Illuminating characteristic patterns of inflammatory dermatoses: A comprehensive dual‐imaging approach using Optical coherence tomography and Line‐field confocal optical coherence tomography

Abstract

Background

Inflammatory skin diseases, such as psoriasis, atopic eczema, and contact dermatitis pose diagnostic challenges due to their diverse clinical presentations and the need for rapid and precise diagnostic assessment.

Objective

While recent studies described non‐invasive imaging devices such as Optical coherence tomography and Line‐field confocal OCT (LC‐OCT) as possible techniques to enable real‐time visualization of pathological features, a standardized analysis and validation has not yet been performed.

Methods

One hundred forty lesions from patients diagnosed with atopic eczema (57), psoriasis (50), and contact dermatitis (33) were imaged using OCT and LC‐OCT. Statistical analysis was employed to assess the significance of their characteristic morphologic features. Additionally, a decision tree algorithm based on Gini's coefficient calculations was developed to identify key attributes and criteria for accurately classifying the disease groups.

Results

Descriptive statistics revealed distinct morphologic features in eczema, psoriasis, and contact dermatitis lesions. Multivariate logistic regression demonstrated the significance of these features, providing a robust differentiation between the three inflammatory conditions. The decision tree algorithm further enhanced classification accuracy by identifying optimal attributes for disease discrimination, highlighting specific morphologic criteria as crucial for rapid diagnosis in the clinical setting.

Conclusion

The combined approach of descriptive statistics, multivariate logistic regression, and a decision tree algorithm provides a thorough understanding of the unique aspects associated with each inflammatory skin disease. This research offers a practical framework for lesion classification, enhancing the interpretability of imaging results for clinicians.

1. INTRODUCTION

Inflammatory skin diseases, including atopic eczema, psoriasis, and contact dermatitis represent a significant healthcare burden worldwide, affecting millions of individuals and often resulting in substantial physical and psychological distress. ¹ , ² , ³ Accurate and timely diagnosis presents a complex challenge to clinicians due to their diverse etiological factors and variable clinical manifestations. ⁴ Traditionally, histopathological examination of skin biopsies has served as the gold standard for diagnosing these conditions, providing insights into the microscopic changes within the skin tissue ⁵ ; however, the histological process is time‐consuming, invasive and may delay the initiation of appropriate treatments.

Recent advances in medical imaging techniques have led to the development of Optical coherence tomography (OCT) ⁶ and Line‐field optical coherence tomography (LC‐OCT), ⁷ offering unprecedented insights into skin microanatomy. OCT, with its ability to capture cross‐sectional images of tissue microstructure at micrometer‐scale resolution, has demonstrated remarkable utility in dermatology. The technique relies on low‐coherence interferometry to generate detailed images by measuring the echo time delay of backscattered light. This enables visualization of the epidermal‐dermal junction, epidermal thickness, and architectural changes in skin lesions. ⁸ , ⁹ LC‐OCT, an advanced iteration of OCT, extends the capabilities by incorporating line‐scanning confocal microscopy, providing real‐time dynamic imaging of tissue up to cellular resolution. ¹⁰ , ¹¹

While recent studies ¹² , ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ and case reports ¹⁸ already described the feasible visualization of characteristic pathological features of inflammatory dermatoses, this prospective study aims to investigate these patterns in a large cohort of 140 lesions, employing a dual‐imaging approach that assembles the strengths of both OCT and LC‐OCT.

In the subsequent sections, we present the findings of our study, by use of a multi‐faceted analytical approach. Descriptive statistics illuminate the morphologic differences between eczema, psoriasis, and contact dermatitis lesions. Multivariate logistic regression serves as a statistical compass, guiding us through the intricacies of these visual disparities. Furthermore, a decision tree algorithm, rooted in the calculation of Gini's coefficient, emerges as a dynamic tool for classifying the diseases based on key imaging attributes.

