High-frequency quantitative ultrasound in characterizing human skin aging: an exploration of the non-modeling and modeling approaches

Abstract

Background

High-frequency quantitative ultrasound (HQUS) technology, with its non-invasiveness, high-resolution, objectivity, and reproducibility, holds significant potential for characterizing skin aging through the analysis of the internal structure of tissues. This study aims to explore a framework for characterizing skin aging assessment through HQUS technology to facilitate the subsequent analysis of skin aging-related studies.

Methods

In this study, an exploration of non-modeling and modeling HQUS techniques in characterizing skin aging was conducted. In particular, we tested the conventional approach using the envelope amplitude, the non-modeling approach based on the small-window entropy and the modeling approach with the Nakagami parameters (m and Ω) at scanning depths of 1 and 1.5 mm, respectively, and discovered that such a characterization framework is well-suitable for quantifying skin aging. These parameters were calculated based on ultrasound backscattered signals at a high frequency of 42 MHz from the facial skin (from the epidermis to the dermis) of 70 female participants aged 24–57 years and then analyzed using the linear fitting and receiver operating characteristic (ROC) curves.

Results

The results show that there exists a linear correlation between all parameters and participant ages at scanning depths of 1 and 1.5 mm, respectively. Among them, the correlation coefficients for parameter m are r²=0.84 (P<0.0001) and r²=0.65 (P<0.0001), which are higher than those for the relative envelope amplitude, small-window entropy, and parameter Ω. Moreover, the parameter m also has the highest area under the curve among the ROC curves, regardless of the scanning depth.

Conclusions

This characterization framework, especially the modeling of the Nakagami parameter m, has great feasibility for the characterization of human skin aging. The proposed framework holds significant potential for assessing the efficacy of facial rejuvenation products, photofacials, and similar treatments.

Introduction

Skin aging is a progressive process marked by the gradual breakdown of skin tissue, with clinical signs appearing in the third decade of life and continuing throughout life. In dermatology, this process is attributed to the exposome, which encompasses the interaction of chronological, environmental, genetic, and hormonal factors (1,2). Advanced skin aging is associated with various disorders, including dryness, itching, infections, autoimmune conditions, and vascular complications (3). Moreover, photoaging, caused by prolonged sun exposure, results in fine and coarse wrinkles, hyperpigmentation, yellowing, and an increased risk of skin cancer (4). Non-melanoma skin cancers are linked to photoaging (5), and approximately 85% of melanoma cases are associated with excessive sun exposure (6). Therefore, accurately characterizing skin aging is crucial for identifying potential skin-related health concerns.

In clinical practice, dermatologists mainly rely on visual observation using scales (7), a subjective method that does not assess the internal structure of the skin. Histological examination is objective but invasive, can cause scarring, and is primarily used for medical diagnosis (8). Consequently, non-invasive methods are essential for objectively characterizing skin tissue and quantitatively assessing skin aging. Speckle imaging provides valuable microscopic information by analyzing the texture of scattered light on the skin surface, particularly in terms of roughness and structural complexity (9). However, this process lacks a unified quantitative standard, as it primarily analyzes surface texture. Roy et al. (10) attempted to address this issue by proposing a three-dimensional model based on fractal parameter analysis to characterize the surface features of aging skin. Unfortunately, this approach is mostly empirical and cannot assess structural changes within the skin, which are influenced by various internal and external factors. To overcome this limitation, line-field confocal optical coherence tomography (LC-OCT) has been used to assess superficial skin aging, offering high resolution to characterize microstructural changes in the epidermis and superficial dermis (11). Nevertheless, its imaging depth is limited to 500 µm, restricting access to deeper dermal layers. By contrast, ultrasound imaging can penetrate deeper into the skin tissue, where aging primarily affects the subepidermal dermal fibers, well beyond the 500 µm limit of LC-OCT.

