Development of The Chinese Version of Ultra-Low Vision Visual Functioning Questionnaire-150

Abstract

Purpose: The purpose of this study was to develop the Chinese version of Ultra-Low Vision Visual Functioning Questionnaire-150 (ULV-VFQ-150) and evaluate its psychometric function.

Methods: A standardized procedure for the translation of ULV-VFQ-150 was carried out, including the forward translation, consistency check, back translation, back review, and coordination. Participants with ultra-low vision (ULV) were recruited for the questionnaire survey. Psychometric characteristics were evaluated using Rasch analysis based on Item Response Theory (IRT), and some items were revised and proofread accordingly.

Results: In total, 70 out of 74 responders completed the Chinese ULV-VFQ-150, of which 10 were excluded because their vision did not meet the criterion of ULV. Therefore, 60 valid questionnaires were analyzed (valid response rate = 81.1%). The average age of eligible responders was 49.0 years (standard deviation = 16.0), with 35% female subjects (21/60). The person measures (ability) ranged from −1.7 to +4.9 logits, and the item measures (difficulty) ranged from −1.6 to +1.2 logits. The mean value of item difficulty and personnel ability were 0.00 and 0.62 logits, respectively. The reliability index was 0.87 for items and 0.99 for persons, and the overall fit is good. The items conform to unidimensionality as indicated by principal component analysis of the residuals.

Conclusions: The Chinese version of ULV-VFQ-150 is a reliable questionnaire for evaluating both visual function and functional vision in people with ULV in China.

Translational Relevance: The Chinese version of ULV-VFQ-150 is a new assessment of the visual function of people with ULV in China.

Introduction

Visual impairment is a condition that affects millions of people worldwide, impairing their daily activities and quality of life. According to a National Low Vision Survey, there were 11.1 million people with low vision in China, with about 20.4% of them severely visually impaired.^1,2 People who have ultra-low vision (ULV) experience profound visual impairment (VA; 20/500–20/1000) or near-total blindness (VA < 20/1000 – light perception).³ Standardized visual acuity tests are often either unreliable or infeasible for people with ULV due to their very limited vision. In addition, conventional visual acuity measures cannot reflect the impact of low vision on people's daily life, nor capture their subjective satisfaction with their visual function.⁴ 

Therefore, there is an increasing need for patient-reported outcomes (PROs) that reflect patients' functional visual ability, particularly for those with severe vision loss. PROs are therefore key to measuring rehabilitation outcomes.⁵ One of the PROs is the visual function scale,⁵ which measures the patient's “functional vision” from four aspects: physical, functional, social, and psychological.⁶ One example of a PRO are various visual function scales, such as the visual function-related quality of life scale developed by the National Eye Institute (25-item National Eye Institute Visual Function Questionnaire [NEI-VFQ-25]),^7,8 as well as the Impact of Vision Impairment (IVI) Profile^9,10 and the Activity Inventory (AI).^11,12 However, most items on the NEI-VFQ-25 typically require the patient's visual acuity that is better than 20/400 (equivalent to a LogMAR visual acuity of 1.3)^7,8; and the IVI Profile intended for persons with visual acuity in the better-seeing eye of less than 20/200 or visual field diameter less than 10 degrees, or both.¹⁰ Only a few items on these scales are suitable for people with ULV. 

A team of experts in low-vision rehabilitation and the Prosthetic Low Vision Rehabilitation (PloVR) research group jointly designed an Ultra-Low Vision Visual Functioning Questionnaire-150 (ULV-VFQ-150¹³; see Supplementary Materials S1). The scale assesses the visual function required by patients in daily life from four aspects: reading, mobility, visual information gathering, and visual motor activities. The questionnaire items cover key factors of visual activity, such as visual contrast, luminance and lighting, size, distance, familiarity, movement, eccentricity, and depth.^13–15 

The ULV-VFQ-150 has been used to assess the visual function and quality of life of patients with ULV in several clinical trials.^15–17 Given that no Chinese visual function scale was available for patients with ULV before our study, we developed the Chinese ULV-VFQ-150 based on its original version through a standardized process of translation, cultural adaptation, and validation. The psychometric properties and measurement validity of the questionnaire in a Chinese ULV population were also reported. 

