Demodex mites are among the most common human parasites, including only two species: Demodex folliculorum and Demodex brevis. Demodex folliculorum is mostly parasitic on eyelash follicles, while D. brevis predominantly inhabits sebaceous glands and meibomian glands (MGs). Demodex folliculorum was first identified in human hosts in 1841, with subsequent confirmation of its presence within the MG ducts in 1875 (1). Furthermore, the prevalence of Demodex infestation demonstrated a significant upward trend with advancing age, peaking in the fifth and sixth decades of life. The infection rate among children is low, at only 13% among those aged 3-15. In contrast, epidemiological data show that the infection rate among people aged 31-50 is 69%, and it approaches 95%–100% in individuals over 70 years (Elston and Elston, 2014; Cheng et al., 2015; Bitton and Aumond, 2021).

Ocular Demodex infestation represents a highly prevalent yet overlooked cause of ocular surface inflammation that contributes to multiple ocular surface disorders, including blepharitis, meibomian gland dysfunction (MGD), recurrent chalazia, and keratoconjunctivitis (Kim et al., 2011; Akkucuk et al., 2023; Yan et al., 2020). Demodex infestation contributes to ocular surface disease mainly through the following mechanisms. The mites can obstruct the sebaceous gland or the MG ducts through mechanical blockage. Demodex mites can also serve as microbial vectors, contributing to the occurrence of disease. Studies have confirmed that D. folliculorum can carry a variety of pathogens such as bacteria, fungi and viruses. Moreover, the mite’s cytoskeleton may serve as a foreign body to trigger a granulomatous reaction, similar to the pathogenesis of chalazion, or stimulate inflammatory and immune responses (Rabensteiner et al., 2019; Cheng et al., 2015; Deng et al., 2024).

MGD secondary to Demodex infestation arises from mechanical ductal occlusion and subsequent inflammatory cascades, and a higher density of Demodex infestation exhibits a positive correlation with the degree of structural and functional deterioration in MGs (Yeu and Koetting, 2025). Liang et al. (2018) demonstrated that D. brevis was associated with severe MGD, characterized by the loss of more than one-third of the MGs in a young patient population. Studies demonstrated that Demodex infestation altered MGs microstructure, including acinar dimensions, fibrosis severity, and meibum reflectivity, with more pronounced effects observed in patients with MGD (4). Moreover, MGs morphological change, mainly including MGs loss or dropout, is associated with the function of MGs (Ding et al., 2024). Clinically, we have observed uneven atrophy of MGs in patients with mite infection. Yin and Gong (2015) also confirmed that uneven atrophy of the MGs did exist, and the nasal segment dropout contributed more significantly than other dropout regions to the development of MGD. Our previous study also demonstrated that Demodex mite infestation could cause uneven gland atrophy and proposed some novel parameters to evaluate uneven atrophy of MGs (Liu et al., 2022; Yu et al., 2022). However, current research on uneven MG atrophy remains limited. There is a lack of effective parameters to accurately assess uneven MGs atrophy in very severe or very mild MGs atrophy. The relationship between uneven gland atrophy and the function of MGs remains unclear. Furthermore, the mechanism of uneven atrophy of MGs caused by Demodex mites is not clear, and further research is needed to explore this.

The energy of a curve is a mathematical tool employed to quantify the shape and characteristics of a curve. A wide range of fields have extensively utilized it, such as computer-aided design (CAD), computer graphics, image processing, and differential geometry. Its applications in medicine are extensive and significant, particularly in the fields of image segmentation (Pamela and Kashyap, 2016) and dual-energy computed tomography (CT) (Lu et al., 2021). By analyzing the attenuation characteristics of rays or tissues, it provides a powerful tool for lesion detection, disease assessment, and monitoring of therapeutic efficacy.

The uneven atrophy of MGs can be reflected in the undulations of the fitted curve of the gland’s lower margin. This links the unevenness to the curve’s morphology. Therefore, we extracted the lower margin curve of the MG and analyzed it using curve energy. This allowed us to investigate whether the characteristics of uneven MG atrophy could serve as an auxiliary diagnostic tool.

