of view; Orientation index: Collagen orientation; Thickness index: Related to thickness of collagen.

Morphological changes of collagen

Through the SHG channel, the individual collagen fibers and the overall morphology of the collagen fibers could be observed. To describe the morphological changes of collagen, we introduced three indicators: thickness, occupation, and orientation [Table 1]. The thickness index was determined by global features, whereas the orientation index was determined by local features. Collagen areas were circled manually to compute the occupation index. The average thickness index was 0.067 in healthy skin and 0.103 at skin lesions, which indicated stronger and denser collagen fibers. The average occupation of collagen on skin lesions was 81.4%, which was slightly higher than that on healthy skin (74.1%). The average orientation index on healthy skin was 0.705, and it was 0.721 on scars. In the image of healthy skin, the collagen fibers exhibit a random orientation where the fibers are seen in various directions. However, in the lesion image, the collagen fibers show a more organized pattern with smaller directional alignments. Independent-sample t-tests showed that the average depth, thickness, occupation of collagen, the DEJ contour ratio, and the orientation of collagen between normal skin and scarred skin differed significantly (P <0.050). And the average depth, collagen occupation, and DEJ contour ratio between keloid and hypertrophic scars also differed significantly (P <0.050), whereas thickness and orientation did not (P = 0.191 and P = 0.356, respectively).

Treatment efficacy

Four of the five patients received intralesional injections with corticosteroids combined with superficial X-ray radiotherapy (three patients received treatment every 7 days, with a total of four treatment visits; one patient was treated once every 2 days for a total of four treatment visits), and one patient was treated with intralesional glucocorticoids (once every 20 days for a total of four treatment visits). In the patient who received intralesional corticosteroid injection every 2 days combined with superficial X-ray radiation therapy, the depth of collagen was reduced from 126.5 μm to 111.5 μm, while other indexes were not significantly affected. The changes in the remaining three patients were similar. In the case of one patient (this patient received treatment every 7 days, for a total of four treatment visits) [Figure 4A–D], the DEJ was imaged more clearly than before treatment (Figure 4A,C: 60 μm of normal skin vs. 60 μm of scar), the content of elastic fibers in scars was increased (white arrow, Figure 4A,C: 100 μm of normal skin vs. 100 μm of scar), and the structure of collagen was clearer than before treatment (red-colored area, Figure 4A,C: 100 μm of normal skin vs. 100 μm of scar). Using image spectral analysis, we found that the depth, thickness, occupation, and orientation of the patient showed improvement after treatment [Figures 4E,G–H], and the DEJ contour ratio was slightly higher after treatment [Figure 4F]. In conclusion, these results validated the validity of indicators and suggested that TPM could facilitate real-time non-invasive observation of treatment effects in clinical practice.

Figure 4.

TPM and indexes performance of scar patient during treatment. (A–C) 3D slices from a tracked patient at day 0 (T0; A), day 14 (T1; B), and day 24 (T2; C) (the red arrow indicates the location of the scar, the yellow arrow indicates a collagen fiber, and the white arrow indicates an elastic fiber); (D) 3D slices from healthy control (the green arrow represents the location of healthy skin); and (E–H) index changes in the patient. Scale bars in TPM images indicate 20 μm. In all figures, TPEF signals are displayed in green, and SHG signals are displayed in red. SHG: Second harmonic generation; TPEF: Two-photon excited fluorescence; TPM: Two-photon microscope; 3D: Three dimensional.

Discussion

In this study, a portable handheld TPM was successfully used to assess scar characteristics in vivo. By analyzing the image-guided spectra of TPEF and SHG, five indexes (DEJ contour ratio, depth, thickness, occupation, and orientation) were obtained, which differed between healthy and scarred tissue. Additionally, we used a portable handheld TPM to assess therapy outcomes in patients with scars. The results showed that the TPM may facilitate non-invasive and real-time monitoring of the pathophysiological changes of scars before clinical observation.