Materials and Methods

Study Population

During the period from May 2021 to December 2021, potential ULV participants were recruited from different settings, including Zhongshan Hospital affiliated with Fudan University, rehabilitation centers, nursing homes, and the Patient Assistance Group for Retinitis Pigmentosa in China. The study was approved by the Ethics Committee of Zhongshan Hospital, Fudan University (Approval Letter No.: B2021-252R) and adhered to the tenets of the Declaration of Helsinki. All participants volunteered to participate in the study and provided informed consent. 

The inclusion criteria were as follows: (1) age 18 years or older; (2) native Chinese language speaker; (3) without any cognitive impairment; and (4) visual acuity within the range of counting fingers/hand motions/light perception through less than or equal to 20/500 in the better eye as determined by their medical records. Those who were not willing to provide information to the researchers were excluded. 

Information about the subjects was collected before the start, including (1) name, gender, age, and education; (2) visual acuity of both eyes; (3) living situation (living alone/living with family); (4) times of going out to socialize a week; and (5) whether they have ever obtained a driving license. 

ULV-VFQ-150 Information

This scale describes the visual aspects related to daily activities (including contrast, environmental lighting, size, distance, familiarity, motion perception, etc.), covering four functional domains: visual information collection, mobility, visual motor, and reading/shape.³ (The distribution of visual aspects for each functional domain is shown in Supplementary Materials S3). The questionnaire includes a total of 150 questions, and can be completed face-to-face, online, or over the telephone between the investigator and participants. The participants will be encouraged to complete the scale as independently as possible. The researchers can simply answer questions about the items that the participants do not understand. 

Development of The Chinese Version of ULV-VFQ

The translation of the original questionnaire into Chinese followed the basic principles and objectives of the scale translation process proposed by the ISPOR Task Force for Translation and Cultural Adaptation.¹⁸ A team of five ophthalmologists proficient in both Chinese and English was established; and Gislin Dagnelie, one of the developers of the original ULV-VFQ, and Liancheng Yang, a bilingual senior System Manager were invited to participate in the development. The translation process involved the following steps: 

• 1. Forward translation: Two forward translators (authors J.C. and J.W.) performed two independent forward translations (from English to Chinese).
• 2. Reconciliation: The team compared the two forward translations and checked the difficult items, and provided the first draft of the questionnaire.
• 3. Back translation: The third translator (author X.W.) performed the back translation. The back-translation was compared to the original English version to identify differences and discrepancies so that revisions could be made to ensure that the translation was identical to the original.
• 4. Back translation review: We contacted Dr. Dagnelie, one of the developers of the original ULV-VFQ, as well as Mr. Liancheng Yang who is Dr. Dagnelie's team member and has proficiency in both Chinese and English, by email to review the back translation. They provided suggestions to clarify any differences between the translated version and the original scale, and all problematic items were modified according to the suggestions.
• 5. Cognitive debriefing: The Chinese version of the questionnaire was administrated to five randomly selected outpatients with ULV to identify any difficulties they encountered when completing the questionnaire and to verify whether they fully understood the meaning of each question. The patients were interviewed via telephone or the internet.
• 6. Review and finalization: After making some minor modifications, we determined the final version of the questionnaire (see Supplementary Materials S2).

Item Response Theory and Rasch Model

The Item Response Theory (IRT), a new standard for scale development and improvement,¹⁹ has clear advantages over the Classical Test Theory (CTT).²⁰ The Rasch model is a latent trait model based on IRT, first proposed by the Danish psychometrician Georg Rasch in 1960. As a gold standard for evaluating item response scales,²¹ the Rasch model assumes that the probability of success is only affected by the item difficulty and the individual ability, both of which are located on the same underlying logit scale. By using the linear logit scale, the probability of a correct response can be transformed into a logarithmic odds unit (i.e. logit) to enable continuous, interval-level measures for the person ability and the item difficulty.²² Previous studies have shown that the Rasch model has strong advantages in data analysis for adjusting and optimizing scale items.²³ 

In the current study, the participants’ responses were recorded in the reverse order of difficulty level, as defined numerically (1 =  impossible to see or do visually; 2 = very difficult; 3 = somewhat difficult; and 4 =  not difficult); responses coded as “not applicable” were treated as missing data and excluded from the analysis. The Andrich polytomous model for joint maximum-likelihood estimation, a popular polytomous Rasch model,²⁴ was used for Rasch analysis by using WINSTEPS (version 5.1.7.0; www.winsteps.com). 