Demodex infestation induces a spectrum of ocular pathologies, including lash folliculitis, glandular duct distortion, and focal meibocyte apoptosis, collectively contributing to destabilization of the ocular surface. Current evidence establishes Demodex as an etiological factor in various anterior segment disorders, including chronic blepharitis, conjunctivitis, and MGD (7, 21). Although previous studies have identified blepharitis as a potential risk factor for MG damage, the specific morphological alterations associated with Demodex blepharitis remain poorly characterized. Current literature predominantly focuses on inflammatory markers and clinical symptoms, with limited data quantifying glandular structural changes through advanced imaging modalities (Deng et al., 2024; Sun et al., 2022)

In this study, we recruited patients with or without Demodex infestation, utilizing a quantitative analysis system powered by AI to evaluate and compare the morphological alterations of MGs. The relationship between ocular surface examination and MG morphology was also analyzed. Our study found that compared to the control group, the Demodex group exhibited significantly lower NIBUT. The shortened non-invasive TBUT indicates compromised tear film stability and heightened ocular surface vulnerability, making the ocular surface more susceptible to external stimuli, which can cause discomfort. No statistically significant variation was observed in TMH, CFS and OSDI between the Demodex group and the normal control group. However, previous studies verified that the ocular discomfort and OSDI score were significantly higher in Demodex infested MGD patients, compared with Demodex-negative MGD patients. It is speculated that Demodex infection may be an important factor in causing or exacerbating damage to the ocular surface and MGs (Yan et al., 2020; Pan and Chen, 2021). Liu et al. (2022) have reported that the ocular Demodex-positive group demonstrated a significant elevation in OSDI scores, meibum quality and tear film instability compared to the ocular Demodex-free group. Our results do not fully align with those reported in their studies. We surmised that the probable cause was that our subjects represented earlier-stage disease where clinical signs are subtler, and this aligns with Sun et al. ’s finding that symptom severity correlates with patient age (Sun et al., 2022).

In terms of MG morphology, we found that the changes of the MG morphology in the upper eyelid are more substantially connected with MGD than those in the lower eyelid in our previous study, as other previous studies have shown that the upper eyelid provides clearer and higher-quality images compared to the lower eyelid, minimizing uneven focus and demonstrating stronger correlations with clinical indicators (Deng et al., 2021; Swiderska et al., 2022). Therefore, this study only recruited participants with upper eyelid involvement. In this study, the lid margin score and meiboscore were significantly higher in the Demodex group than in the normal control group. These findings align with prior studies reporting elevated lid margin inflammation and gland atrophy in Demodex blepharitis which is attributed to mite-induced mechanical damage and bacterial vector effects (Yan et al., 2020; Liu et al., 2022). However, conventional metrics such as MG height, width, tortuosity, and density have not adequately captured the overall features of uneven MG morphological changes in prior research (Zhang et al., 2022; Wei et al., 2024). As Arita et al. (2014) demonstrated, these parameters only assess static gland attributes, not distribution patterns, where central glands show significant shortening without changes in width (Wei et al., 2024). Similarly, Zhang et al. (2022) demonstrated that global MG density correlates strongly with MGD severity but may overlook regional pathology patterns. It was observed that the distribution of MG dropout exhibits increased unevenness among patients affected by Demodex mite infestation. Several parameters are introduced to measure data discreteness, such as variance, standard deviation, or the total variation-based uneven index (Liu et al., 2022; Yu et al., 2022). The Uneven index based on total variation, is a novel index to evaluate the overall features of MG atrophy, which refers to the addition of the absolute value of the difference in the height of each adjacent pair of glands in the meibography images and can be used as a quantitative measure of the unevenness of MG atrophy. However, the computational complexity of this metric limits its practical utility in time-constrained clinical settings.

To address these limitations, we introduced the index of AI based Energy Curve in a different part of meibography. Several AI models for MG analysis (Li et al., 2024) have already been proposed. Although ResNet-based methods (Saha et al., 2022) excel at meiboscore grading, they cannot quantify spatial heterogeneity - the key innovation of our metric. This capability echoes Khan et al.'s GAN approach (Khan et al., 2021) for boundary enhancement but provides clinically actionable quantification. The energy curve quantifies uneven atrophy by using the two-dimensional curvature energy of the glandular margin, surpassing the limitations of traditional static metrics such as length and width. It enables rapid, non-invasive, and equipment-agnostic application, but it also imposes stringent requirements on the quality of meibography images. The analysis involves three types of metrics: the raw curve energy value, the tarsal margin corrected value, and the density normalized energy indices. The raw value directly quantifies the absolute curvature of the MG margin but is confounded by the intrinsic morphology of the tarsal. This interference can be mitigated by correcting the tarsal margin. Dividing these indices by gland density further removes scale effects arising from variable gland counts, yielding more reliable results.

