ABSTRACT
Neovascular age-related macular degeneration (nAMD) can result in blindness if left untreated, and patients often require repeated anti-vascular endothelial growth factor injections. Although, the treat-and-extend method is becoming popular to reduce vision loss attributed to recurrence, it may pose a risk of overtreatment. This study aimed to develop a deep learning model based on DenseNet201 to predict nAMD recurrence within 3 months after confirming dry-up 1 month following three loading injections in treatment-naïve patients. A dataset of 1076 spectral domain optical coherence tomography (OCT) images from 269 patients diagnosed with nAMD was used. The performance of the model was compared with that of 6 ophthalmologists, using 100 randomly selected samples. The DenseNet201-based model achieved 53.0% accuracy in predicting nAMD recurrence using a single pre-injection image and 60.2% accuracy after viewing all the images immediately after the 1st, 2nd, and 3rd injections. The model outperformed experienced ophthalmologists, with an average accuracy of 52.17% using a single pre-injection image and 53.3% after examining four images before and after three loading injections. In conclusion, the artificial intelligence model demonstrated a promising ability to predict nAMD recurrence using OCT images and outperformed experienced ophthalmologists. These findings suggest that deep learning models can assist in nAMD recurrence prediction, thus improving patient outcomes and optimizing treatment strategies.
Subject(s)
Macular Degeneration , Wet Macular Degeneration , Humans , Tomography, Optical Coherence/methods , Artificial Intelligence , Retrospective Studies , Neural Networks, Computer , Macular Degeneration/diagnostic imaging , Intravitreal Injections , Angiogenesis Inhibitors/therapeutic use , Wet Macular Degeneration/diagnostic imaging , Wet Macular Degeneration/drug therapy , RanibizumabABSTRACT
This study investigated the changes in the ganglion cell complex (GCC) of patients with acute central serous chorioretinopathy (CSC) following focal laser photocoagulation (FLP) and sought to determine its correlation with visual acuity (VA). Our retrospective study was conducted on 30 patients diagnosed with acute CSC between January 2015 and April 2022, who underwent FLP within 3 months of symptom onset. The study assessed GCC changes by measuring the thickness of its inner retinal layers-retinal nerve fiber layer (RNFL), ganglion cell layer (GCL), and inner plexiform layer (IPL) using optical coherence tomography (OCT). GCC thickness was compared between the affected and unaffected eyes and a healthy control group. VA was also assessed at baseline and at 1, 3, and 6 months post-treatment. VA showed significant improvement from 0.20 ± 0.14 at baseline to 0.10 ± 0.12 logMAR at 6 months post-treatment (p = 0.003). There was no significant change in GCC thickness over the 6-month period. No significant differences in GCC thickness were observed when comparing CSC eyes with fellow eyes or with normal controls at any time point. Final VA was significantly related only to baseline VA, with no correlation found with other factors, including RNFL, GCL, and IPL thickness. In summary, for patients with acute CSC undergoing FLP, our findings indicate that there is no significant change in GCC thickness detectable by OCT before and after the resolution of subretinal fluid (SRF), despite improvements in VA post-laser treatment. This suggests that any potential impact of FLP on GCC thickness may be minimal and not discernible with the current measurement methods, such as OCT, emphasizing that VA improvements may be primarily associated with alterations in the outer retina rather than the inner retina. Further studies with extended follow-up durations are warranted to evaluate any potential long-term changes in GCC.