Quantitative ultrasound (QUS) techniques are widely used for the non-invasive tissue characterization of soft tissues and organs (12,13). Texture analysis of ultrasound images has been shown to estimate skin aging throughout the dermal layer (14), however, it typically relies on processed grayscale images rather than raw data. In contrast, most QUS techniques utilize unprocessed radiofrequency (RF) signals, which contain richer information. Moreover, the QUS resolution is related to ultrasound transmission frequency: lower frequencies penetrate deeper tissues but offer reduced resolution, while higher frequencies provide superior resolution for visualizing superficial skin layers (15). Given that dermal changes are closely associated with skin aging, skin tissue characterization using high-frequency ultrasound RF data needs to be urgently explored.

QUS techniques include a wide range of methods, primarily spectral-based parameterization, computational complexity, envelope statistics, elastography, and shear-wave imaging. Currently, spectral-based parametrization, computational complexity, and envelope statistics cannot be applied to traditional clinical ultrasound equipment. However, they have recently been successful in many research applications, with new exploratory possibilities (16), all aimed at obtaining valuable information on the microstructure of the underlying tissues. Specifically, techniques based on spectral parameterization are mainly based on the estimation of the backscattering coefficient, attenuation, and physical properties of the scatterers. Numerous applications of these techniques for skin tissue characterization have been proposed (17-20). However, the effectiveness of the methods described in the above literature for quantifying skin aging significantly depends on factors including the behavior of light upon striking and penetrating the skin, the performance of the utilized equipment, and the proficiency of the operator. Consequently, exploring QUS methods that can be applied to skin aging is necessary.

Non-modeling approaches have demonstrated superior performance in tissue characterization, particularly ultrasound kurtosis and entropy. Among these, entropy has been most widely used because it can quantitatively portray alterations in the microstructures of scattering media (21,22) and can be interpreted as a probability density function reflecting physical properties (23). Using raw ultrasound RF data to calculate entropy mitigates biases from demodulation methods. However, limitations such as insufficient spatial resolution, boundary artifacts, and the prerequisites for raw data can restrict its practical use (24). To address these, Tsui et al. (24) proposed the entropy based on a small sliding window for tissue characterization. This approach has gained considerable attention in studies on breast lesions (25), hepatic steatosis (26), Duchenne muscular dystrophy (27), and osteoporosis (28). Meanwhile, the modeling approaches for envelope statistics, including the Homodyned K (HK) and Nakagami distributions, are commonly promoted in tissue characterization. The HK distribution, recognized for its physical relevance, has been applied to tissue microstructure characterization (29,30), including breast lesions (31), fatty liver diagnosis (32), liver fibrosis assessment (33), and lymphoma quantification (34). However, its complexity and unreliable estimation make the Nakagami distribution more favor for tissue characterization (12). The Nakagami distribution provides a simpler approximation of ultrasonic scattering and effectively identifies the backscattered signals reflecting the tissue microstructure. Recent studies have validated its use in thermal ablation monitoring (35), breast tumor characterization (36), cataract detection (37), carotid artery health monitoring (38), myocardial anisotropy assessment (39), and malnutrition evaluation (40).

The primary objective of this study was to explore the feasibility and capability of both non-modeling (small-window entropy) and modeling (Nakagami distribution) approaches based on high-frequency ultrasound for characterizing human skin aging. In addition, our findings may contribute to evaluating the efficacy of facial rejuvenation, early diagnosis of abnormal skin aging, and the adjunct diagnosis of dermatological diseases involving the dermis for clinical applications. We present this article in accordance with the STROBE reporting checklist (available at https://qims.amegroups.com/article/view/10.21037/qims-24-1753/rc).

Methods

Ultrasound acquisition and data set

A future-oriented study was conducted to characterize human skin aging in 70 healthy female individuals. The ultrasound RF data were obtained from the external surface of the facial skin of the participants with a specific focus on the marionette fold and soft triangle regions, which are prominent areas displaying noticeable signs of aging. In this study, the participants were divided into seven age groups, each consisting of 10 women. Their ages ranged from 23.8 to 56.8 years, with a mean age of 39.6 years. Participant characteristics, including age, sex, body mass index, and smoking status, were noted.