In this study, a statistically significant difference was observed in energy curve indices between the Demodex group and the normal control, with the uneven atrophy in the Demodex group being more severe than that in the normal control. This finding left us with one hint that the Energy Curve could distinguish between Demodex infestation patients and DED patients. We hypothesized that the observed relationship may be explained by two distinct pathways: (Rabensteiner et al., 2019): Demodex infestation can directly damage MGs through mechanical obstruction and inflammatory cascades, leading to measurable morphological changes; (Elston and Elston, 2014); pre-existing gland dysfunction may create a favorable microenvironment for mite proliferation. This bidirectional relationship highlights the importance of comprehensive clinical evaluation, rather than relying solely on isolated parameters for diagnosis.

In current practice, identifying Demodex infections typically relies on clinical symptoms and the presence of cylindrical dandruff on eyelashes. Subsequently, confirmation of Demodex infestation is achieved through positive findings from light microscopic examination of eyelashes or confocal microscopy. Sometimes we do not pay attention to the condition of eyelashes, and only do a dry eye examination, which will easily lead to a missed diagnosis. Meibography, a non-invasive imaging technique, enables real-time visualization of MG morphology and is widely integrated into dry eye diagnostic workflows. We are the first to utilize meibography to calculate the Energy Curve to evaluate the morphological unevenness of MGs. However, we found that the calculation result of the Energy Curve may be influenced by the MG density. We attempted to use MG density to correct the Energy Curve. Notably, the normalized Energy Curve index demonstrated robust diagnostic performance (AUC = 0.897), highlighting its potential for clinical implementation. Its simplified visualization facilitates rapid screening by non-specialists and also aids clinicians in making more accurate diagnostic and treatment decisions. By identifying irregular atrophy of the MGs, the energy curve can provide guidance for targeted interventions and monitor treatment efficacy over time. However, it is worth noting that similar MG morphological changes can occasionally be caused by other diseases in clinical practice. Therefore, this index should be used as a supplementary tool to direct Demodex detection or additional clinical assessments, rather than a replacement. Combining the energy curve biomarker with other diagnostic tools can enhance diagnostic specificity and provide a more comprehensive evaluation.

There were several limitations in this study. First, we did not include the parameters from the lower eyelid, which may introduce selection bias given known topographic differences in MG morphology between eyelids. Future research should expand morphological assessments to include lower eyelid meibography, addressing potential topographic variability in gland dropout patterns. Concurrently, we will expand the sample size by including more subjects to enhance the representativeness and generalizability of the study results. Second, under the present inspection conditions, we can not precisely count the number of Demodex mites within each eyelash follicle. Therefore, we were unable to determine the relationship between glandular atrophy and Demodex quantity in individual glands. Future investigations should focus on elucidating the relationship between Demodex mite burden and both clinical manifestations and structural alterations in MG architecture, particularly in early-stage cohorts. Lastly, the cross-sectional nature of this study limits its ability to establish causal relationships. Future longitudinal studies could provide more robust evidence by tracking changes in MG parameters and clinical symptoms over time in a cohort of patients with Demodex blepharitis. Furthermore, we will also explore more refined representations of edge energy. For example, we will integrate multi - scale edge information, enhance sensitivity to curvature, and design adaptive weighting mechanisms for complex boundary shapes. This will further improve the accuracy of quantitative analysis and its practical value in clinical applications.

To quantify the irregular atrophy of MGs caused by Demodex infestation, we developed the energy curve metric. As a non-invasive biomarker, it’s rapid and easy to perform, making it suitable for early diagnosis, especially in settings with limited resources. The introduction of the energy curve also offers a novel approach for future research into the mechanisms of MG uneven atrophy.