ABSTRACT
In this retrospective longitudinal cohort study, we investigated the temporal changes in the peripapillary retinal nerve fiber layer (pRNFL) and inner retinal thickness in patients with acute central serous chorioretinopathy (CSC) using spectral-domain optical coherence tomography (SD-OCT). We followed up with these patients for 6 months, and during this period, the thickness of the pRNFL and the ganglion cell complex (GCC) in CSC patients were compared with the eyes of normal healthy individuals. The study also examined the correlation between the pRNFL thickness, GCC thickness, and visual acuity. The research sample consisted of 67 patients (43 male and 24 female) with an average age of 49.72 ± 9.87 years. The initial findings showed no significant differences in the pRNFL and GCC thickness between the study and fellow eye, study and normal healthy eyes, and fellow and normal healthy eyes. There was no significant difference in the pRNFL and GCC thickness when comparing the study eye with the fellow eye for 6 months. In the study eye, no significant difference was observed when comparing the initial GCC and pRNFL thickness with those at 1, 3, and 6 months. Visual acuity improved significantly from 0.18 ± 0.23 logMAR to 0.04 ± 0.06 logMAR (p < 0.001). The GCC and pRNFL thickness did not significantly affect visual acuity. In conclusion, acute CSC patients did not show significant changes in the pRNFL and inner retinal thickness, suggesting that the GCC and pRNFL do not substantially influence the short-term visual prognosis in these patients.
Subject(s)
Central Serous Chorioretinopathy , Humans , Female , Male , Adult , Middle Aged , Central Serous Chorioretinopathy/diagnostic imaging , Tomography, Optical Coherence , Longitudinal Studies , Retrospective Studies , Retina/diagnostic imaging , Acute DiseaseABSTRACT
BACKGROUND: Although previous research has made substantial progress in developing high-performance artificial intelligence (AI)-based computer-aided diagnosis (AI-CAD) systems in various medical domains, little attention has been paid to developing and evaluating AI-CAD system in ophthalmology, particularly for diagnosing retinal diseases using optical coherence tomography (OCT) images. OBJECTIVE: This diagnostic study aimed to determine the usefulness of a proposed AI-CAD system in assisting ophthalmologists with the diagnosis of central serous chorioretinopathy (CSC), which is known to be difficult to diagnose, using OCT images. METHODS: For the training and evaluation of the proposed deep learning model, 1693 OCT images were collected and annotated. The data set included 929 and 764 cases of acute and chronic CSC, respectively. In total, 66 ophthalmologists (2 groups: 36 retina and 30 nonretina specialists) participated in the observer performance test. To evaluate the deep learning algorithm used in the proposed AI-CAD system, the training, validation, and test sets were split in an 8:1:1 ratio. Further, 100 randomly sampled OCT images from the test set were used for the observer performance test, and the participants were instructed to select a CSC subtype for each of these images. Each image was provided under different conditions: (1) without AI assistance, (2) with AI assistance with a probability score, and (3) with AI assistance with a probability score and visual evidence heatmap. The sensitivity, specificity, and area under the receiver operating characteristic curve were used to measure the diagnostic performance of the model and ophthalmologists. RESULTS: The proposed system achieved a high detection performance (99% of the area under the curve) for CSC, outperforming the 66 ophthalmologists who participated in the observer performance test. In both groups, ophthalmologists with the support of AI assistance with a probability score and visual evidence heatmap achieved the highest mean diagnostic performance compared with that of those subjected to other conditions (without AI assistance or with AI assistance with a probability score). Nonretina specialists achieved expert-level diagnostic performance with the support of the proposed AI-CAD system. CONCLUSIONS: Our proposed AI-CAD system improved the diagnosis of CSC by ophthalmologists, which may support decision-making regarding retinal disease detection and alleviate the workload of ophthalmologists.
Subject(s)
Central Serous Chorioretinopathy , Diagnosis, Computer-Assisted , Humans , Algorithms , Artificial Intelligence , Central Serous Chorioretinopathy/diagnostic imaging , Computers , Deep LearningABSTRACT
This study investigated the long-term effects of epiretinal membrane (ERM) surgery on peripapillary retinal nerve fiber layer (RNFL) thickness using optical coherence tomography (OCT) images. We included 30 patients with idiopathic ERM who underwent a vitrectomy for ERM removal with internal limiting membrane peeling. The patients were followed up for 5 years after surgery, and their medical records were reviewed for best-corrected visual acuity (BCVA) and OCT parameters. The study population comprised 24 females (80.0%), and the mean age was 65.4 ± 7.2 years. The baseline BCVA significantly improved from 0.28 ± 0.24 to 0.12 ± 0.09 logMAR (p < 0.001) 1 year after surgery and continued to improve for 5 years after surgery. The peripapillary RNFL thickness initially increased after surgery and then gradually decreased. The peripapillary RNFL thicknesses of the global and temporal sectors showed significant reductions 2 years after surgery, whereas those of the nasal sectors did not significantly change. The peripapillary RNFL thickness was thinner in the global and temporal areas of the operated eyes than in those of the fellow eyes 4 and 5 years after surgery. In conclusion, peripapillary RNFL thicknesses decreased in the global and temporal areas after ERM surgery, whereas peripapillary RNFL thicknesses in the nasal sectors did not change significantly during the long-term follow-up.