Ultrasound scanning and the VISIA skin test

Ultrasound RF signals were acquired using a portable ultrasound research scanner (Vantage 128, Verasonics, Inc., Kirkland, WA, USA) equipped with a linear array and a 42-MHz central frequency linear array transducer (L38-22v CMUT, KOLO Medical, Inc., Jiangsu, China). The transducer has 256 elements and a pitch length of approximately 0.069 mm. Participants in the seven groups underwent standard-of-care ultrasound examinations. Each participant was examined by a qualified ultrasound sonographer at the School of Medicine. Specialized dermatologists performed skin tests using VISIA (CANFIELD Imaging Systems, Canfield Scientific, Inc., Fairfield, NJ, USA) on participants from seven age groups. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments, and was approved by the Ethics Committee of the Affiliated Hospital of Yunnan University (No. 2023183). Each participant provided written informed consent, and all images were anonymized.

Small-window entropy

Shannon proposed entropy as a method for quantifying information (21). Let $K$ be a discrete random variable and $p\left(k\right)$ be the output probability function. $k\in K$, then the entropy is given by $H\left(k\right)$

$$H\left(k\right)={\displaystyle {\sum }_{k\in K}p\left(k\right)\log }p\left(k\right)$$ [1]

where $p\left(k\right)$ ranges from 0 to 1.

In biomedical ultrasound, the entropy of a probability distribution serves as an indicator for interpreting the inherent uncertainty or unpredictability in backscattered RF signals (41). The entropy $H_c$ of the backscattered data can be mathematically expressed as the negation of the logarithm of the probability distribution of the signal:

$$H_c=-{\displaystyle {\int }_{y_{\min }}^{y_{\max }}w\left(y\right)}\log 2\left[w\left(y\right)\right]dy$$ [2]

where $y_{\min }$ and $y_{\max }$ represent the minimal and maximal values of the envelope amplitudes, respectively, and $w\left(y\right)$ denotes the statistical histogram of the RF signals.

Considering the complexity of biological tissues, the entropy value is usually calculated over a small imaging region determined by a sliding window, which is called the small-window entropy. For small-window entropy imaging of the entire frame of backscattered RF signals, the following steps are taken:

1. A local RF signal is acquired using a sliding window from the entire frame of backscattered RF signals.

2. Derive the density function of the local RF signal.

3. Calculate the entropy value of the local RF signal by Eq. [2].

4. The entire frame of the backscattered RF signals was traversed by the sliding window, and steps I, II, and III were repeated.

5. Produce entropy imaging results for the entire frame of backscatter RF signals.

Nakagami distribution

The Nakagami distribution was initially used for radio link modeling (42). Nowadays, the Nakagami distribution is also usually found in QUS, which is given as

$$N\left(x\right)=\frac{2m^m}{\Gamma \left(m\right){\Omega }^m}{x}^{2m-1}{e}^{-\frac{m{x}^2}{\Omega }}U\left(x\right)$$ [3]

where $x$ represents the possible values of the random variable $X$ of the backscattered envelopes, $N\left(x\right)$ represents the probability density function of $x$, $\Gamma (.)$ represents the Euler gamma function, and $U(.)$ represents the unit-step function.

The Nakagami distribution is defined by two parameters, which given as follows:

$$m=\frac{{\left[E\left(X^2\right)\right]}^2}{E{\left[X^2-E\left(X^2\right)\right]}^2}$$ [4]

$$\Omega =E\left(X^2\right)$$ [5]

where $E(.)$ represents the expected value. The shape parameter, represented by m, reflects the information related to envelope statistics. A transition in the m value from 0 to 1 displays a change in the envelope statistics from a pre-Rayleigh to Rayleigh distribution. When the m value exceeds 1, it can be inferred that the backscattered statistics align with post-Rayleigh distributions. In particular, the Nakagami distribution imposes a requirement where the parameter m must be equal to or greater than 0.5. When m satisfies this condition, it is denoted explicitly as the Nakagami parameter (43). The scaling parameter, denoted as Ω, is used to quantify the mean power in the envelope signal.