We assessed the Chinese ULV-VFQ-150 by the following measures: (1) Person-Item Map (Wright map): The Rasch model transforms the individual ability and item difficulty of nonlinear data into equidistant logit values by logarithmic transformation, and then puts both into the same scale. The Wright map clearly presents the targeting of the test to the sample, as well as the targeting of individual items to persons. In general, a difference between the mean person and item measure of more than 1.0 logit indicates notable mistargeting.²⁵ (2) Category Probability Curve: A category probability curve is used to assess the behavior of a rating scale and explore if the category thresholds are ordered. Rating scale functioning can be assessed visually on a category probability curve graph which displays the likelihood of each category being selected over the range of measurement of an underlying trait.²⁶ (3) Fit Statistics: Fit statistics are used to assess the reliability and the validity of the scale, testing how well the data fit the Rasch model.²⁷ The overall fit of the data to the model was assessed using the following fit statistics: infit/outfit Mean-square (Mnsq) and infit/outfit Z-std. Both infit and outfit Mean-squares have an expected value of 1 logit, with the acceptable fit criterion of 0.5 to 1.5 logits.^22,28,29 The bubble diagrams are used to visually reflect the fit of item difficulty and the individual ability. Each bubble represents a person/item, the bubble location represents the size of the Infit Z-std. Ideally, the bubble should be close to the central axis (i.e. the position of zero).³⁰ When the Z-std is greater than 4, it indicates that the item or the person is under-fitted (i.e. it suggests that the person may have responded indiscriminately or was disturbed by the content measured by the questionnaire). The distribution of visual aspects and domains supported participants' descriptions of what enabled them to perform certain tasks. (4) Differential Item Functioning (DIF) analysis: Participants' characteristics, such as age, gender, and survey method, may affect their responses to a questionnaire, leading to functional differences in the scale items that favor certain groups. Therefore, DIF analysis was performed to ensure the fairness and effectiveness of the questionnaire. (5) Principal Component Analysis (PCA) – Unidimensionality: The Rasch model is subject to logistic distribution and is a logit model based on the assumption of a single dimension, which means that the answer of each item in the response to each item in the scale must be independent of others. When the scale does not meet the assumption, it fails to have unidimensionality.³¹ For a unidimensional scale, the principal components should interpret most of the variance. Unidimensionality requires that the explained variance of the original data must be greater than 60% and the unexplained variance of the first component less than 10%. 

Results

Participant Demographics

A total of 70 out of 74 filled Chinese ULV-VFQ-150 questionnaires (response rate = 94.6%) were collected, 10 of which were excluded because the participants' vision did not meet the inclusion criteria. Therefore, 60 valid questionnaires (valid response rate = 81.1%) were analyzed. Table 1 depicts the demographic and other characteristics of the study participants (n = 60). The median age was 44 years (range = 21 to 76 years), with the mean of 49.0 years (SD = 16.0). Twenty-one (35%) participants were women. Best-corrected visual acuity in the better-seeing eye was recorded. 

Table 1.

 

Characteristics of Sample as Number (%)

 

Person-Item Mapping

In the Rasch model, item difficulty and individual ability (visual function and functional vision) were plotted side by side on the Wright map (person-item map). As shown in Figure 1, the vertical axis is a logit scale, and the letter M on both sides of the vertical axis represents the mean value of each variable. The left side of the vertical axis displays the ability distribution of the participants, showing that the subject's ability gradually increases from bottom to top; and the right side of the vertical axis is the distribution of item difficulty, with the item difficulty increasing in ascending order. As shown, the person measure ranges from −1.7 to +4.9 logits; and the item measure ranges from −1.6 to +1.2 logits. The person ability has a mean offset of 0.62 logits compared to the item difficulty mean which is conventionally set at 0.0 logits. Differences between person and item means can be used to assess how well the difficulty of items in the scale matches the abilities of the individuals. If the difference was greater than 1 logit, it indicates a significant mistargeting.^25,32 It can be seen from Figure 1 that the subject difficulty is concentrated at a moderate level, and the distribution is also relatively uniform. Most subjects' abilities were successfully matched with the item difficulty, but the person measures of those with better visual function (22 of the 60 participants, with their visual acuity either 20/667 or 20/500; see Table 1) failed to be matched. It is important to note that ULV-VFQ was developed to capture the visual ability of individuals with severe vision loss, especially those with very limited visual reserve (e.g. light perception).¹³ To better validate the Chinese ULV-VFQ, a population with more profound visual impairment would be more appropriate. To meet this end and also to test the robustness of the estimations, we conducted a sensitivity analysis that compared the item measures with and without those whose person measures failed to be matched by the items (i.e. the participants with their visual acuity ranging from light perception (LP) through 20/500 versus those from LP through 20/1000; see Table 1). The analysis showed that there was virtually no difference (mean difference = −0.0006 logits, paired Student t-test, t = −0.038, P = 0.970). Nevertheless, based on the current data, the overall matching in this study is acceptable. 