ABSTRACT
Myopic choroidal neovascularization (mCNV) is a common cause of vision loss in patients with pathological myopia. However, predicting the visual prognosis of patients with mCNV remains challenging. This study aimed to develop an artificial intelligence (AI) model to predict visual acuity (VA) in patients with mCNV. This study included 279 patients with mCNV at baseline; patient data were collected, including optical coherence tomography (OCT) images, VA, and demographic information. Two models were developed: one comprising horizontal/vertical OCT images (H/V cuts) and the second comprising 25 volume scan images. The coefficient of determination (R2) and root mean square error (RMSE) were computed to evaluate the performance of the trained network. The models achieved high performance in predicting VA after 1 (R2 = 0.911, RMSE = 0.151), 2 (R2 = 0.894, RMSE = 0.254), and 3 (R2 = 0.891, RMSE = 0.227) years. Using multiple-volume scanning, OCT images enhanced the performance of the models relative to using only H/V cuts. This study proposes AI models to predict VA in patients with mCNV. The models achieved high performance by incorporating the baseline VA, OCT images, and post-injection data. This model could assist in predicting the visual prognosis and evaluating treatment outcomes in patients with mCNV undergoing intravitreal anti-vascular endothelial growth factor therapy.
ABSTRACT
Acupuncture is recognized as a component of alternative medicine and is increasingly used worldwide. Many studies have shown the various effects of acupuncture around the eyes for ophthalmologic or nonophthalmologic conditions. For ophthalmologic conditions, the effect of acupuncture on dry eye syndrome, glaucoma, myopia, amblyopia, ophthalmoplegia, allergic rhinoconjunctivitis, blepharospasm, and blepharoptosis has been reported. Recently, several studies on dry eye syndrome have been reported and are in the spotlight. However, given the variety of study designs and reported outcomes of periocular acupuncture, research is still inconclusive, and further studies are required. In addition, although a systematic and reliable safety assessment is required, to the best of our knowledge, there have been no reports of a literature review of ocular complications resulting from periocular acupuncture. This review collected cases of ocular injury as severe adverse events from previously published case reports of periocular acupuncture. A total of 14 case reports (15 eyes of 14 patients) of adverse events published between 1982 and 2020 were identified. This review article provides a summary of the reported cases and suggestions for the prevention and management of better visual function prognosis.
Subject(s)
Acupuncture Therapy , Amblyopia , Blepharoptosis , Dry Eye Syndromes , Glaucoma , Humans , Acupuncture Therapy/adverse effects , Acupuncture Therapy/methods , Glaucoma/etiology , Dry Eye Syndromes/etiology , Blepharoptosis/etiologyABSTRACT
This study reported the short-term effects of intravitreal brolucizumab (IVB) on peripapillary retinal nerve fiber layer (RNFL) thickness in patients with neovascular age-related macular degeneration (nAMD). This retrospective observational case series included patients with nAMD treated with other anti-vascular endothelial growth factor (anti-VEGF) agents and subsequently switched to IVB because of poor response to those other agents on spectral domain optical coherence tomography (SD-OCT). Best-corrected visual acuity (BCVA), intraocular pressure, funduscopy, and SD-OCT were assessed at baseline, 2 weeks, 1 month, and 3 months after injection. Twenty-two patients were included in the study. In the IVB group, BCVA significantly improved 3 months after injection compared with baseline (0.45 ± 0.25 vs. 0.38 ± 0.25, p = 0.012). During the 3-month follow-up, compared with baseline, RNFL thicknesses of the global, superior temporal, inferior temporal, inferior nasal, nasal, and superior nasal sectors did not change substantially in the IVB group. However, temporal RNFL thickness significantly decreased at 1 month (p = 0.045), and the significance was lost at 3 months (p = 0.378). The central macular thickness of treated eyes significantly decreased compared with the baseline at every follow-up visit. IVB in patients with nAMD had morphological and functional visual gain effects without RNFL thinning during the short-term follow-up.