These two Nakagami distribution parameters can be used for parametric ultrasound imaging (43). Briefly, Nakagami parametric imaging is performed by processing the envelope of the backscattered RF signals using a sliding window. The main steps are as follows:

1. Acquire a local envelope signal with a sliding window from the entire frame of the backscattered envelope signals.

2. Compute the two parameters of the local envelope signal using Eqs. [4,5].

3. Traverse the entire frame of the envelope signals in a certain step and repeat steps 1 and 2.

4. Generate the Nakagami parameter imaging results.

Data processing

Figure 1 presents the algorithmic workflow for evaluating skin aging using the envelope amplitude, small-window entropy, and Nakagami parameters (m and Ω), comprising the following steps. First, a portable ultrasound transducer was used to acquire raw data from facial skin, as described in the section “ Ultrasound scanning and the VISIA skin test”, and delay-and-sum beamforming was performed to generate the RF data. In the demodulation process, the Hilbert transform was applied to acquire the relative envelope of the RF data, and the B-mode images were obtained after logarithmic compression for the relative envelope amplitude. A sliding window traversed the entire imaging region as a specific step to determine the computational kernel. The RF data and the relative envelope in the current computational kernel were processed to calculate the small-window entropy and Nakagami parameters, respectively. As window slides, the calculation results for the entire imaging region were obtained for the characterization of human skin aging.

Figure 1.

Algorithmic flowchart of the different parameter values for characterization of human skin aging. RF, radiofrequency.

Statistical analysis

Dermal skin thickness varies among individuals owing to individual differences, as observed in a survey (44). Signs of skin aging are more apparent in the marionette fold and soft triangle regions, where the average dermal thicknesses are approximately 1 and 1.5 mm, respectively (45). Accordingly, we conducted ultrasound examinations to obtain backscattered RF data at scanning depths of 1 and 1.5 mm, respectively. MATLAB (version R2018b, MathWorks Inc., Natick, MA, USA) was used for offline data processing. The values of the relative envelope amplitude, small-window entropy, and Nakagami parameters as functions of the skin aging level are presented in terms of median and quartile ranges. To compare the correlation between each parameter and the degree of human skin aging, the probability value P and Pearson’s correlation coefficient r were computed (P<0.05 was deemed statistically significant). The area under the receiver operating characteristic (AUROC) curve was calculated by analyzing the receiver operating characteristic (ROC) curves with a 95% confidence interval. All were aimed at exploring the feasibility of non-modeling and modeling approaches for the characterization of human skin aging. GraphPad Prism (version 9.0, Domatics, Inc., Boston, MA, USA) was used for all statistical analyses.

Results

Validation of the database

The study collected a human skin database consisting of RF data from seven age groups, with each group comprising ten individual females. Detailed information about the participants is listed in Table S1. In addition, full-face images of each participant were acquired using the VISIA imaging system to verify whether the ultrasound RF data gathered from a participant conformed to the natural pattern for their age group. A series of typical images of the participants are shown in Figure S1. The participants in Figure S1A-S1G) are sequentially shown in an increasing order of their age. As shown in the figure, with increasing age, the facial wrinkles of the seven subjects gradually became more prominent and deeper. Concurrently, typical aging features, such as enlarged pores and pigmentation, were also observed. This result suggests that the degree of facial skin aging is positively correlated with the age of the participants in the study dataset.