2. MATERIALS AND METHODS

OCT images were acquired using the commercially distributed Vivosight (Michelson Diagnostics, Maidstone, Kent, UK). This frequency domain OCT is based on Michelson interferometry, has a lateral optical resolution of 7.5 μm, and an axial resolution of 10 μm, combined with a penetration depth of up to 1–1.5 mm. A more detailed description of the device is provided elsewhere. ⁸

LC‐OCT was conducted with the deepLiveTM System (DAMAE, Paris, France). Combining the principles of reflectance confocal microscopy (RCM) and OCT, LC‐ OCT has improved penetration depth compared to RCM and improved resolution compared to conventional OCT. Moreover, LC‐OCT can generate vertical and horizontal sections in real‐time, and three‐dimensional (3D) images in seconds. Further technical details are provided elsewhere. ¹⁹

A total of 140 lesions were included in the study, comprising patients with histologically or clinically confirmed atopic eczema (57), psoriasis (50), or allergic contact dermatitis derived from positive patch testing (33). For OCT 88 lesions (41 atopic eczema, 26 psoriasis and 21 contact dermatitis), and for LC‐OCT 134 lesions (56 atopic eczema, 45 psoriasis and 33 contact dermatitis) were acquired.

Each lesion was imaged using the vertical‐ (en‐coupe), horizontal‐ (en‐face), and 3D‐mode (the latter only possible in LC‐OCT) following a standardized protocol.

In determining the morphologic criteria to be examined, we were guided by previously described visual patterns and established histological features. ¹² , ¹³ , ¹⁸ , ²⁰

The following morphologic OCT and LC‐OCT features were analyzed: erosion, serum crust, acanthosis, hyperkeratosis, parakeratosis, spongiosis, inflammatory cells in the upper dermis, bright inflammatory cells in the epidermis, intraepidermal macro vesicle, intraepidermal micro vesicle, elongated rete ridges, dilated vessels, bright inflammatory cells in the dermis, subepidermal macrovesicle, subepidermal microvesicle, munro‐abscess, and vesicles with detached cells.

All images were analyzed by three experienced examiners (C.R., F.D., E.S.) who agreed on the outcome for each patient. All examiners were blinded to clinical and/or histological diagnosis.

Descriptive statistical analyses were employed to characterize the morphologic features of eczema, psoriasis, and contact dermatitis lesions. The occurrence of each criterion was assessed, and a chi‐square test was performed to evaluate the differences among the three disease groups. A post hoc analysis with the Bonferroni method was then applied to compare each combination of diagnosis.

To assess the significance of morphologic differences between the disease groups while controlling for potential confounding variables, multivariate logistic regression was performed using the backward elimination approach in which non‐significance was considered as the main specified level for effect removal: the less significant predictor was removed until all model predictors were considered significant (p‐values <0.05). This statistical approach allowed us to identify key imaging features that were independently associated with each inflammatory skin condition.

A decision tree algorithm comprising three steps (obtained from a code developed by Sébastien Fischman) was constructed to create a classification model based on the imaging attributes identified in the dataset. The algorithm employed the Gini's coefficient to determine the optimal splitting criteria at each node, facilitating the hierarchical classification of lesions into eczema, psoriasis, or contact dermatitis categories. Furthermore, Pearson's correlation coefficients were performed for the correlation of each technique between the different criteria.

Statistical analysis was performed using R software (version 4.3.1).

The study was conducted according to the principles of the Declaration of Helsinki and international guidelines concerning human studies. It was approved by the local ethics committee (Nr. 17‐699), and written informed consent was obtained from each patient.

3. RESULTS

Among the enrolled patients, the mean age was 51 (±18.4) years, with a slight male predominance comprising 54.3%.

3.1. Descriptive statistics

OCT and LC‐OCT imaging confirmed previously reported distinct visual features among the different cutaneous lesions. Atopic eczema lesions exhibited a characteristic pattern of epidermal hyperplasia and spongiosis (Figure 1A), while psoriasis lesions demonstrated epidermal acanthosis and elongation of rete ridges (Figure 1B). Contact dermatitis lesions displayed a diverse range of features, including epidermal vesiculation and variable degrees of inflammatory cell infiltrates (Figure 1C). For detailed descriptive statistics see Table 1.

