Texture analysis of orbital magnetic resonance imaging for monitoring and predicting treatment response to glucocorticoids in patients with thyroid-associated ophthalmopathy

Abstract

Purpose: To evaluate the value of magnetic resonance imaging (MRI)-based texture analysis of extraocular muscle (EOM) and orbital fat (OF) in monitoring and predicting the response to glucocorticoid (GC) therapy in patients with thyroid-associated ophthalmopathy (TAO). Methods: Thirty-seven active and moderate-to-severe TAO patients (responders, n=23; unresponders, n=14) were retrospectively enrolled. MRI-based texture parameters (entropy, uniformity, skewness and kurtosis) of EOM and OF were measured before and after GC therapy, and compared between groups. Correlations between the changes of clinical activity score (CAS) and imaging parameters before and after treatment were assessed. Receiver operating characteristic curves were used to evaluate the predictive value of identified variables. Results: Responsive TAOs showed significantly decreased entropy and increased uniformity at EOM and OF after GC therapy (p<0.01), while unresponders showed no significance. Changes of entropy and uniformity at EOM and OF were significantly correlated with changes of CAS before and after treatment (p<0.05). Responders showed significantly lower entropy and higher uniformity at EOM than unresponders before treatment (p<0.01). Entropy and uniformity of EOM and disease duration were identified as independent predictors for responsive TAOs. Combination of all three variables demonstrated optimal efficiency (area under curve, 0.802) and sensitivity (82.6%), and disease duration alone demonstrated optimal specificity (100%) for predicting responsive TAOs. Conclusion: MRI-based texture analysis can reflect histopathological heterogeneity of orbital tissues. It could be useful for monitoring and predicting the response to GC in TAO patients.

T2‐Weighted MR Imaging‐Derived Radiomics for Pretreatment Determination of Therapeutic Response to Glucocorticoid in Patients With Thyroid‐Associated Ophthalmopathy: Comparison With Semiquantitative Evaluation

“…Components of inflammatory edema show a good response to systemic GC administration, while those of the fibrotic compounds do not 30 . According to previous suggestions, responsive TAOs commonly demonstrate more homogeneous edema within EOMs; therefore, they manifest a satisfactory therapeutic response to GC 13,14,32 . However, unresponsive TAOs usually have greater tissue complexity and more fibrous scarring within edema; hence, they may have a lower therapeutic response 13,14,32 .…”

“…In accordance with the detected results from the semiquantitative imaging model, original_firstorder_Minimum showed a positive relationship with the therapeutic response. Considering the prior knowledge that the responsive group may have more homogeneous edematous environment of EOMs resulting in smaller span of SI and less heterogeneity, 14,32 the inverse relationship between original_firstorder_Range, as well as squareroot_glszm_GrayLevelNonUniformity and wavelet.LLL_firstorder_Skewness, 33 and therapeutic response would also be reasonable. Through uniting both low‐level and high‐dimensional features derived from T 2 WI, it is verified that the utilization of radiomics could also facilitate the decision support regarding the pretreatment determination of therapeutic response in patients with TAO.…”

“…30 According to previous suggestions, responsive TAOs commonly demonstrate more homogeneous edema within EOMs; therefore, they manifest a satisfactory therapeutic response to GC. 13,14,32 However, unresponsive TAOs usually have greater tissue complexity and more fibrous scarring within edema; hence, they may have a lower therapeutic response. 13,14,32 Given a strong association between the decreased SI on T 2 WI and fibrotic component, we speculated that the obviously reduced SIR min of the unresponsive group found in this study may also be correlated with the increased proportion of focal fibrosis under the general inflammatory environment.…”

Predictive markers for anti-inflammatory treatment response in thyroid eye disease

“…( 59 ) simply chose decrease in CAS by at least 2 points, while Wang et al. ( 65 ) chose a compound criteria with decrease in CAS by at least 2 points accompanied with at least one of the other improvements (i.e., decrease of proptosis ≥ 2 mm, decrease of lid width ≥2 mm, etc.). Such difference in evaluation would make it hard to compare the exact predictive value of different biomarkers among studies.…”

“…By combining imaging biomarkers with clinical markers, Wang et al. ( 65 ) constructed a prediction model based on entropy of EOM, uniformity of EOM and disease duration (AUC=0.802). Another worth-noting point in this study is their employment of texture analysis in orbital MRI, which is a prospective and advanced post-processing method.…”