Artificial Intelligence for Evaluation of Retinal Vasculopathy in Facioscapulohumeral Dystrophy Using OCT Angiography: A Case Series

Abstract

Facioscapulohumeral muscular dystrophy (FSHD) is a slowly progressive muscular dystrophy with a wide range of manifestations including retinal vasculopathy. This study aimed to analyse retinal vascular involvement in FSHD patients using fundus photographs and optical coherence tomography-angiography (OCT-A) scans, evaluated through artificial intelligence (AI). Thirty-three patients with a diagnosis of FSHD (mean age 50.4 ± 17.4 years) were retrospectively evaluated and neurological and ophthalmological data were collected. Increased tortuosity of the retinal arteries was qualitatively observed in 77% of the included eyes. The tortuosity index (TI), vessel density (VD), and foveal avascular zone (FAZ) area were calculated by processing OCT-A images through AI. The TI of the superficial capillary plexus (SCP) was increased (p < 0.001), while the TI of the deep capillary plexus (DCP) was decreased in FSHD patients in comparison to controls (p = 0.05). VD scores for both the SCP and the DCP results increased in FSHD patients (p = 0.0001 and p = 0.0004, respectively). With increasing age, VD and the total number of vascular branches showed a decrease (p = 0.008 and p < 0.001, respectively) in the SCP. A moderate correlation between VD and EcoRI fragment length was identified as well (r = 0.35, p = 0.048). For the DCP, a decreased FAZ area was found in FSHD patients in comparison to controls (t (53) = −6.89, p = 0.01). A better understanding of retinal vasculopathy through OCT-A can support some hypotheses on the disease pathogenesis and provide quantitative parameters potentially useful as disease biomarkers. In addition, our study validated the application of a complex toolchain of AI using both ImageJ and Matlab to OCT-A angiograms.