FIGURE 1.

LC‐OCT images (image size 1.2 × 0.5 mm², lateral/axial resolution: 1.1 × 1.3 μm) showing characteristic morphologic features including atopic eczema (A) with epidermal hyperplasia (white bracket) and spongiosis (white star), psoriasis vulgaris (B) with epidermal acanthosis and elongation of rete ridges (white arrows) as well as contact dermatitis (C) with macrovesicles and detached cells (exemplary white circle).

TABLE 1.

Descriptive statistics of LC‐OCT and OCT imaging in patients with psoriasis, eczema, and contact dermatitis regarding characteristic morphologic features.

	LC‐OCT							OCT
	Psoriasis	Atopic eczema	Contact dermatitis	p‐value	p‐value	p‐value	p‐value	Psoriasis	Atopic eczema	Contact dermatitis	p‐value	p‐value	p‐value	p‐value
	(N = 45)	(N = 56)	(N = 33)	Chi2	Pso vs. E	Pso vs. CD	E vs. CD	(N = 26)	(N = 41)	(N = 21)	Chi2	Pso vs. E	Pso vs. CD	E vs. CD
Erosion	9 (20.0)	8 (14.3)	3 (9.1)	0.4	1	0.66	1	3 (11.5)	15 (36.6)	8 (38.1)	0.06	0.08	0.12	1
Serum crust	8 (17.8)	8 (14.3)	2 (6.1)	0.32	1	0.53	0.93	2 (7.7)	8 (19.5)	4 (19.1)	0.39	0.88	1	1
Acanthosis	34 (75.6)	34 (60.7)	21 (63.6)	0.27	0.41	0.95	1	24 (92.3)	32 (78.1)	6 (28.6)	<0.001	0.54	<0.001	<0.001
Hyperkeratosis	43 (95.6)	33 (58.9)	4 (12.1)	<0.001	<0.001	<0.001	<0.001	20 (76.9)	28 (68.3)	1 (4.8)	<0.001	1	<0.001	<0.001
Parakeratosis	22 (48.9)	4 (7.1)	0 (0.0)	<0.001	<0.001	<0.001	0.87	6 (23.1)	2 (4.9)	0 (0.0)	0.01	0.14	0.08	1
Spongiosis	21 (46.7)	53 (94.6)	31 (93.9)	<0.001	<0.001	<0.001	1	21 (80.8)	27 (65.9)	21 (100.0)	0.008	0.8	0.17	0.004
Inflamm. Infiltrate upper dermis	7 (15.6)	40 (71.4)	18 (54.6)	<0.001	<0.001	0.001	0.34	11 (42.3)	15 (36.6)	17 (81.0)	0.003	1	0.03	0.004
Inflamm. Infiltrate epidermis	25 (55.6)	45 (80.4)	20 (60.6)	0.02	0.03	1	0.15	23 (88.5)	35 (85.4)	21 (100.0)	0.19	1	0.73	0.26
Intraepidermal macro‐vesicles	1 (2.2)	13 (23.2)	21 (63.6)	<0.001	0.008	<0.001	<0.001	7 (26.9)	20 (48.8)	11 (52.4)	0.13	0.37	0.39	1
Intraepidermal micro‐vesicles	26 (57.8)	40 (71.4)	32 (97.0)	<0.001	0.62	<0.001	0.01	24 (92.3)	38 (92.7)	20 (95.2)	0.91	1	1	1
Elongated rete ridges	27 (60.0)	26 (46.4)	4 (12.1)	<0.001	0.69	<0.001	0.003	16 (61.5)	18 (43.9)	8 (38.1)	0.22	0.64	0.44	1
Dilated vessels	32 (71.1)	51 (91.1)	24 (72.7)	0.02	0.05	1	0.1	17 (65.4)	32 (78.1)	16 (76.2)	0.5	0.82	1	1
Inflamm. Infiltrate dermis	12 (26.7)	6 (10.7)	2 (6.1)	0.02	0.19	0.1	1	20 (76.9)	27 (65.9)	18 (85.7)	0.22	1	1	0.41
Munro micro abscess	23 (51.1)	0 (0.0)	0 (0.0)	<0.001	<0.001	<0.001	1	3 (30.8)	0 (0.0)	0 (0.0)	<0.001	<0.001	0.02	1
Vesicles with detached cells	2 (4.4)	3 (5.4)	19 (57.6)	<0.001	1	<0.001	<0.001	3 (11.5)	1 (2.6)	10 (47.6)	<0.001	0.14	0.04	<0.001