Figure 1.

 

Map of persons and items.

Category Probability Curve

The category probability curves did not show any signs of disordered thresholds, as the calibration of the categories increased in an orderly way (shown in Fig. 2). Four response categories were identified for all items (omitting the category “not applicable” as missing data). 

Figure 2.

 

Category probability curves.

Generally, observations in the higher categories have to be produced by higher measures, and the average measures by category must progress in a monotonic way on the rating scale. Otherwise, the meaning of the rating scale is uncertain for the data set, and consequently, any derived measures are of doubtful utility.³³ The rating scale diagnostics output is shown in Table 2. All category frequency counts are large, which means that locally stable estimates of the rating scale structure can be produced. The average measures increase monotonically with the rating scale category from −3.17 to −0.78 logits (a jump of 2.4), from −0.78 to 0.58 logits (a jump of 1.4), and then from 0.58 to 1.91 logits (a jump of 2.5). The table also shows the fit of each rating scale category to the unidimensional Rasch model, well under the criterion of Mean-square statistic less than 2.0.³³ Taken together, the step calibrations are ordered. 

Table 2.

 

Analysis of Rating Scale of Chinese ULV-VFQ-150 Data

 

The category probability curves illustrate the ordered threshold. The four curves from left to right represent four response categories (1 =  impossible to see or do visually; 2 = very difficult; 3 = somewhat difficult; and 4 =  not difficult). 

Fit Statistics

The Rasch analysis indicated high individual reliability (0.99) and high item reliability (0.87), with acceptable levels for both measures. This suggests a high degree of scale reliability and estimation precision for most participants. 

The average infit Mean-squares (Mnsq) for persons was 0.85 and for items was 0.96, whereas the average outfit Mnsq for persons and items were both 0.84. The expected value of both infit and outfit Mean-square is 1, and an acceptable criterion for fit is between 0.5 and 1.5.²² Our findings demonstrate that both the items and persons provided reliable contributions to the measurement of the latent variable, namely visual ability. This result further indicates that the data are in alignment with the Rasch model. Bubble charts can help to identify potential items that are not suitable for the Rasch model. As shown in Figure 3 and Figure 4, the overall performance of item infit and outfit Mean-square (Mnsq) and person infit and outfit Mean-square (Mnsq) were acceptable with satisfactory fit to the Rasch model. Item 12 and 6 participants’ infit and outfit Mnsq exceeded 1.5 (underfit). 

Figure 3.

 

Bubble chart of item Mean-square (Mnsq). Standardized fit statistics: item (N = 150) measures. Infit Mean-square (top) and Outfit Mean-square (bottom).

Figure 4.

 

Bubble chart of person Mean-square (Mnsq). Standardized fit statistics: person (N = 60) measures. Infit Mean-square (top) and Outfit Mean-square (bottom).

In Z-std bubble charts, z-scores greater than 4 are generally considered unfit.⁶ Figure 5 is the item Z-std bubble chart that shows that all bubbles were basically positioned at a difficulty value of zero, except for the same item (item 12, “When entering a building, how difficult is it? To adjust from a bright sunny day to the dim interior lighting after 5 minutes?”). The z-score for this item was over 4 (underfit). Its purpose was to assess the ability of the adaptation to light changes, which is subjective and variable. In our study, 70% of the participants had retinitis pigmentosa, which is associated with poor light adaptability.¹³ Despite this, the questionnaire demonstrated good overall fitting statistics. 

Figure 5.

 

Bubble chart of item Z-std. Standardized fit statistics: item (N = 150) measures. Infit Z-std (top) and Outfit Z-std (bottom).

In the person Z-std bubble chart (Fig. 6), the z-scores of the same 6 participants were greater than 4, which indicated that their visual function responses were more difficult to predict. 