Results of human skin-aging characterization

As an intuitive, qualitative analysis method, the parameter values calculated using non-modeling and modeling approaches (including small-window entropy and Nakagami distribution) can be imagined as color coding. Figure 2 shows the imaging results of the ultrasound B-mode, small-window entropy, and Nakagami parameters for each age group. The four columns from left to right show the results of B-mode, small-window entropy, Nakagami parameter (m), and Nakagami parameter (Ω) imaging. Seven rows from top to bottom show the results from seven age groups of 25–55 years. For the B-mode images and small-window entropy, no considerable differences were observed in the results between the age groups. Nevertheless, the brightness of the Nakagami parameter (m) images decreased with increasing age in each age group, particularly in the imaging results from Figure 2, A3-G3. This demonstrates that such a characterization framework, based on the modeling of the Nakagami parameter m, may be suitable for the quantitative characterization of human skin aging.

Figure 2.

Typical images of ultrasound parametric imaging for each age group. Four columns of imaging results from left to right are B-mode images, the small-window entropy imaging, the Nakagami parameters (m) imaging, and the Nakagami parameters (Ω) imaging, respectively. The seven rows of imaging results from top to bottom are from seven age groups of 25–55 years, respectively.

However, the colors in the parametric imaging results were only given based on the pixel intensity for visualization purposes. Using statistical descriptions of the parameter values to quantify the characteristics of the parametric ultrasound images is recommended. Therefore, we performed more quantitative comparisons using ultrasound parameter values. Specifically, Figures 3,4 present box plots of the relative envelope amplitude, small-window entropy, Nakagami parameter (m), and Nakagami parameter (Ω) for participants in the seven age groups at scanning depths of 1 mm and 1.5 mm, respectively. At a depth of 1 mm, all median values gradually with age as follows: the relative envelope amplitude from 2.02×10³ to 1.43×10³, the small-window entropy from 4.16 to 4.10, the Nakagami parameter (m) from 0.89 to 0.78, and the Nakagami parameter (Ω) from 5.15×10⁶ to 2.58×10⁶. At a depth of 1.5 mm, a declining trend was observed. The median values of the four parameters decreased by 0.54×10³, 0.03, 0.07, and 0.98×10⁶, respectively. This indicates that skin aging causes changes in signal distribution and information uncertainty. This further demonstrated the feasibility and effectiveness of the proposed framework for characterizing human skin aging. More specifically, the relative envelope amplitude and Nakagami parameter (Ω) demonstrated fluctuating decreases, revealing their limited capacity for quantitative characterization of human skin aging.

Figure 3.

Box plots corresponding to different age groups. (A) Differences in the age-group distribution of relative envelope amplitude. (B) Differences in the age-group distribution of small-window entropy. (C) Differences in the age-group distribution of Nakagami parameter (m). (D) Differences in the age-group distribution of Nakagami parameter (Ω). Data are expressed through box plots, and the scanning depth is 1 mm.

Figure 4.

Box plots corresponding to different age groups. (A) Differences in the age-group distribution of relative envelope amplitude. (B) Differences in the age-group distribution of small-window entropy. (C) Differences in the age-group distribution of Nakagami parameter (m). (D) Differences in the age-group distribution of Nakagami parameter (Ω). Data are expressed through box plots, and the scanning depth is 1.5 mm.

The linear fitting for the results of the relative envelope amplitude small-window entropy, m, and Ω were measured at scanning depths of 1 and 1.5 mm, respectively. The results are shown in Figures 5,6. As we can see, the linear fittings at a depth of 1 mm indicate that the correlation coefficients (r² value) between the four parameters and age were 0.46 (P<0.0001), 0.70 (P<0.0001), 0.84 (P<0.0001), and 0.20 (P<0.05), respectively. At a depth of 1.5 mm, the correlation coefficients (r² value) were 0.17 (P<0.05), 0.47 (P<0.0001), 0.65 (P<0.0001), and 0.0014 (P<0.05), respectively. Obviously, whether the depth is 1 or 1.5 mm, the parameter m and small-window entropy present a stronger correlation with age variations than the other two parameters. This suggests that reliance on pixel brightness variations and signal shape changes alone may not comprehensively describe skin aging. Furthermore, it indicates that the non-modeling small-window entropy and the modeling Nakagami distribution are applicable for the high-frequency ultrasound quantitative characterization of human skin aging. Moreover, of the four parameters, the parameter m has the highest correlation value, suggesting that this characterization framework based on the modeling approach may show better skin-aging characterization performance than the non-modeling approach.