CD, contact dermatitis; E, eczema; Pso, psoriasis, numbers in brackets represent their respective percentage value.

3.2. Multivariate logistic regression

The multivariate logistic regression analysis demonstrated the significance of these morphologic features in distinguishing between eczema, psoriasis, and contact dermatitis lesions.

3.2.1. Psoriasis

In LC‐OCT imaging of psoriasis lesions was strongly associated with hyperkeratosis (OR 43.37 [6.58–534.88]; p < 0.001) and absence of inflammatory infiltrate in the upper dermis (OR 0.02 [0.003–0.11]; p < 0.001) or spongiosis (OR 0.03 [0.004–0.16]; p < 0.001), while in OCT imaging of psoriasis lesions was associated with parakeratosis (OR 29.06 [3.78–347.41]; p = 0.002) and also hyperkeratosis (OR 7.29 [2.13–29.58]; p = 0.002), aligning with the well‐established histopathologic features of psoriatic plaques.

3.2.2. Atopic eczema

The logistic regression model for eczema revealed a strong association in LC‐OCT imaging with spongiosis (OR 33.23 [6.59–255.25]; p < 0.001) and inflammatory infiltrate in the upper dermis (OR 18.17 [5.83–71.48]; p < 0.001). These features, characteristic of eczematous lesions, contributed significantly to the predictive model. For OCT imaging regression analysis showed highly significant results for the absence of vesicles with detached cells (OR 0.01 [0.0006–0.012]; p < 0.001). The heterogeneity observed in eczema lesions underscored the need for a nuanced imaging approach, further highlighting the diagnostic utility of the dual‐imaging modality.

3.2.3. Contact dermatitis

The multivariate logistic regression analysis for contact dermatitis demonstrated the presence of vesicles with detached cells (LC‐OCT: OR 18.73 [5.29–83.12]; p < 0.001; OCT: OR 13.42 [2.49–109.63]; p = 0.005) and the absence of hyperkeratosis to be statistically significant (LC‐OCT: OR 0.06 [0.01–0.19]; p < 0.001; OCT: OR 0.02 [0.0006–0.14]; p = 0.002). See also Tables 2 and 3.

TABLE 2.

Multivariate analysis of LC‐OCT imaging criteria.

LC‐OCT	OR (95% CI)	p‐value
Psoriasis
Inflamm. infiltrate upper dermis	0.02 (0.003–0.11)	<0.001
Spongiosis	0.03 (0.004–0.16)	<0.001
Hyperkeratosis	42.37 (6.58–534.88)	<0.001
Inflamm. infiltrate dermis	11.85 (1.97–116.37)	0.01
Intraepidermal macro vesicles	0.03 (0.001–0.38)	0.02
Eczema
Inflamm. infiltrate upper dermis	18.17 (5.83–71.48)	<0.001
Vesicles with detached cells	0.03 (0.005–0.12)	<0.001
Spongiosis	33.23 (6.59–255.25)	<0.001
Inflamm. infiltrate dermis	0.11 (0.02–0.49)	0.005
Acanthosis	0.18 (0.05–0.57)	0.006
Inflamm. infiltrate epidermis	6.51 (1.88–28.00)	0.006
Contact dermatitis
Hyperkeratosis	0.06 (0.01–0.19)	<0.001
Vesicles with detached cells	18.73 (5.29–83.12)	<0.001

TABLE 3.

Multivariate analysis of OCT imaging criteria.