Figure 6.

 

Bubble chart of person Z-std. Standardized fit statistics: person (N = 60) measures. Infit Z-std (top) and Outfit Z-std (bottom).

After carefully reviewing the details of these participants, there were no specific findings concerning their severity of visual impairment, age, and sociological characteristics (Supplementary Material S4). Of note, people who overfit were more predictable and do not add any new information for the model fitting. 

Visual Domains and Aspects

In Figure 7, item measures are color-coded according to visual aspects (top) and visual domains (bottom). The distribution of visual aspects is similar to that of its original version¹³: items associated with environmental lighting and luminance were simpler to determine (as indicated by lower values in the figure); in contrast, items determined by motion perception, size, and familiarity required more visual effort (as shown by higher item measures). The range of values for items influenced by contrast was wider. 

Figure 7.

 

Item measures as visual aspects (N = 150; top) and domains (N = 150; bottom).

For the visual domains, items determined by mobility were easier (lower in the figure), whereas items determined by reading/shape require more vision (higher item measures); items determined by information gathering ranged wider. 

DIF Analysis

One of the requirements of the Rasch model is that the scale should be coordinated between different subgroups (e.g. age and gender).³⁴ Because all the subjects included in this study were Chinese, there was no difference in the item function in terms of ethnicity. To determine whether there was any difference in the response among other characteristics of the participants, DIF analysis was conducted through three dimensions as follows: (1) online versus telephone administration, (2) gender (male versus female), and (3) age (>49 years versus ≤49 years, because the mean age of the participants was 49 years old). Because the DIF analysis was performed on 150 items simultaneously to test the hypothesis of no difference in any of the items across different groups, we applied the Bonferroni correction to the tests for 150 items of the scale, so the P value for statistical significance was adjusted from 0.05 to 0.0003.^13,35,36 Along any of the three dimensions examined, no items reached statistical significance, thus no evidence for model differences was revealed. Then, we went through those with P values < 0.05 and |DIF contrast| >1.0 logit (see Supplementary Materials S5) along the 3 aforementioned dimensions, and made reasoned judgment for any evident DIF.³⁷ We noticed that only when comparing female with male subjects, 9 items (item 20, 37, 38, 41, 60, 81, 86, 105, 140, and 9/150 = 0.06) demonstrated |DIF contrast| >1.0 logit (female versus male, ranged = −1.26 −1.11, mean = −0.40, SD = 1.11) with P < 0.05. For a 150-item scale, the potential impact of these items on the measurement could be negligible.^35,38 The items were inspected one by one, and there was no obvious bias against male or female subjects visually. Therefore, the Chinese ULV-VFQ-150 is largely free of DIF. 

Principal Component Analysis

The PCA of the residuals is one of the methods to determine whether a scale is unidimensional. For a unidimensional scale, the principal component should explain most of the variance. Besides, a level of 60% of the variance explained by the raw data and less than 10% of the unexplained variance of the first residual component are necessary for a unidimensional scale.³⁹ 

The explained variance was estimated to be 63.2%, and PCA was performed on the unexplained variance or the residual (36.8% of the variance) to discover any systemic deviation from the model. As shown in Table 3, the first principal component explained only 6.9% of the total variance, and each component after that explained even less. It met the criteria for unidimensionality, as described above. There was no major systematic deviation from the Rasch model, which indicated that the data were consistent and unidimensional, and the scale measurements were meaningful. 

Table 3.

 

Principal Component Analysis

 

Discussion

The ULV-VFQ was originally developed in English. Jeter et al. reported that ULV-VFQ-150 has excellent psychometric properties and superior measurement validity in the American ULV population.¹³ In this study, the Chinese version of ULV-VFQ-150 was first developed and then psychometrically assessed by using the Rasch model in a Chinese ULV population. The present study confirmed that the Chinese version of ULV-VFQ-150 is a reliable, valid, and unidimensional questionnaire. 

Several scales measuring visual function or vision-related quality of life have been translated and validated, including the National Eye Institute Visual Functioning Questionnaire (NEI-VFQ-25),⁸ Catquest-9SF,⁴⁰ the IVI Profile,⁴¹ etc. Although these questionnaires have been reported to have good psychometric properties, they were not designed to measure the rudimentary visual function in ULV populations. Thus, most of the items in those questionnaires would be rated as impossible by patients with ULV, as a scale can only measure the abilities that fall in the range covered by the items.¹⁴ The Chinese version of ULV-VFQ-150 developed in the present study captures both the visual function and self-report functional vision in the Chinese ULV population. By quantifying the vision, it can not only serve the Chinese patients with ULV seeking low-vision aids but also those participating in the trials of retinal prosthesis implantation, gene therapy, or other novel treatment modalities. 

