Introduction

Unilateral congenital palsy of the superior oblique muscle (SOM) is the most common cause of vertical strabismus (incidence in children: 2.7 to 3.4 per 100,000 children per year)1. In the absence of a confirmation test, the diagnosis of unilateral congenital SOM palsy is most often presumed based on a range of clinical arguments.

Advances in magnetic resonance imaging (MRI) have broadened our understanding of SOM palsy, which is now considered as a congenital cranial dysinnervation disorder corresponding to abnormal cranial nerve development (i.e. hypoplastic or aplastic) causing muscle dysinnervation2. Several studies support the diagnostic usefulness of orbital MRI to visualize trochlear nerve absence and SOM atrophy hypoplasia3,4,5,6. Trochlear nerve status is clinically decisive, once congenital palsy is clinically diagnosed, treatment, whether surgical or non-surgical, can proceed regardless of muscle size. In a routine clinical setting, imaging of the cisternal portion of trochlear nerve by MRI can be challenging7 and therefore, non-visualization of the trochlear nerve does not necessarily indicate the absence of the nerve or its atrophy. Technical limitations, including the small size of the trochlear nerve as well as inter-reader variability in nerve identification, can affect the reliability of its assessment.

In contrast, imaging of the oculomotor muscles is easily performed. Quantitative assessment of the SOM can provide valuable complementary information, particularly in patients with an unvisualized trochlear nerve or with asymmetrical trochlear nerves, by revealing the consequences of nerve impairment on muscle trophicity in addition to the clinical orthoptic evaluation.

By considering oculomotricity as a whole through combined clinical, neural, and muscular assessments, orbital muscle evaluation on MRI assumes its full and complementary role within this integrated diagnostic framework. A qualitative and quantitative analysis of the muscle therefore appears pertinent to establish a comprehensive baseline profile before therapeutic decisions are made. Little is known concerning the optimal procedure for assessing SOM atrophy, either based on a qualitative or quantitative approach. Regarding the quantitative approach, there is a need to define optimal quantitative criteria to diagnose SOM atrophy on orbital MRI, which could effectively distinguish abnormal SOM in patients with unilateral congenital superior oblique palsy from controls8.

This study aimed to (1) compare the qualitative and quantitative approaches and (2) provide an optimal MRI-based quantitative strategy to diagnose SOM atrophy in patients with unilateral congenital SOM palsy.

We describe the Eye-Surface-Area (ESA) method which refers to a three-step diagnostic workflow designed to identify SOM atrophy on MRI. It includes: (1) a qualitative visual assessment, (2) a surface measurement based on manual contouring of the muscle cross-section, and (3) an area calculation using the muscle’s orthogonal diameters. This study provides the first validation of a composite MRI-based diagnostic threshold, with reproducibility confirmed across both ophthalmologic and radiologic evaluations.

Methods

Study design

This retrospective controlled study conducted at the University Hospital of Tours, France, spanned between January 2005 and June 2019. Patients diagnosed with unilateral congenital superior oblique palsy who underwent orbital MRI were included, without any age limit. The control population was paired by age and sex. Patients who performed an MRI including orbital sequences were included if they were visual symptom-free except for decreased visual acuity. For cases, exclusion criteria encompassed a history of head trauma or brain disease, prior strabismus or cranial surgery, hypertropia in the opposite eye during ipsilateral gaze or in the Bielschowsky head-tilt test (suggestive of bilateral superior oblique palsy). For controls, the exclusion criteria were as follows: history of ophthalmic disease, oculomotor palsy, severe myopia, thyroid-associated orbitopathy.

For a secondary analysis, to account for potential trophic differences between children and adults, we divided patients and controls into two groups (≤ 18 years and > 18 years), provided descriptive data, and performed simple comparison analyses using the Student’s t-test (paired for intra-subject eyes’ comparison, and unpaired for inter-subject comparison.