OCT	OR (95% CI)	p‐value
Psoriasis
Parakeratosis	29.06 (3.87–347.41)	0.002
Hyperkeratosis	7.29 (2.13–29.58)	0.002
Erosion	0.06 (0.005–0.31)	0.004
Intraepidermal macro vesicles	0.18 (0.04–0.62)	0.10
Eczema
Vesicles with detached cells	0.01 (0.0006–0.12)	<0.001
Intraepidermal macro vesicles	4.69 (1.57–15.60)	0.008
Erosion	5.39 (1.56–22.79)	0.01
Spongiosis	0.25 (0.06–0.88)	0.03
Contact dermatitis
Hyperkeratosis	0.02 (0.0006–0.14)	0.002
Subepidermal microvesicles	14.92 (2.88–118.84)	0.003
Vesicles with detached cells	13.42 (2.49–109.63)	0.005

3.3. Decision tree algorithm

The decision tree model, rooted in the Gini's coefficient calculations, provided a dynamic framework for classifying lesions based on key imaging attributes. The algorithm identified specific morphologic criteria at each node, supporting a hierarchical categorization into eczema, psoriasis, or contact dermatitis (Figure 2).

FIGURE 2.

LC‐OCT (A) and OCT (B) Decision Tree Algorithm. Pso = psoriasis, E = eczema, CD = contact dermatitis.

3.3.1. Root node

Epidermal thickness (= hyperkeratosis) emerged as the primary discriminator, reflecting the overarching importance of this feature in distinguishing between the three inflammatory conditions.

3.3.2. Terminal nodes

The decision tree algorithm culminated in terminal nodes that corresponded to specific diagnoses. These terminal nodes were characterized by a combination of features such as spongiosis for eczema, elongation of rete ridges for psoriasis, and a diverse set of features for contact dermatitis.

3.4. Overall performance

The dual‐imaging approach, investigating OCT and LC‐OCT, exhibited a high overall accuracy in classifying lesions into eczema, psoriasis, and contact dermatitis categories. Sensitivity and specificity values further underscored the robustness of each imaging modality in correctly identifying the inflammatory skin condition.

4. DISCUSSION

The integration of Optical coherence tomography (OCT) and Line‐field confocal optical coherence tomography (LC‐OCT) in our study has confirmed the previous reported spectrum of morphologic features to distinguish eczema, psoriasis, and contact dermatitis lesions. ¹² , ¹³ , ¹⁴ , ¹⁷ , ¹⁸ Our findings therefore not only contribute to the growing body of knowledge in dermatologic imaging but also hold significant implications for advancing diagnostic precision and therapeutic strategies in inflammatory skin diseases.

The distinct morphologic features identified through descriptive statistics align with established histopathologic characteristics of atopic eczema, psoriasis, and contact dermatitis. ²⁰ Eczema lesions, characterized by spongiosis and increased epidermal thickness, corroborate with the known features of this inflammatory condition, emphasizing the reliability of the dual‐imaging approach in capturing these subtle changes. ²¹ Psoriasis lesions, typified by acanthosis and elongation of rete ridges, echo the well‐documented hallmarks of psoriatic plaques, reinforcing the diagnostic accuracy afforded by our imaging modality. ²² Interestingly, in our multivariate analysis elongated rete ridge did not stand out significantly comparing psoriasis versus eczema or contact dermatitis, while inflammatory cells in the epidermis which we expected being more commonly in dermatitis lesions showed a significant occurrence in psoriasis vulgaris. Our understanding is that inflammatory dermatoses can show very heterogeneous and diverse pathologies and can indeed often present elongated rete ridges and inflammatory cells in all manifestations, probably also depending on the severity of the disease. Conceivably, by looking more specific at the amplitude of elongated rete ridges or the amount or distributional pattern of inflammatory cells, the criteria would have stood out more specific.

The heterogeneity observed in contact dermatitis lesions, ranging from vesiculation to variable architectural changes, also reflects the diverse inflammatory responses associated with this condition. ²³ Nevertheless, our study underscores the capacity of OCT and LC‐OCT to capture these nuanced morphologic details, providing a comprehensive depiction of inflammatory skin diseases.

