We examined MRI EOM findings in a homogenous group of nine patients with CPEO caused by single mtDNA deletions compared with those in age- and gender-matched healthy controls. We describe differences in EOM radiological appearance, cross-sectional area, and T2 in CPEO patients compared to controls, and a significant negative correlation between EOM T2 and ROEM, suggesting MRI can provide functionally relevant measures of CPEO disease severity.
Earlier imaging studies have identified EOM involvement as characteristic of CPEO [4–7, 11]. However, these are generally in smaller, patient-only case series with genetically undetermined disease without quantitative assessment in all five EOMs. We have, for the first time, performed a systematic evaluation of EOM MRI signal changes in comparison with matched controls, and correlated MRI measurements of EOM size and condition with clinical functional measurements.
Previous reports of EOM atrophy in CPEO are mixed and rather limited: Carlow et al.  examined MR, LR and IR volume in eight clinically diagnosed CPEO patients showing an overall 41 % reduction vs. controls. In a smaller series, EOM atrophy was reported on CT imaging of four clinically diagnosed CPEO patients . Ortube et al.  detected isolated atrophy of SR and SO in five CPEO patients compared with controls. Finally, more recent work confirmed a primary myopathic rather than supranuclear pathway dysfunction aetiology for the limitation of eye movement in CPEO as evidenced by significant EOM atrophy in patients with single or multiple mtDNA deletions compared with controls . Our EOM cross-sectional area results, incorporating SR and SO muscles in addition to MR, LR and IR, were consistent with these findings.
EOM atrophy is reported in a number of other neuromuscular disorders. Cranial nerve palsies (III, IV or VI) cause atrophy which is restricted to the muscle innervated by the affected nerve, thereby making it straight-forward to distinguish the affected muscle radiologically . EOM atrophy is variably apparent in myasthenia gravis, but consistently seen in cases left untreated . Radiological atrophy of clinically affected muscles has also been found in congenital fibrosis of EOMs type 1  and type 2  (now known to be dysinnervation syndromes). Taken together, these studies suggest that EOM atrophy is a feature of late-stage muscle pathology irrespective of the underlying cause, comparable to the atrophy of skeletal limb muscles on MRI in both neurogenic and myopathic disorders .
While EOM cross-sectional area has previously been investigated in CPEO, correlations between muscle dimensions and other quantitative MRI measures, and clinical functional measurements, have not been reported. Such correlations, if observed, would support the validity of MRI measures as biologically relevant disease indices. Although we found no correlation between muscle atrophy and reduced ROEM or disease duration, there was a significant negative correlation between T2 and ROEM in the patients. This indicates that although measures relating to EOM volume may not be useful to quantify severity in patients with long-established disease, measures such as EOM T2, reflecting variations in muscle composition rather than size, may provide valid alternative functionally relevant markers. Future studies will be required to establish the sensitivity of EOM T2 to longitudinal change, and to establish the sensitivity of the measure to partial volume confounds due to surrounding intra-orbital fat.
In addition to atrophic changes, we also detected EOM MRI signal abnormalities in patients with CPEO, apparent qualitatively as an increased frequency of T1w hyperintense streaks with very occasional STIR hyperintensity. Central streaks of EOM hyperintensity on T1w MRI have previously been reported in CPEO , although we also saw similar appearances in some of the control EOMs. While the frequency of these hyperintensities was greater in the patient group, their presence in the healthy control images suggests such changes are not specific markers of CPEO pathology. Similar changes have also been shown to occur with muscle atrophy in patients with congenital fibrosis of the EOMs type 1 .
This is the first study to measure EOM T2 in a genetically homogenous CPEO cohort. While acute inflammation prolongs T2, and is used to quantify EOM inflammation in thyroid-related ophthalmopathy [15–19], inflammation is unlikely to be a pathological substrate of EOM T2 change in CPEO, since inflammatory infiltrates are usually absent in the muscle biopsies of patients with mitochondrial disease. This is supported by the predominant absence of EOM STIR hyperintensities in our patient images, and suggests increased intramuscular lipid as the source of T2 elevation since our T2 relaxometry acquisition was not fat-suppressed. A further analysis (results not shown) that attempted to fit a bi-exponential model to the multi-echo decay data was not sufficient to obtain putative lipid and muscle-water components with independent T2 values, and more sophisticated future measurements will be required to unambiguously establish the origin of the total muscle T2 elevation. A further caveat to interpretation of the T2 findings is the potential susceptibility of the measurements to contamination from the surrounding fatty substrate, despite our precautions in defining conservative ROI boundaries. Nevertheless, elevated patient EOM T2 compared with controls was consistent and, coupled with the correlation observed between EOM T2 and ROEM measures, should motivate future more sophisticated investigations.
Accurate, valid and reliable biomarkers of disease progression are lacking in mitochondrial disease but will be required to assess the efficacy of novel therapies in treatment trials. We have demonstrated the feasibility of quantitative MRI to provide measures reflecting disease severity in individual EOMs in patients with CPEO. While EOM atrophy is a clear radiological feature, its uniformity in patients and lack of correlation with ROEM suggests it might not provide a meaningful measure of progression in patients studied a significant time after disease onset. In contrast, EOM T2 prolongation was associated with cross-sectional variations in ROEM in our study group, and so may represent a useful biomarker for future clinical trials involving patients with similar disease durations. Future investigations will evaluate the added value of combining T2 measurement with the technically challenging but promising approach of EOM quantitative diffusion weighted imaging .
In conclusion, we have examined the EOM MRI appearance of patients with CPEO due to single mtDNA deletions and demonstrated atrophy of all EOMs, more frequent T1w hyperintensities in patients than controls and uniform prolongation of EOM T2 in patients. A correlation between EOM T2 and reduced ocular excursion facility suggests that this parameter might provide a valid, clinically relevant, objective measure of disease severity useful to monitor patients with established CPEO with possible value in future clinical trials.