Ethics

This study was performed in accordance with the ethical standards of the Declaration of Helsinki and its amendments. All data were collected and processed anonymously. Due to the retrospective nature of the study, the patients were not required to provide informed consent in agreement with the French Law. This research was approved by the Ethics Committee of the French national Society of Ophthalmology that approved it (IRB: 00008855). All experiments were performed in accordance with the relevant guidelines and regulations.

Diagnostic criteria for unilateral congenital superior oblique palsy

The diagnosis of unilateral congenital superior oblique palsy relied on the presence of some or all of the following criteria: vertical, oblique, or torsional binocular diplopia; hypertropia in the primary position of the affected eye, increasing in contralateral gaze (over-elevation in adduction); positive Bielschowsky head-tilt test, deemed significant if the hypertropia in the affected eye was ≥ 5 prism diopters when the head was tilted on the ipsilateral side compared to the contralateral side; limitation of depression in down-gaze on the affected eye; head tilt to the contralateral side; face turn towards the contralateral shoulder or chin down9.

The presumption of the congenital origin of superior oblique palsy was made when there was a documented history or photographic evidence of long-standing strabismus, anomalous head posture since infancy or facial asymmetry.

Imaging protocol

Patients underwent a 1.5 or 3 Tesla (T) MRI, using an 8- or 20-channel head coil. MRI examinations were performed while asking the subjects’ to keep their eyes closed. Subjects were instructed to mentally focus on an imaginary point to ensure that both eyes were immobile and oriented the most straight ahead despite strabismus. No specific additional device was used, and no oculodynamic MRI was performed. The MRI examination was conducted using a head coil to obtain high-resolution images10,11. The 1.5 T scanners included the Aera (Siemens Healthineers, Erlangen, Germany) and the Signa HDx (GE Healthcare, USA), while the 3 T system was the Verio (Siemens Healthineers, Erlangen, Germany). The orbital MRI protocols across the three scanners, while not identical, were carefully harmonized to ensure comparable image quality (see detailed imaging parameters, Supplemental Material 1). All scanners underwent regular vendor calibration as part of routine clinical quality control. Most patients and controls were scanned using 1.5T MRI systems (total n = 91; Aera: n = 57, Signa HDx: n = 34), while 3T MRI was used in 29 cases (24%). The MRI protocol included at least thin coronal T2-weighted imaging (WI) covering the orbital surface (slice thickness = 2.5–3 mm). Orbital coronal T1-WI and high-resolution-3D-T2-gradient-echo-WI were usually performed for patients. The optimized high-resolution 3D T2-weighted imaging (T2-WI) protocol, used primarily for accurate measurements and multiplanar reconstructions, is as follows:

  • Aera 1.5T: axial CISS sequence (Constructive Interference In Steady State): duration: 3’20; slice thickness 3D isodimensional 0.6 mm, Repetition Time 6.21ms, Echo Time 3.11ms; Flip Angle 62; matrix 512 × 448; Echo Train Length: 1.

  • Signa HDx 1.5T: axial FIESTA sequence (Fast Imaging Employing Steady state Acquisition): duration: 3’10, slice thickness 3D isodimensional 0.6 mm, interslice gap 0.3 mm, Repetition Time 5.3 ms, Echo Time 2.14 ms; Flip Angle 65; matrix 288 × 256; Echo Train Length: 1.

  • Verio 3T: axial CISS sequence: duration: 3’10; slice thickness 3D isodimensional 0.7 mm, Repetition Time 6.37ms, Echo Time 2.73ms; Flip Angle 49; matrix 320 × 307; Echo Train Length: 1.

