Deep learning-based software is developed to assist physicians in terms of diagnosis; however, its clinical application is still under investigation. We integrated deep-learning-based software for diabetic retinopathy (DR) grading into the clinical workflow of an endocrinology department where endocrinologists grade for retinal images and evaluated the influence of its implementation. A total of 1432 images from 716 patients and 1400 images from 700 patients were collected before and after implementation, respectively. Using the grading by ophthalmologists as the reference standard, the sensitivity, specificity, and area under the receiver operating characteristic curve (AUC) to detect referable DR (RDR) were 0.91 (0.87–0.96), 0.90 (0.87–0.92), and 0.90 (0.87–0.93) at the image level; and 0.91 (0.81–0.97), 0.84 (0.80–0.87), and 0.87 (0.83–0.91) at the patient level. The monthly RDR rate dropped from 55.1% to 43.0% after implementation. The monthly percentage of finishing grading within the allotted time increased from 66.8% to 77.6%. There was a wide range of agreement values between the software and endocrinologists after implementation (kappa values of 0.17–0.65). In conclusion, we observed the clinical influence of deep-learning-based software on graders without the retinal subspecialty. However, the validation using images from local datasets is recommended before clinical implementation.
To investigate the effect and complications of Combined Endoscope assisted Procedures (CEaP): endoscopic cyclophotocoagulation and pars plana ablation (ECP-plus), along with endoscopic panretinal photocoagulation (PRP). The study design is a retrospective and noncomparative interventional case series from a tertiary referral center in Taiwan. Patients experiencing vessel growth at the iris and anterior chamber angle, along with an IOP > 21 mmHg were included. Twenty-five eyes from 23 patients were included over a 24-month period. After the procedures, all of them had a lower IOP value than their preoperative value. The mean IOP was 38.2± 7.1 mm Hg preoperatively, and 10.2± 4.7 mmHg (1 day), 13.8± 4.6 mmHg (1 week), 15.0± 5.3 mmHg (2 weeks), 17.4± 4.7 mmHg (1 month), 16.6± 4.1 mmHg (3 months), 16.0± 5.0 mmHg (6 months), and 15.7± 5.5 mmHg (12 months) postoperatively. At the 6th and 12th months, the IOP stabilized rate was 84% and 75%, respectively. Complications in the initial postoperative period (< 3 months) included uveitis (24%), and hyphema (16%), which were both resolved in the early postoperative period. Complications beyond 6 months included hypotony and phthisis bulbi in two patients (8%) in our study. There was no subject who suffered from retinal detachment, endophthalmitis or any other severe complications. The results of this study show that CEaP offers positive results in IOP lowering and NV regression. Additionally, CEaP is a complete treatment for NVG in controlling IOP and NV growth. The IOP lowering effects can be sustained upon completion of the treatment.
Optical coherence tomography angiography (OCTA) is a popular medical imaging technology that can quickly establish a three-dimensional model of the fundus without dye injection. However the number of images in a model is quite large, so finding the lesions through image processing technology can greatly reduce the time required for the judgment of the condition. This paper proposes a method for finding choroidal neovascularization (CNV) in OCTA images. Among the several characteristics of CNV, the larger turning angle of blood vessels is a relatively clear feature, so we will use this property to find out whether there is CNV in an OCTA image. We will transform the color space to CIELAB space, and extract the L-channel prior to preceding to the next step. We will then use some image segmentation methods to find the clearer vessel region. Finally, we will detect the CNV through certain morphology methods. The experimental result shows that our proposed method can effectively find the CNV in the OCTA image, meaning that we can make automated judgments through this method in the future and reduce the time necessary for human judgment.
Purpose. To investigate effects and complications of endoscopic vitrectomy combined with 3D heads‐up viewing system in treating traumatic ocular injury. Patients and Methods. This is a retrospective interventional case series in a tertiary referral center in Taiwan, and we included patients of traumatic ocular injury, and they underwent endoscopic vitrectomy combined with a 3D heads‐up viewing system. Results. Fourteen eyes of traumatic globe injury from 14 patients were studied over a 30‐month period. Preoperative VA ranged from no light perception (NLP) to 6/6. Postoperative visual acuity improved in 11 of the 14 eyes (79%). Until 6 months after surgery, all eyes had attached retina. The median logMAR BCVA was 2.4 at the first visit and 1.19 at the last visit (p = 0.0028). No subject suffered from retinal detachment, endophthalmitis, or other severe complications. Conclusions. Vitrectomy using endoscopy combined with 3D heads‐up viewing system allowed early evaluation and intervention in traumatic ocular injuries. Most of our cases showed both anatomical and visual acuity improvements.