The decision tree algorithm furthermore provides a dynamic framework for lesion classification, which can give a hint to a probability of the correct diagnosis. The terminal nodes, characterized by specific combinations of features, offer a practical guide for clinicians, enhancing the interpretability of imaging results and being supportive for accurate diagnosis. We have to emphasize, that these features should not be seen as knock‐out criteria and even when a specific feature is not present, the diagnosis cannot be totally withdrawn. Due to the heterogenous presentation of inflammatory diseases, this approach should consider all logistic regression criteria.

Since the evaluation and interpretation of OCT and LC‐OCT images occurred totally independently, the juxtaposition of OCT and LC‐OCT has illuminated the nuanced capabilities of each imaging modality. LC‐OCT, with its superior resolution, emerges as a powerful tool for unraveling microarchitectural changes, offering a more detailed view of morphologic features compared to OCT. The heightened ability of LC‐OCT to capture fine structural nuances becomes particularly relevant in the examination of inflammatory dermatoses, where pathology predominantly manifests in the upper layers of the skin. The prominence of LC‐OCT in showcasing these microstructural alterations may account for the identification of more significant features, contributing to its enhanced diagnostic potential in the context of inflammatory skin conditions. However, it is crucial to acknowledge that the strengths of OCT persist in visualizing deeper processes, exemplified by its efficacy in discriminating skin tumors or deeper inflammatory processes. This duality of strengths between OCT and LC‐OCT underscores the importance of a tailored approach in dermatologic imaging, capitalizing on the unique attributes of each modality to comprehensively characterize skin pathologies at various depths. A correlation map comparing morphologic criteria from both imaging techniques is shown in Figure 3. (Further detailed correlation maps of each individual imaging technique are provided in Supplementary Material S1 and S2).

FIGURE 3.

Direct juxtaposition of morphologic criteria using OCT and LC‐OCT imaging, values representing Pearson's correlation coefficients.

Nevertheless, it is also imperative to acknowledge certain limitations in our study. Not only due to the physical limited penetration depth of LC‐OCT but also due to thickened epidermal plaques, crusting or inhomogeneous contact of the handheld probe with skin surface, the assessment of deeper morphologic features can be challenging. In such cases OCT imaging with higher penetration and is less susceptible to errors compared to LC‐OCT imaging.

Future longitudinal studies are needed to validate the stability of the identified imaging features and their potential variations during the course of inflammatory skin diseases. ²⁴ Additionally, expanding the dataset to include a broader spectrum of inflammatory skin conditions would augment the generalizability of our findings, fostering a more comprehensive understanding of the diagnostic potential of both imaging methods.

5. CONCLUSION

In conclusion, our study harnesses the power of OCT and LC‐OCT imaging to delineate the morphologic intricacies of inflammatory dermatoses, aligning with established histopathologic characteristics. The identified predictors in regression analysis serve as robust indicators for distinguishing between atopic eczema, psoriasis, and contact dermatitis. Additionally, the integration of a decision tree algorithm supports the understanding of the distinct features characterizing each inflammatory skin condition. Therefore, a holistic assessment offers a practical and hierarchical framework for lesion classification, enhancing the interpretability of imaging results for clinicians.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest.

ETHICS APPROVAL

Approved by the local ethics committee (Ludwig‐Maximilians University Munich; Project‐Nr. 17‐699).

CONSENT STATEMENT

The authors obtained written consent from patients for their photographs and medical information to be published in print and online and with the understanding that this information may be publicly available. Patient consent forms were not provided to the journal but are retained by the authors.

Supporting information

Supporting Information

Supporting Information

Deußing M, Ruini C, Nutz M, Kerl‐French K, Hartmann D, French LE, et al. Illuminating characteristic patterns of inflammatory dermatoses: A comprehensive dual‐imaging approach using Optical coherence tomography and Line‐field confocal optical coherence tomography. Skin Res Technol. 2024;30:e13833. 10.1111/srt.13833

DATA AVAILABILITY STATEMENT

The data underlying this article will be shared on reasonable request to the corresponding author.