Image analysis

To enhance reproducibility, the multiplanar reconstruction tool was used to run through each muscle with longitudinal and perpendicular reconstructions, instead of measurements made on a classical coronal plane, which can lead to an incorrect perception of the muscle section slices (Fig. 1). Each muscle was reconstructed using a 3D tool to obtain a standardized and reproducible cross-sectional slice at its thickest region. We studied the SOM and IR muscles for each eye, blindly to palsy status but not the diagnosis. IR muscle measurements were considered as the control data, assuming that IR morphology is not affected by SOM palsy. According to the ophthalmological observation, each eye was defined as paralyzed (P) or non-paralyzed (NP). All radiological data were collected blindly to the palsy status of each patient. For both quantitative and qualitative assessments, double independent reading was performed for each patient with SOM palsy by an experienced ophthalmologist (MJL) and an experienced neuroradiologist (CC), and independent reading by an experienced radiologist was performed for control subjects (AM). All images were analyzed using the Carestream-Philips workstation. For discordant measurements in the patient group (i.e. ≥ 3 mm2 for area and ≥ 5 mm2 for surface) on inferior rectus (IR) muscle readings between both readers, a third reading was performed by a radiologist (BD), and the measurements from this third reading were used for the final analysis.

Fig. 1
figure 1

Magnetic resonance imaging-based assessment of the surface and area of the superior oblique muscle (SOM). Imaging sequence: High-resolution-3D-T2-gradient-echo CISS sequence centered on the SOM. Left column: native view, not used for study measurements. Right column: multiplanar reconstructions - perpendicular (axial view) and longitudinal (sagittal view) along the muscle axis (arrows) using an oblique three-dimensional tool, representing the measurement methodology applied in the study. The coronal muscle section slice is the plane of interest for SOM analysis. SOM analysis row: demonstrates measurement differences between native and reconstructed SOM analysis. Surface measurement: performed using the “irregular region of interest” tool. Area calculation: defined as width (larger axis of the muscle section) multiplied by length (shortest axis).

Full size image

Qualitative assessment

The observer had to visually estimate the presence of SOM atrophy as (1) if present or (0) if absent and indicate the side of the atrophy.

Quantitative assessment

Surface and area were assessed for the quantitative analysis (Fig. 1):

  • Surface measurement: by manually contouring the muscle section with the “irregular region of interest” tool (in mm2).

  • Area calculation: width measurement defined as the larger axis of the muscle section and length measurement defined as the shortest axis; area was then calculated as follows: area = width x length (in mm²).

Ratios were calculated for patients and controls. IR muscle measurements were considered as the control data, assuming that IR morphology is not affected by SOM palsy. Acknowledging that coexisting oculomotor abnormalities could theoretically alter IR size, calculating right-to-left IR ratio was intended to ensure that IR morphology was not significantly influenced by SOM palsy, and allow comprehensive comparison with normative data from literature.

  • For patients with SOM palsy, the side of paralyzed (P) SOM was given by the ophthalmological examination.

    • Ratio between surface (s) of SOM paralyzed and non-paralyzed muscle (sSOMP/sSOMNP), and ratio between IR muscle of paralyzed and non-paralyzed side were calculated (sIRP/sIRNP).

    • Identical ratios were calculated for areas (a): aSOMP/aSOMNP and aIRP/aIRNP.

  • For control subjects, identical ratios were calculated, using right/left (R/L) lateralization, giving four ratios, for surface and area:

    • surface ratio right and left SOM.

    • surface ratio right and left IR.

    • area ratio right and left SOM.

    • area ratio right and left IR.

The most sensitive and specific threshold for the diagnosis of SOM palsy using the MRI-based quantitative method was investigated.

Statistical analysis

Statistical analyses were conducted to compare qualitative and quantitative assessments of superior oblique muscle atrophy. Continuous variables are presented as medians with interquartile ranges (IQRs). Variability in measurements depending on the MRI magnetic field strength and imaging protocol (2D versus 3D) was analyzed using the Wilcoxon rank-sum test, and the Kruskal–Wallis test for variables with more than two categories (e.g., slice thickness and MRI scanner type). Comparisons between independent groups were performed using unpaired Student’s t-tests for normally distributed data and Mann-Whitney U tests for non-parametric variables. The normality of the data was assessed using the Shapiro-Wilk test. Categorical variables were analyzed using Chi-square or Fisher’s exact tests as appropriate. Inter- and intra-observer reliability was evaluated using Cohen’s kappa coefficients for categorical assessments and intraclass correlation coefficients (ICCs) for quantitative measurements. Diagnostic performance of quantitative ratios was assessed using Receiver Operating Characteristic (ROC) curve analysis, with optimal thresholds determined by the Youden index to maximize sensitivity and specificity. Correlations between continuous variables were examined using Spearman’s rank correlation coefficient. Statistical significance was set at p < 0.05 for all tests. Analyses were performed using the R software, version 4.1.1. For the final quantitative analysis, data from the most experienced neuroradiologist (CC) was analyzed.

