Application of deep learning image assessment software VeriSee™ for diabetic retinopathy screening.

PURPOSE: To develop a deep learning image assessment software VeriSee™ and to validate its accuracy in grading the severity of diabetic retinopathy (DR). METHODS: Diabetic patients who underwent single-field, nonmydriatic, 45-degree color retinal fundus photography at National Taiwan University Hospital between July 2007 and June 2017 were retrospectively recruited. A total of 7524 judgeable color fundus images were collected and were graded for the severity of DR by ophthalmologists. Among these pictures, 5649 along with another 31,612 color fundus images from the EyePACS dataset were used for model training of VeriSee™. The other 1875 images were used for validation and were graded for the severity of DR by VeriSee™, ophthalmologists, and internal physicians. Area under the receiver operating characteristic curve (AUC) for VeriSee™, and the sensitivities and specificities for VeriSee™, ophthalmologists, and internal physicians in diagnosing DR were calculated. RESULTS: The AUCs for VeriSee™ in diagnosing any DR, referable DR and proliferative diabetic retinopathy (PDR) were 0.955, 0.955 and 0.984, respectively. VeriSee™ had better sensitivities in diagnosing any DR and PDR (92.2% and 90.9%, respectively) than internal physicians (64.3% and 20.6%, respectively) (P < 0.001 for both). VeriSee™ also had better sensitivities in diagnosing any DR and referable DR (92.2% and 89.2%, respectively) than ophthalmologists (86.9% and 71.1%, respectively) (P < 0.001 for both), while ophthalmologists had better specificities. CONCLUSION: VeriSee™ had good sensitivity and specificity in grading the severity of DR from color fundus images. It may offer clinical assistance to non-ophthalmologists in DR screening with nonmydriatic retinal fundus photography.

Full-thickness macular hole formation in proliferative diabetic retinopathy.

Twenty-one consecutive patients (21 eyes) having proliferative diabetic retinopathy (PDR) and fibrovascular proliferation (FVP) with optical coherence tomography (OCT) available before and after full-thickness macular hole (FTMH) formation were retrospectively reviewed. Four types of FTMH formation pathways in PDR were identified and were quite different from those in idiopathic conditions. The activity, severity and locations of FVP varied in PDR eyes destined to develop FTMHs. Type 1 was characterized by epiretinal membrane (ERM) and/or vitreomacular traction (VMT) inducing foveoschisis, intraretinal cysts or foveal detachment, followed by formation of a FTMH or macular hole retinal detachment (MHRD). In type 2, ERM and/or FVP induced lamellar macular hole (LMH) with foveoschisis, followed by the formation of FTMH or MHRD. Type 3 was characterized by the initial tractional retinal detachment (TRD) with foveal cysts and/or foveoschisis and the subsequent formation of MHRD. Type 4 was characterized by TRD associated with foveal thinning, ensued by the formation of MHRD. The severity of FVP was grade 2 in 66.7% of eyes in both types 1 and 4, and grade 3 in 75% of eyes in type 3 while the severity of FVP was more evenly distributed in type 2.

Vitrectomy with internal limiting membrane peeling versus nonsurgical treatment for diabetic macular edema with massive hard exudates.

PURPOSE: To compare the anatomical and functional outcomes of severe diabetic macular edema (DME) with massive hard exudates managed by pars plana vitrectomy (PPV) with internal limiting membrane (ILM) peeling or nonsurgical treatment. METHODS: We retrospectively reviewed 40 eyes with DME and massive hard exudates treated with either PPV with ILM peeling (vitrectomy group, 21 eyes) or nonsurgical treatment with anti-vascular endothelium growth factor (VEGF) and/or steroids (nonsurgical group, 19 eyes). Changes in best-corrected visual acuity (BCVA) and central retinal thickness (CRT) and resolution of macular hard exudates were compared between the two groups. RESULTS: After treatment, CRT decreased steadily in the vitrectomy group but fluctuated in the nonsurgical group. Compared with eyes in the nonsurgical group, eyes in the vitrectomy group had better visual improvement (P < 0.05 at 6 and 12 months and the final visit) and greater decrease in CRT (P < 0.05 at 3 and 6 months and the final visit) after adjustment for baseline BCVA. Hard exudates resolved more rapidly in the vitrectomy group than in the nonsurgical group, with 94.1% versus 47.4% eyes showing significant absorption after 6 months of the treatment (P = 0.003). In the vitrectomy group, 62% eyes did not require any further injections for treating DME after the operation. CONCLUSIONS: PPV with ILM peeling resulted in rapid resolution of hard exudates with significant anatomical and functional improvement in DME with massive hard exudates.

Using optical coherence tomography to assess the role of age and region in corneal epithelium and palisades of vogt.

Using spectral-domain optical coherence tomography (OCT) to observe the morphology and epithelial thickness (ET) of the palisades of Vogt (POV), and to evaluate the role of age and region on these structures.One hundred twelve eyes of 112 healthy subjects were enrolled and divided into 4 groups: A (0-19), B (20-39), C (40-59), and D (≥60 years old). RTvue-100 OCT was applied on the cornea and the limbus. The morphology of the subepithelial stroma underneath the epithelium of POV was classified into typical and atypical types. Maximum ET of POV was measured manually from OCT images.The positive rate of typical POV in superior, nasal, temporal, and inferior limbus was: Group A: 100%, 69.2%, 65.4%, 100%; Group B: 100%, 73.5%, 61.8%, 94.1%; Group C: 95.8%, 41.7%, 37.5%, 83.3%; Group D: 67.9%, 0%, 3.6%, 25%, showing a significant decreasing tendency with age. The maximum ET of POV in superior, nasal, temporal, and inferior limbus was: Group A: 103.5 ±â€Š10.1 um, 89.2 ±â€Š9.7 um, 87.9 ±â€Š13.6 um, 104.7 ±â€Š14.1 um; Group B: 111.4 ±â€Š15.8 um, 85.3 ±â€Š9.9 um, 88.2 ±â€Š8.6 um, 112.6 ±â€Š19.7 um; Group C: 116.4 ±â€Š16.4 um, 82.8 ±â€Š11.6 um, 87.0 ±â€Š11.6 um, 120.0 ±â€Š25.6 um; Group D: 96.3 ±â€Š17.9 um, 73.8 ±â€Š15.9 um, 79.2 ±â€Š16.7 um, 87.4 ±â€Š18.5 um. Age-dependent change was observed. In general, the maximum ET of POV in superior/inferior quadrants was thicker than the other 2 quadrants.Spectral-domain OCT is a useful tool to observe the limbal microstructure and provide invaluable information. Aging and anatomic regions had significant effects on the microstructure of these areas.