Figure 5.

Linear relationship between different ages and parameters. (A) Linear fitting curves of relative envelope amplitude in all age groups. (B) Linear fitting curves of small-window entropy in all age groups. (C) Linear fitting curves of Nakagami parameter (m) in all age groups. (D) Linear fitting curves of Nakagami parameter (Ω) in all age groups. The scanning depth is 1 mm.

Figure 6.

Linear relationship between different ages and parameters. (A) Linear fitting curves of relative envelope amplitude in all age groups. (B) Linear fitting curves of small-window entropy in all age groups. (C) Linear fitting curves of Nakagami parameter (m) in all age groups. (D) Linear fitting curves of Nakagami parameter (Ω) in all age groups. The scanning depth is 1.5 mm.

We calculated the AUROC values of the relative envelope amplitude, small-window entropy, and Nakagami parameters (m and Ω) at six age differences to further explore the feasibility of frameworks for non-modeling and modeling approaches in characterizing human skin aging. Six age differences were produced by subtracting the ages of any two age groups; 6 age differences were produced, which are 5, 10, 15, 20, 25, and 30 years. In the ROC curve, the vertical Y-axis denotes sensitivity, which represents the true positive rate, whereas the horizontal X-axis corresponds to 1 − specificity, which represents the false positive rate. The closer the ROC curve approaches the Y-axis, the lower the false-positive rate. Figure 7 shows the relevant results. More specifically, the AUROC values obtained at the scanning depth of 1 mm using the relative envelope amplitude were 0.51, 0.77, 0.71, 0.68, 0.88, and 0.99; those obtained using the small-window entropy were 0.63, 0.86, 0.87, 0.96, 0.98, and 0.98; those obtained using the Nakagami parameter (m) were 0.62, 0.91, 0.93, 1, 1, and 1; and those obtained using the Nakagami parameter (Ω) were 0.55, 0.60, 0.59, 0.60, 0.51, and 0.98.

Figure 7.

The ROC curves obtained from the ultrasound envelope amplitude, Nakagami parameters (m and Ω), and small-window entropy parameters at intervals of 5, 10, 15, 20, 25, and 30 years, respectively. The scanning depth was set at 1 mm. AUROC, area under the receiver operating characteristic curve; ROC, receiver operating characteristic.

Notably, m consistently showed the highest AUROC values among the computed parameters, suggesting a superior discriminatory ability in this specific context. Simultaneously, the results at a scanning depth of 1.5 mm are similar to those mentioned earlier, as shown in Figure 8. The AUROC values of the relative envelope amplitude varied between 0.51 and 0.90, the small-window entropy demonstrated a progression from 0.71 to 1, the Nakagami parameter (m) ranged from 0.78 to 1, and the Nakagami parameter (Ω) ranged from 0.53 to 0.66. In summary, both the Nakagami distribution and small-window entropy displayed excellent results, particularly the Nakagami distribution. These results indicate the utility of ultrasound small-window entropy and Nakagami distribution for characterizing human skin aging. The findings also demonstrate that frameworks for non-modeling and modeling approaches are feasible for high-frequency ultrasound quantitative characterization. The ROC curves of these models show their performance at different thresholds. To better understand these models, the parameter values under the optimal thresholds were selected and analyzed. Performance comparisons between the B-mode values and ultrasound-quantified parameters at depths of 1 and 1.5 mm are summarized in Tables S2,S3.

Figure 8.

The ROC curves obtained from the ultrasound envelope amplitude, Nakagami parameters (m and Ω), and small-window entropy parameters at intervals of 5, 10, 15, 20, 25, and 30 years, respectively. The scanning depth was set at 1.5 mm. AUROC, area under the receiver operating characteristic curve; ROC, receiver operating characteristic.