Results

A total of 120 subjects were included in this study (median age: 12 years; IQR1-3: 5–42 years; males: 73 (61%)). Baseline characteristics are summarized in Table 1. Sixty patients with unilateral congenital SOM palsy were age and sex matched with 60 controls (respectively p = 0.286 and p = 0.854). No significant variability due to magnetic field strength, protocol, MRI scanner, or slice thickness was observed (see Supplemental material 2).

Table 1 Baseline characteristics.
Full size table

Inter-rater reliability

Inter-rater agreement was good to excellent for both quantitative measurements:

  • Surfaces SOM paralyzed/SOM non paralyzed = 0.86 (CI95% [0.78–0.92]).

  • Areas SOM paralyzed/SOM non paralyzed = 0.86 (CI95% [0.77–0.91]).

  • Surfaces SOM paralyzed/IR paralyzed = 1.0 (CI95% [1–1]).

  • Areas SOM paralyzed/IR paralyzed = 0.79 (CI95% [0.68–0.87]).

  • Surfaces IR paralyzed/IR non paralyzed = 0.66 (CI95% [0.49–0.78]).

  • Areas IR paralyzed/IR non paralyzed = 0.65 (CI95% [0.48–0.78]).

Qualitative assessment was also found to yield good reproducibility. Kappa coefficient for right and left SOM palsy was 0.75 (95% CI [0.48–1.00]) and 0.84 (95% CI [0.64–1.00]), respectively (Supplemental material 3).

Qualitative results for superior oblique muscle atrophy

Qualitative SOM atrophy was found for 28 patients (46.7%) by operator 1 (MJL) and 29 patients (48.3%) for operator 2 (CC). No cases of qualitative SOM atrophy were reported in the controls.

Quantitative results for superior oblique muscle atrophy

In the 60 patients, median SOM surface and area were significantly lower on the paralyzed side compared to the contralateral eye:

  • Median surface SOM paralyzed vs. non-paralyzed: 8.4 mm² vs. 13.5 mm² (p < 0.001).

  • Median area SOM paralyzed vs. non-paralyzed: 9 mm² vs. 16.0 mm² (p < 0.001).

No other differences were found for median surfaces and areas concerning the other muscles in both patient and control groups (Table 2).

Table 2 Magnetic resonance imaging-based quantitative assessment of oculomotor muscles.
Full size table

Threshold for the diagnosis of superior oblique muscle palsy

Using the Youden index, the best thresholds for diagnosing SOM palsy based on surface and area were established as follows (Table 3):

Table 3 Best thresholds for the diagnosis of superior oblique palsy according to superior oblique muscle surface and area SOM: superior oblique muscle.
Full size table

  • Surface SOM paralyzed = 9.4 mm², AUC = 0.74, CI95%[0.65–0.83], sensitivity = 0.60, specificity = 0.80, PPV = 0.75.

  • Area SOM paralyzed = 11.9 mm², AUC = 0.73, CI95%0.64–0.82], sensitivity = 0.65, specificity = 0.82, PPV = 0.78.

  • Surfaces SOM paralyzed/SOM not paralyzed = 0.90, AUC = 0.85, CI95%[0.77–0.92], sensitivity = 0.73, specificity = 0.88.

  • Areas SOM paralyzed/SOM not paralyzed = 0.80, AUC 0.80 CI95%[0.72–0.90], sensitivity = 0.63, specificity = 0.95.

