Development and Validation of a Deep Learning –Based System for Detection of AIDS- Related Cytomegalovirus Retinitis in Ultra-Wide-Field Fundus Images

Background: Ophthalmological screening for cytomegalovirus retinitis (CMVR) for HIV/AIDS patients is important. The application of a deep learning (DL) system to AIDS-related CMVR with ultra-wide-field (UWF) fundus images is promising, but the feasibility and efficiency of this method have not been studied. Methods: We independently developed and internally validated a DL system for identifying active CMVR, inactive CMVR and non-CMVR in 6960 UWF fundus images from 862 AIDS patients and validated the system in a prospective and an external validation data set using area under curve (AUC), accuracy, sensitivity, and specificity. A heat map identified the most important area (lesions) used by the DL system for differentiating CMVR. This trial is registered with ClinicalTrials.gov, number NCT04831333. Findings The DL system showed AUCs of 0·945 (95% confidence interval [CI]: 0·929, 0·962), 0·964 (95% CI: 0·870, 0·999) and 0·968 (95% CI: 0·860, 1·000) for detecting active CMVR from non-CMVR and 0·923 (95% CI: 0·908, 0·938), 0·902 (0·857, 0·948) and 0·884 (0·851, 0·917) for detecting active CMVR from non-CMVR in the internal cross-validation, external validation, and prospective validation, respectively. It also showed the ability to differentiate active CMVR from non-CMVR and inactive CMVR as well as to identify active CMVR and inactive CMVR from non-CMVR (all AUCs in the three independent data sets >0·900). The heat maps successfully highlighted lesion locations. Interpretation: Our system showed reliable performance for detecting AIDS-rated CMVR. DL technology is promising for screening and differentiating CMVR in AIDS patients. Clinical Trial: This trial is registered with ClinicalTrials.gov, number  NCT04831333. Funding: Scientific Research Project of Beijing Youan Hospital, CCMU, 2018 (YNKTQN20180201); Capital Health Research and Development of Special (2020-1-2052); Science & Technology Project of Beijing Municipal Science & Technology Commission (Z181100001818003); Beijing Municipal Administration of Hospitals’ Ascent Plan (DFL20150201). Declaration of Interest: The authors have no conflicts of interest to declare. Ethical Approval: This study was conducted in accordance with the Declaration of Helsinki. Both the Ethics Committee of Beijing YouAn Hospital (LL-2018-150-K) and the Ethics Committee of Beijing Tongren Hospital (TRECKY2018-056) approved the study. Written informed consent was obtained from each subject.