Deep learning algorithms have been used to detect diabetic retinopathy (DR) with specialist-level accuracy. This study aims to validate one such algorithm on a large-scale clinical population, and compare the algorithm performance with that of human graders. A total of 25,326 gradable retinal images of patients with diabetes from the community-based, nationwide screening program of DR in Thailand were analyzed for DR severity and referable diabetic macular edema (DME). Grades adjudicated by a panel of international retinal specialists served as the reference standard. Relative to human graders, for detecting referable DR (moderate NPDR or worse), the deep learning algorithm had significantly higher sensitivity (0.97 vs. 0.74, p < 0.001), and a slightly lower specificity (0.96 vs. 0.98, p < 0.001). Higher sensitivity of the algorithm was also observed for each of the categories of severe or worse NPDR, PDR, and DME ( p < 0.001 for all comparisons). The quadratic-weighted kappa for determination of DR severity levels by the algorithm and human graders was 0.85 and 0.78 respectively ( p < 0.001 for the difference). Across different severity levels of DR for determining referable disease, deep learning significantly reduced the false negative rate (by 23%) at the cost of slightly higher false positive rates (2%). Deep learning algorithms may serve as a valuable tool for DR screening.

BackgroundCataract surgery is the most common surgical procedure in ophthalmology. Biometry data and accurate intraocular lens (IOL) calculations are very important in achieving the desired refractive outcomes. The aim of this study was to compare measurements using a new optical low coherence reflectometry (OLCR) biometer (OA-2000) and the gold standard partial coherence interferometry (PCI) optical biometer (IOLMaster 500).MethodsOcular biometry of cataract patients were measured by the OA-2000 and IOLMaster 500 to compare keratometry (K), axial length (AL), anterior chamber depth (ACD), white-to-white (WTW) diameter, and IOL power using the SRK/T formula.ResultsOne hundred and two eyes of 68 cataract patients were evaluated with the two optical biometers. The mean values of the AL, K, ACD, and WTW differed very little (OCLR biometer, 23.12 mm, 44.50 diopters (D), 3.06, and 11.64 mm, respectively; PCI biometer, 23.18 mm, 44.6 D, 3.15, and 11.86 mm, respectively), but the differences were significant (all, p ≤ 0.05). The AL, K, and ACD showed excellent correlations (r = 0.999, 0.980, and 0.824, respectively; all p < 0.001); however, there was a weak correlation of the WTW diameter between the two devices (r = 0.256). The IOL powers using the SRK-T formula derived from both instruments were very similar, with an excellent correlation (r = 0.989). The mean difference between the two instruments was 0.32 D.ConclusionsThe OLCR biometer showed very a strong agreement with the standard PCI optical biometer for almost all ocular biometry measurements, except for the WTW diameter.Trial registrationTCTR20160614003; date 06/09/2016; ‘retrospectively registered’.

Aim: To assess the central corneal thickness (CCT) and endothelial cell loss after manual small-incision cataract surgery and phacoemulsification in patients with white cataract. Material and methods: This is a comparative, prospective, non-randomized study on 42 patients with white cataract, who underwent cataract surgery. The patients were divided into manual small-incision cataract surgery (21 eyes, MSICS group) and phacoemulsification cataract surgery group (21 eyes, phaco group). The endothelial cell density (ECD), central cornea thickness (CCT), and corrected distance visual acuity (CDVA) were evaluated at 1 day, 1 week, 4 weeks, and 3 months postoperatively. The results of 20 cases of nuclear sclerosis grade II-III (LOCS III) who underwent phacoemulsification by the same surgeon were also compared. Propensity scoring was used to adjust for confounding by selection bias. Results: The CCT increased after surgery in both groups. The thickness was greater in the phaco group on first day postoperatively (73 µ increase in MSICS group and 138 µ in phaco group, p=0.008) and it returned to preoperative levels 1 month postoperatively. The endothelial cell loss was lower in the MSICS group at 3 months postoperatively (11.8% in MSICS group and 15.8% in phaco group, p=0.111). The CDVA was not different in both groups at 1 week and 4 weeks postoperatively (p>0.05). Conclusions: Manual small-incision cataract surgery for white cataract provided less central corneal thickness changes compared to conventional phacoemulsification. Abbreviations: CCT = central corneal thickness; ECD = endothelial cell density; CDVA = corrected distance visual acuity; APT = absolute phacoemulsification time; EPT = effective phacoemulsification time; MSICS = Manual small-incision cataract surgery in white cataract; Phaco II = Phacoemulsification in white cataract; Phaco I = phacoemulsification in NS 2 + Cataract; Phaco = Phacoemulsification in white cataract; APACRS = Asia-Pacific Association of Cataract and Refractive Surgeons

PurposeTo evaluate the effectiveness of subconjunctival anesthesia as compared to retrobulbar anesthesia for pain control during manual small-incision cataract surgery (MSICS) performed by third-year residents.DesignA randomized, controlled trial.Patients and methodsA total of 150 patients undergoing routine cataract surgery were randomly assigned to receive either subconjunctival anesthesia (group 1, n = 75) or retrobulbar anesthesia (group 2, n = 75). Third-year residents performed MSICS using the modified Blumenthal technique. Subconjunctival anesthesia was administered by injecting 2% xylocaine with adrenalin into the superior conjunctiva, and retrobulbar anesthesia by injecting 2 mL of 2% xylocaine with adrenalin into the retrobulbar space. We studied the following variables: intraoperative pain score rated on a 100-point visual analog scale (VAS), operative time, and injection and operative complications.ResultsA mean age of 69 vs 70 years, an operative time of 47.1 (SD, 9.9) min vs 47.7 (10.9) min, and a median (interquartile range) pain score of 40 (range, 20–70) vs 40 (range, 20–50) were observed in the subconjunctival and the retrobulbar groups, respectively. The injection complication of subconjunctival hemorrhage was significantly higher in the subconjunctival group (25.3%) compared to the retrobulbar group (1.3%). The operative complication rate between groups was not different (P > 0.05).ConclusionBoth, superior subconjunctival anesthesia and retrobulbar anesthesia were effective during MSICS when used in a residency training program.

No abstract