The best threshold for diagnosing SOM palsy based on ratios of SOM surface and area between patients and controls are presented in Table 4. The composite ratio of SOM surface < 0.9 or area < 0.8 showed the best compromise between sensitivity and specificity. With this composite ratio, 47/60 (78.3%) and 8/60 (13.3%) were diagnosed in the patient and control groups, respectively. A case of quantitative measurement-based SOM atrophy, with uncertain qualitative measurement-based SOM atrophy, is presented in supplemental material 4.

Table 4 Best thresholds for the diagnosis of superior oblique palsy according to SOM ratios between patients and controls, based on surface, area and composite criteria of surface OR area.
Full size table

Muscle parameters according to age

Supplemental material 5 summarizes the mean (± standard deviation) muscle parameters based on age group (i.e. children vs. adults) for both patients and controls, including thickness and width for both SOM and IR muscle. Notably, the mean SOM surface was 7.19 ± 3.6 mm2 in the paralyzed eye for children and 11.95 ± 5.1 mm2 for adults (p = 0.496), and 10.89 ± 3.0 mm2 in the non-paralyzed eye for children and 15.91 ± 3.2 mm2 for adults (p < 0.001).

Discussion

The detection of SOM atrophy on MRI can be refined through a combined qualitative and quantitative approach. The quantitative method is reproducible between a neuroradiologist and an ophthalmologist, provided that appropriate reconstructions are used. A composite quantitative threshold can be defined, with a surface ratio of 0.9 or an area ratio of 0.8 (paralyzed vs. non-paralyzed side), correctly diagnosing 78.3% of cases, whereas qualitative analysis alone proves inferior, identifying only 46.7 to 48.3% of cases. Based on these results, we describe the ESA method consisting of three steps, (1) qualitative assessment, (2) surface measurement, and (3) area measurement, as a reliable method for detecting SOM atrophy.

Qualitative assessment of superior oblique muscle atrophy has a lower diagnostic yield

Our findings suggest that qualitative assessment of SOM atrophy has a lower diagnostic yield compared to quantitative approaches. While previous studies have explored quantitative strategies, this is the first to evaluate an eye-based assessment of SOM atrophy. In one out of two cases, the qualitative approach is sufficient to identify SOM atrophy. Visual qualitative assessment of SOM atrophy had a good correlation with the quantitative evaluation. Based on our findings, we suggest that, to minimize missed cases, the search for SOM atrophy should begin with a qualitative assessment, followed by the standardized quantitative approach described in this study if no atrophy is detected.

A combined threshold for quantitative assessment of superior oblique muscle atrophy

Over the past three decades, MRI-based understanding of SOM palsy has significantly advanced12,13,14,15,16,17,18,19,20. Numerous studies have examined MRI findings in SOM palsy, particularly regarding trochlear nerve absence and SOM atrophy. SOM atrophy on MRI is now well-established as a hallmark of unilateral congenital SOM palsy when compared to the non-paretic SOM, although the degree of atrophy can vary. While early studies focused on SOM volume12,13, cross-sectional area measurements offer a simpler and more practical approach. However, the MRI-based definition of SOM atrophy remains inconsistent. Establishing a reliable imaging biomarker would be of great value for diagnosing patients with hypertropia, given the condition’s variable clinical presentation. To achieve this, a biomarker capable of effectively distinguishing SOM atrophy in patients with vertical strabismus from controls is needed. A reliable biomarker could enhance diagnosis and facilitate further research by clarifying clinical heterogeneity.

Analyzing the trochlear nerve may be challenging as this requires high-resolution, thin-slice 3D-MRI sequences centered on the mesencephalo-pontine junction. Furthermore, with the exception of a few studies, no large-scale investigation of trochlear nerve visibility using 1.5T or 3T MRI has been conducted, and its assessment may require an experienced neuroradiologist7. As such, analysis of the trochlear nerve was not considered a sufficiently strong parameter to include in our database. Additionally, Yang et al., 2015, demonstrated that the ratio of superior oblique muscle area and volume between the paretic and normal sides exhibits excellent predictive accuracy for diagnosing trochlear nerve absence in congenital superior oblique palsy8. Although current decision-making primarily relies on clinical findings and nerve visualization, muscular information obtained through imaging may, in the future, play a complementary role in refining the therapeutic strategy. Considering two patients with similar clinical findings and an unvisualized trochlear nerve, a surgical rather than a medical treatment decision might be more easily justified in the patient whose ESA results demonstrate SOM atrophy. Our current results are not yet sufficient to directly influence the therapeutic strategy bu they nevertheless represent an essential preliminary step toward that goal.