Discussion

In this study, we explore the feasibility of using a high-frequency quantitative ultrasound (HQUS)-based framework to characterize human skin aging based on non-modeling and modeling approaches in QUS techniques. Two typical characterization frameworks of the non-modeling small-window entropy and modeling Nakagami distribution have been applied to show their performance superiority. The experimental results demonstrated that conventional ultrasound envelope amplitude cannot quantify skin aging, while both Nakagami distribution and small-window entropy are helpful in characterizing skin aging. Notably, the modeling framework using the Nakagami parameter (m) demonstrates superior accuracy in estimating skin aging compared to the non-modeling small-window entropy. The study findings may contribute to the evaluation of facial rejuvenation efficacy, early diagnosis of abnormal skin aging, and the auxiliary diagnosis of dermatological diseases involving the dermis for clinical applications. It must be emphasized, however, that this study is a preliminary exploration of characterizing human skin aging through the HQUS technique, employing both non-modeling and modeling approaches.

Notably, proposed frameworks based on the QUS techniques for assessing human skin aging should comply with the following requirements: (I) changes in the microstructure of skin tissue during human aging can be described and interpreted from the histological viewpoint; (II) the aging process of human skin tissues can be adequately simulated; (III) vital biological information can be represented. Therefore, it is particularly crucial to choose the appropriate approaches. As mentioned in the Introduction section, the QUS techniques of the non-modeling and modeling approaches have superior tissue characterization performance. Although the non-modeling approaches with increased analytical complexity and cumbersome calculations, such as small-window entropy, may offer significant physical insights, their estimators are challenging to implement, which can affect quantitative estimation. This is why the non-modeling approaches work poorly in some instances (46).

In addition, age-group division in this experiment also needs to be mentioned. Following the “Five Age Groups” rule, the average population lifespan is presupposed to be 100 years. This is then divided into five age groups, each with a 20-year interval. The five age groups are thus 0–19, 20–39, 40–59, 60–79, and >80 years. In this study, we selected 70 women with an age range between the second age group (20–39 years) and the third age group (40–59 years old) for the following reasons: (I) people in the first age group show no visible signs of aging because they are in the growing and developmental stages; (II) people in the second and third age groups tended to be in a more mature phase where a number of physiological and skin tissue changes are apparent; (III) people in the fifth age group have more intricate causes of skin tissue aging, making it difficult to discern the independent effects of skin aging; (IV) people in their first age group include minors, which has ethical implications. Similarly, participants were classified into seven age groups at a 5-year interval to validate the feasibility of the current approach, better capture tissue changes, and observe more distinct microstructural differences. The experimental results demonstrated the validity of this division.

Eventually, there will be some future work to concentrate on the following aspects. Facial skin shares a similar structure and follows the same process as other parts of the body when aging. As such, we consider this method suitable for human skin. However, more comprehensive experimental validation is required before clinical application. In addition, the quantitative assessment of human skin aging has been explored only for women, whereas the same work should be undertaken for men to eliminate the influence of gender. Notably, the number of participants was limited. The dataset had only 70 cases; therefore, the results must be confirmed on a larger dataset before applying our proposed characterization frameworks in a practice setting. Finally, the ultrasound data should be acquired by the same ultrasound expert using the same system settings to reduce the dependence of the ultrasound data on the operator.

Conclusions

Non-modeling and modeling approaches are typical QUS methods that play an essential role in ultrasound tissue characterization. These techniques, which rely on ultrasound RF signals, offer significant potential for characterizing human skin aging. This study uses the HQUS technique to analyze and correlate skin aging in 70 female participants based on recorded data. Notably, it is the first to explore the feasibility of using non-modeling small-window entropy and modeling Nakagami distribution approaches for HQUS-based characterization of skin aging. Furthermore, the results confirmed that the accuracy of skin aging characterization was enhanced using the modeling Nakagami distribution compared to the non-modeling small-window entropy approach.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments, and was approved by the Ethics Committee of the Affiliated Hospital of Yunnan University (No. 2023183). Each participant provided written informed consent, and all images were anonymized.