During the Chinesization of the scale, some necessary modifications have been made in terms of cultural adjustment. For example, something or some activities mentioned in the original text may not be familiar to most Chinese people living in Mainland China. Those things or activities were then replaced by some that shared similar characteristics but were more common in China. The items in question were listed as follows: item 14 (“When sitting down for dinner, how difficult is it to visually locate the silverware on a white table in a dimly lit room?”), silver was replaced by stainless steel because stainless steel tableware with metallic luster shining like silver is more prevalent in China; item 79 (“When shopping in the supermarket, how difficult is it to see the difference between a white bagel and pumpernickel bagel?”), the question aims at the comparison of two similar-shaped objects that vary considerably in color, but the shape of objects does not need to be perfectly round like bagels that are not familiar to most Chinese people. So, the final version became “When shopping in the supermarket, how difficult is it to see the difference between white steamed bread and miscellaneous grain steamed bread?”; item 95 (“When watching a baseball pitcher wind up on a large screen TV, how difficult is it to see his white uniform against the dark green playing field?”), the word “Baseball” was changed to “soccer”; for item 113 (“When walking through the woods on a sunny day, how difficult is it for you to see a deer from 30 yards away?”), we thought it was relatively impossible for a person with ULV to spot a deer in the jungle here in China. The context here, according to the author of the original scale, is that the animal moves around freely and comes up accidentally to the observer, and therefore, a zoo does not fit this point. The distance between the animal and the person can be changed with the size of the animal, thus the final version was determined as “When you are hanging out in a park, how difficult is it for you to see a stray dog 10 meters away?” 

In the fitting statistical analysis, one item and six participants’ parameters were under-fitted. After excluding the better-sighted respondents, no significant change in item measures was found. The effect of outliers on the overall fit was negligible. The category rating scale of the Chinese ULV-VFQ-150 worked in an orderly manner. There was no significant DIF and the scale was confirmed to be unidimensional. These results were consistent with the findings of the original ULV-VFQ survey in the US population.³ 

The current study also had some limitations. First, although the inclusion criterion did not specify the cause of visual impairment, 70% of the 60 eligible patients with ULV had been diagnosed with retinitis pigmentosa. Therefore, it is unable to determine how well the finding of the present study could be generalized to the ULV population with other diagnoses. Second, due to the study's cross-sectional nature, it cannot examine the test–retest reliability of the assessments. Last, due to the difficulty of recruiting individuals with more severely damaged visual function, a higher than expected proportion (37%) of participants with relatively better visual function (visual acuity 20/667–20/500) were recruited in the current study, which compromised the evaluation of person-item targeting to some extent. However, our sensitivity analysis showed that the estimates were robust and not significantly affected by the inclusion of more able individuals. This was demonstrated by comparing the item measures with and without those individuals. 

The ULV-VFQ-150 comprehensively covers different aspects of functional visual domains,³ with detailed content which provides an item bank for other related questionnaires. It has high reliability but also takes a lot of time to complete. Additionally, some items are relatively similar.⁶ In the present study, it took around 40 minutes to 1 hour to complete the questionnaire for each test. In subsequent studies, we plan to develop and validate a shorter version of the Chinese ULV-VFQ that can be used more efficiently. 

Conclusion

The Chinese version of ULV-VFQ-150 has demonstrated good overall psychometric function for the Chinese population, measuring a unidimensional latent trait. It has proven to be a reliable and valid instrument for the assessing the visual function of patients with ULV. Further studies are required to validate the questionnaire in more diverse populations. 

Acknowledgments

The authors would like to acknowledge the important contributions of Liancheng Yang., who reviewed and provided suggestions for the back translation. 

Funding Information: The authors are grateful for the financial support from the Natural Science Foundation of Xiamen (3502Z20227277). 

Disclosure: J. Cong, None; X. Wu, None; J. Wang, None; C. Feng, None; Y. Wu, None; G. Dagnelie, None; Y. Yuan, None