The composite ratio of SOM surface < 0.9 or area < 0.8 between the paretic and non-paretic sides achieved the optimal balance between sensitivity and specificity, correctly diagnosing 8 out of 10 patients in our cohort including both adults and children with unilateral congenital SOM palsy and yielding a 10% false positive rate in the control group. Yang et al.8 showed that the ratio of SOM surface, when ≤0.75, was associated with diagnosis of trochlear nerve absence. However, our work differs from Yang et al.8 who also measured surface of SOM cross-section, but on a strict orbital coronal plane, not strictly coronal to each oculomotor muscle plane. This may lead to less reproducible measurements, away from anatomic reality. In our study, the high inter-reader intraclass coefficients for surface measurements compared to area measurements suggest that the surface delineation tool may be more suitable for evaluating orbital muscles than simple 2D width and length values. Therefore, the ESA method is a standardized, stepwise approach that combines qualitative assessment, manual surface measurement, and geometric area calculation. In contrast to prior studies that require high-resolution volumetric reconstructions or specialized neuroimaging protocols, the ESA method can be applied using routine clinical MRI sequences and basic radiological tools. Additionally, we define practical diagnostic thresholds (surface ratio < 0.9 and area ratio < 0.8), which offer reproducible criteria to support the diagnosis of congenital SOM palsy, even by experienced ophthalmologists without neuroradiological training.

Both surface and area measurements were considered, even though they essentially reflect the same morphological muscle property. Calculating area based on width and thickness measurements allowed us to align our methodology with previous literature on orbital muscle assessment, which frequently reports two-dimensional radial measurements, whereas surface evaluation is less commonly used21,22,23,24. We considered that focusing on a single cross-sectional plane of the muscle would provide a reasonable reflection of muscle’s trophicity. However, this approach has inherent limitation as it does account for the entire muscle volume, the selection of the thickest slice may vary between readers, and the superior oblique muscle itself is elongated and fusiform. Three-dimensional volumetric measurement or multi-slice integration would undoubtedly offer a more accurate estimation of true muscle volume, but this was not feasible in our cohort since some patients had only two-dimensional MRI sequences. Although such an approach would be promising, volumetric segmentation would be more time-consuming and would likely benefit from automated or semi-automated processing methods.

To the best of our knowledge, this is the first study to propose a muscle-axis SOM analysis and compare inter-rater reproducibility for both qualitative and quantitative approaches. Reproducibility was found to be high between both readers, suggesting that in a clinical setting, experienced ophthalmologists perform as well as experienced radiologists, provided that the MRI resolution is appropriate. These findings highlight the potential for ophthalmologists to contribute to MRI interpretation in clinical settings, reducing dependence on radiologists for specific cases. By providing a structured, reproducible method, the ESA method can facilitate earlier diagnosis of SOM palsy in routine practice. It may reduce the number of missed cases, especially in ambiguous clinical presentations, and enable ophthalmologists or trained technicians to perform reliable assessments. However, further studies with more readers are needed to confirm these results.

Focusing on adults’ measurements, the mean IR muscle thickness of 4.5 mm is consistent with values reported in the literature (ranging from 3.2 to 5.1 mm)21,22,23,24, supporting the representativeness of our study population. The mean right-to-left IR surface ratio was 0.99 in both patients and controls, confirming the robustness of the measurements and suggesting that IR morphology was not significantly influenced by SOM palsy. It is worth noting that the purpose of the IR measurement and the derived SO/IR ratio was primarily to ensure intra- and inter-subject comparability and to validate the consistency of our cohort with published data, rather than to propose a clinically applicable ratio.