Based on normal skin pathology, the DEJ, composed of semidesmosomes, anchoring filaments, basal lamina, anchoring fibrils, and elastic fibrils,^[30] is an important physiological characteristic to distinguish the dermis from the epidermis. Therefore, by comparing the depth of the DEJ between scars and normal skin, we obtained the average depth of the first appearance of scar and normal skin collagen, which indicated that the epidermal thickness of skin lesions was deeper than that of uninjured skin and could be used as one of the differential points for early diagnosis of scars. Based on the flattening of the DEJ, we calculated the DEJ contour ratio and the occupation index. The differences in DEJ contour ratio between scar tissue and normal skin suggested that skin lesions showed more undulating folds. Occupation is a novel index that we propose here for the first time, which is defined as the ratio of the area of collagen to FOV (i.e., DEJ or the position where collagen fibers first occur). The difference in occupation index indicates that the area of collagen is increased at the site where collagen first occurs in injured and uninjured skin, which is caused by the disappearance of the DEJ and excessive proliferation of fibroblasts in skin lesions. Thus, the difference between the DEJ contour ratio and the occupation index can also be used as one of the auxiliary criteria for the early diagnosis of scarring. Scars were characterized according to local fibroblast proliferation and excessive collagen production,^[31,32] and previous studies demonstrated that the Fourier transform can be used to analyze the texture of collagen, providing information in the spatial frequency domain, such as the image's spatial frequency distribution.^[33,34,35] Therefore, we used 2D FFT to calculate the thickness of collagen fibers for the first time in this study. The results showed that the collagen in the lesions was denser than that in normal skin. This is in accordance with the histopathological changes of scars and shows that TPM can produce histopathological results non-invasively.

The arrangement of collagen in normal skin is regulated and exhibits preferred orientations, whereas collagen in hypertrophic scar tissue is arranged in a more disorganized manner, sometimes disrupted, and shows swirled patterns.^[19] However, in our study, the calculation results of the orientation index showed that the collagen arrangement in scar tissue was more directional than that in normal tissue. The in vivo TPM images also showed that the collagen in normal tissue was smaller in diameter and disorderly in direction, whereas that in scars was dense and thicker in diameter, and the arrangement of collagen fibers was in an organized direction. The reasons for this discrepancy may be as follows: (1) we used in vivo imaging, and the imaging interface was parallel to the epidermis (X-Y axis), whereas in vitro tissues are mostly perpendicular to the epidermis, and the imaging results may thus differ; (2) differences in scar type, as we included keloids and hypertrophic scars; (3) imaging depth may differ; (4) and differences in imaging devices and computational methods may also introduce bias.

Keloids are fibroproliferative lesions characterized by the production of scar tissue that spreads beyond the limits of the original wound. Hypertrophic scars are elevated red scars that are typically contained within the borders of the original injury. In contrast with hypertrophic scars, keloid scars exhibit a papillary dermis that is non-fibrotic and an overlaying epidermis that is relatively unflattened.^[36,37] However, without histological diagnosis, keloids and hypertrophic scars with similar clinical symptoms are easily misinterpreted clinically. We thus compared the indicators of keloid and hypertrophic scars (due to the small amount of data on atrophic scars, these were not compared with keloid and hypertrophic scars), and the average depth, collagen occupation, and DEJ contour ratio differed significantly, whereas thickness and orientation did not. This indicates that keloids develop on deeper lesions, and they exhibit more collagen in DEJ and more fluctuation in DEJ, compared with hypertrophic scars, which suggests that TPM usage may help distinguish keloids from hypertrophic scars in vivo.

The indexes proposed in this study can relatively comprehensively reflect the status of pathological scars in vivo, which may be used to evaluate treatment efficacy. Further, our results support the applicability of portable handheld TPM in clinical scenarios. Further studies should be conducted to expand the range of diseases, including multicenter clinical follow-up observational studies. Moreover, the optical design and patient interface should be improved, and the rapid scanning of large areas of the skin while maintaining sub-micron resolution should be facilitated.

The limitations of our study include a small sample size for an observational study; thus, further research on a larger sample is required for confirmation. Moreover, the study duration was short. As treatment changes require time, longer observation periods would be of use for confirmation. Finally, we did not obtain pre- and post-treatment histopathological images of patients to evaluate the agreement between TPM and biopsy, but the non-invasiveness of our study is also a strength.

In summary, our study facilitated real-time, non-invasive in vivo observation of the pathophysiological state of scars using a portable handheld TPM. The differences in DEJ contour ratio, depth, thickness, occupation, and orientation may be of use for rapid early diagnosis and differential diagnosis of scars. Observations of patients with pathological scars before and after treatment provide some evidence for the clinical applicability of portable handheld TPM and suggest that the TPM is an important label-free clinical high-resolution imaging tool for in vivo skin histology to facilitate treatment evaluation.

Acknowledgments

The authors would like to thank Yuxin Wang, Mingjie Pan, and Longhao Cao for assisting in imaging TPM pictures; Yang Huang for his technical support; Ying Chen and Dr Ruixing Yu for helping to collect patient data; and all the patients who participated in this study for their cooperation and support.

Funding

This work was supported by grants from Beijing Municipal Science and Technology Commission Medicine Collaborative Science and Technology Innovation Research Project (No. Z191100007719001) and To Establish a Database and Study the Imaging Features of Common Skin Diseases based on Two-photon Imaging Technology (No. SK2021090379-1).

Conflicts of interest

None.