Search

This study presents a fully automated image informatics framework. The framework is combined with a deep learning (DL) approach to automatically predict visual acuity outcomes for people undergoing surgery for idiopathic full-thickness macular holes using 3D spectral-domain optical coherence tomography (SD-OCT) images. To overcome the impact of high variation in real-world image quality on the robustness of DL models, comprehensive imaging data pre-processing, quality assurance, and anomaly detection procedures were utilised. We then implemented, trained, and tested nine state-of-the-art DL predictive models through our designed loss function with multiple 2D input channels on the imaging dataset. Finally, we quantitatively compared the models using four evaluation metrics. Overall, the predictive model achieved a MAE of 6.47 ETDRS letters score, demonstrating high predictability. This confirms that our fully automated approach with input from seven central SD-OCT images from each patient can robustly predict visual acuity measurements. Further research will focus on adapting 3D DL-based predictive models and the uncertainty of 2D and 3D DL-based predictive models.

Abstract Globally, 43 million people are living with HIV, 90% in developing countries. Increasing life expectancy with combination antiretroviral therapy (cART) results in chronic complications, including HIV-associated neurocognitive disorders (HAND) and eye diseases. HAND screening is currently challenging. Our aim was to evaluate clinical utility of retinopathy as a screening measure of HAND in older cART-treated individuals in Tanzania and feasibility of smartphone-based retinal screening in this low-resource setting. A cross-sectional systematic sample aged ≥ 50-years attending routine HIV follow-up in Tanzania were comprehensively assessed for HAND by American Academy of Neurology criteria and received ophthalmic assessment including smartphone-based retinal imaging. HAND and ophthalmic assessments were independent and blinded. Diagnostic accuracy was evaluated by AUROC curves. Of 129 individuals assessed, 69.8% were visually impaired. Thirteen had retinopathy. HAND prevalence was 66.7%. Retinopathy was significantly associated with HAND but HIV-disease factors (CD4, viral load) were not. Diagnostic accuracy of retinopathy for HAND was poor (AUROC 0.545-0.617) but specificity and positive predictive value were high. We conclude that ocular pathology and HAND appear highly prevalent in this low-resource setting. Although retinal screening cannot be used alone identify HAND, prioritization of individuals with abnormal retinal screening is a potential strategy in low-resource settings.