Although this study is monocentric and retrospective, its originality lies in a previously unproposed measurement method, optimized to obtain a strict slice section for each muscle, with very good inter-rater agreement. A strict coronal view analysis does not account for the anatomical structure of orbital muscles, which do not have a strictly anteroposterior orientation, potentially leading to less reproducible measurements. There remains room for improvement in future studies by controlling for eye movements during MRI, although eye-tracking technology is not yet routinely available.

Concerning data heterogeneity due to different MRI scanners, we emphasize that patients could undergo orbital MRI on any of the available scanners while maintaining similarly high image quality, thereby avoiding any loss of diagnostic opportunity related to scanner type. Differences in magnetic field strength (1.5T vs. 3T) may affect image resolution, signal-to-noise ratio, and contrast. While 3T MRI theoretically provides higher spatial resolution and improved soft tissue contrast, potentially enhancing the visualization of small structures such as the superior oblique muscle, we did not observe inter-rater differences related to magnetic field strength, slice thickness, or MRI scanner. Additionally, variations in MRI protocols and slice thickness were controlled by using the IR muscle as a reference structure, which demonstrated no variability across scans. The high inter-rater reproducibility further supports that measurements were robust despite minor technical variations. We acknowledge that restricting analyses to a single scanner would have been more appropriate for approaches relying heavily on imaging parameters beyond visual perception, such as radiomics. However, for our purpose focusing on the analysis of clearly visible anatomical structures, the combination of data from different scanners was considered appropriate. MRI is more resource-intensive than a standard clinical exam; however, our method was specifically designed to be reproducible, applicable to routine MRI sequences, and interpretable by both experienced radiologists and trained ophthalmologists.

Although controls were age and sex-matched, no orthoptic examination was performed. However, controls did not report any diplopia, suggesting the absence of patent strabismus, but latent strabismus could not be ruled out.

Even though the ESA method was used in our cohort on muscle ratios accounting for age-related differences and compared to age and gender-matched controls, pediatric and adult-onset of unilateral SOM palsy may vary in their clinical presentations. Mean SOM and IR muscle measurements were indeed higher in adults, in line with physiological data on orbital muscles trophicity24,25,26. We intentionally did not conduct a stratified analysis by age group, as the ESA methodology aims to provide an intra-subject ratio that can be universally used, regardless of age or gender.

The ESA method offers a practical, reproducible, and accessible approach for the assessment of SOM atrophy with a potential to serve as a standardized diagnostic tool, improving the early and reliable detection of congenital SOM palsy. The data presented here are not yet sufficient to directly inform therapeutic decision-making, as larger and multicentric cohorts will be required to strengthen the evidence. However, in patients with vertical strabismus who do not meet all the diagnostic criteria for SOM palsy, radiological confirmation remains important to better characterize these patients and provide a reliable diagnosis. Beyond conventional clinical and trochlear nerve assessment, orbital muscle imaging data deserve to be analyzed in relation to post-therapeutic outcomes. Although current decision-making primarily relies on clinical findings and nerve visualization, muscular information obtained through imaging may, in the future, play a complementary role in refining the therapeutic strategy.

Conclusion

For patients with unilateral congenital SOM palsy, SOM atrophy assessed by MRI is a valuable additional tool in the diagnostic process, as a complement to trochlear nerve visualization. We recommend a simple and innovative measurement methodology that analyses along the muscle axis using multiplanar reconstruction and the surface delineation tool. The current findings lead to the conclusion that discriminative thresholds to conclude to SOM atrophy are SOM surface < 0.9 or area < 0.8. The Eye-Surface-Area (ESA) method involves three key steps: (1) qualitative assessment, (2) surface measurement, and (3) area measurement. This structured approach reliably identifies SOM atrophy and is suitable for use by both radiologists and ophthalmologists. However, there is still room for improvement as this ratio missed 20% of cases and yielded a 10% false positive rate in controls.