Subject(s)
Macular Degeneration , Ranibizumab , Humans , Angiogenesis Inhibitors , Retrospective Studies , Vascular Endothelial Growth Factor A/pharmacology , Retinal Ganglion Cells , Intravitreal Injections , Macular Degeneration/drug therapy , Nerve Fibers , Tomography, Optical CoherenceABSTRACT
Importance: It remains unclear whether comorbidities in patients with retinal artery occlusion (RAO), a rare retinal vascular disorder, differ by subtype and whether mortality is higher. Objective: To examine the nationwide incidence of clinically diagnosed, nonarteritic RAO, causes of death, and mortality rate in patients with RAO compared with that in the general population in Korea. Design, Setting, and Participants: This retrospective, population-based cohort study examined National Health Insurance Service claims data from 2002 to 2018. The population of South Korea was 49â¯705â¯663, according to the 2015 census. Data were analyzed from February 9, 2021, to July 30, 2022. Main Outcomes and Measures: The nationwide incidence of any RAO, including central RAO (CRAO; International Statistical Classification of Diseases and Related Health Problems, Tenth Revision [ICD-10] code, H34.1) and noncentral RAO (other RAO; ICD-10 code, H34.2) was estimated using National Health Insurance Service claims data from 2002 to 2018, with 2002 to 2004 as the washout period. Furthermore, the causes of death were evaluated and the standardized mortality ratio was estimated. The primary outcomes were the incidence of RAO per 100â¯000 person-years and the standardized mortality ratio (SMR). Results: A total of 51â¯326 patients with RAO were identified (28â¯857 [56.2%] men; mean [SD] age at index date: 63.6 [14.1] years). The nationwide incidence of any RAO was 7.38 (95% CI, 7.32-7.44) per 100â¯000 person-years. The incidence rate of noncentral RAO was 5.12 (95% CI, 5.07-5.18), more than twice that of CRAO (2.25 [95% CI, 2.22-2.29]). Mortality was higher in patients with any RAO than in the general population (SMR, 7.33 [95% CI, 7.15-7.50]). The SMR for CRAO (9.95 [95% CI, 9.61-10.29]) and for noncentral RAO (5.97 [95% CI, 5.78-6.16]) showed a tendency toward a gradual decrease with increasing age. The top 3 causes of death in patients with RAO were diseases of the circulatory system (28.8%), neoplasms (25.1%), and diseases of the respiratory system (10.2%). Conclusions and Relevance: This cohort study found that the incidence rate of noncentral RAO was higher than that of CRAO, whereas SMR was higher for CRAO than noncentral RAO. Patients with RAO show higher mortality than the general population, with circulatory system disease as the leading cause of death. These findings suggest that it is necessary to investigate the risk of cardiovascular or cerebrovascular disease in patients newly diagnosed with RAO.
Subject(s)
Retinal Artery Occlusion , Male , Humans , Adolescent , Female , Retrospective Studies , Cohort Studies , Incidence , Retinal Artery Occlusion/epidemiology , Retinal Artery Occlusion/etiology , Republic of Korea/epidemiologyABSTRACT
The purpose of this study was to investigate the long-term changes in the choroidal thickness in central retinal vein occlusion (CRVO) following anti-vascular endothelial growth factor (VEGF) treatment. This retrospective study included 41 eyes from 41 patients with treatment-naïve unilateral CRVO. We compared the best-corrected visual acuity (BCVA), subfoveal choroidal thickness (SFCT), and central macular thickness (CMT) of CRVO eyes with those of fellow eyes at baseline, 12 months, and 24 months. Baseline SFCT was significantly higher in CRVO eyes than in fellow eyes (p < 0.001); however, there was no significant difference in the SFCT between CRVO eyes and fellow eyes at 12 months and 24 months. When compared with baseline SFCT, SFCT significantly decreased at 12 months and 24 months in CRVO eyes (all p < 0.001). In patients with unilateral CRVO, SFCT in the CRVO eye was significantly thicker than in the fellow eye at baseline, and after 12 months and 24 months, there was no difference from the fellow eye.
Subject(s)
Retinal Vein Occlusion , Humans , Retinal Vein Occlusion/diagnostic imaging , Retrospective Studies , Choroid/diagnostic imaging , PatientsABSTRACT
Neovascular age-related macular degeneration (nAMD) and central serous chorioretinopathy (CSC) are two of the most common macular diseases. This study proposes a convolutional neural network (CNN)-based deep learning model for classifying the subtypes of nAMD (polypoidal choroidal vasculopathy, retinal angiomatous proliferation, and typical nAMD) and CSC (chronic CSC and acute CSC) and healthy individuals using single spectral-domain optical coherence tomography (SD-OCT) images. The proposed model was trained and tested using 6063 SD-OCT images from 521 patients and 47 healthy participants. We used three well-known CNN architectures (VGG-16, VGG-19, and ResNet) and two customized classification layers. Additionally, transfer learning and mix-up-based data augmentation were applied to improve robustness and accuracy. Our model demonstrated high accuracies of 99.7% and 91.1% in the nAMD and CSC classification and retinopathy (nAMD and CSC) subtype classification, including normal participants, respectively. Furthermore, we performed an external test to compare the classification accuracy with that of eight ophthalmologists, and our model showed the highest accuracy. The region determined to be important for classification by the model was confirmed using gradient-weighted class activation mapping. The model's clinical criteria were similar to that of the ophthalmologists.
ABSTRACT
This study aimed to investigate changes in choroidal thickness after pars plana vitrectomy (PPV) with and without air tamponade in patients with idiopathic epiretinal membrane (ERM). We retrospectively reviewed 61 patients with ERM who underwent a 25-gauge transconjunctival sutureless PPV. The patients were divided into two groups: the air tamponade group (30 eyes) and the nontamponade group (31 eyes). Subfoveal choroidal thickness (SFCT) was analyzed over 12 months. No significant differences were seen between the two groups at baseline. For all patients, the SFCT was significantly thicker at 1 month after surgery and significantly thinner at 6 and 12 months after surgery than preoperative values. In the subgroup analysis, there was no significant difference in SFCT 3 months after surgery compared with the preoperative values in both groups, but SFCT was significantly lower 6 months after surgery in both groups. In conclusion, our results showed that choroidal thickness temporarily increased after surgery and then gradually decreased until 12 months after the PPV for ERM.
ABSTRACT
In this study we evaluated the effect of intravitreal brolucizumab injections on choroidal thickness in patients with neovascular age-related macular degeneration (nAMD) who previously showed an incomplete response to anti-vascular endothelial growth factor treatment. A total of thirty-four eyes from 34 patients were included in this study. The patients received an average of 2.4 ± 1.1 brolucizumab injections with the mean follow-up period of 4.9 ± 2.0 months. After their first brolucizumab treatment, the central foveal thickness (CFT) and subfoveal choroidal thickness (SFCT) were significantly decreased from 431.6 ± 190.0 µm and 193.9 ± 75.1 µm to 274.6 ± 109.4 µm (P < 0.001) and 169.4 ± 71.1 µm (P < 0.001), respectively. However, there were no improvements in visual acuity. Patients were divided into three subgroups according to the number of brolucizumab treatments: one, two, and three or more injections. In all three subgroups, the CFT and SFCT were significantly reduced compared to baseline at all time points of brolucizumab injections. In conclusion, choroidal thickness was significantly reduced after intravitreal brolucizumab injections as a switching treatment in patients with nAMD.
Subject(s)
Macular Degeneration , Ranibizumab , Humans , Angiogenesis Inhibitors/therapeutic use , Retrospective Studies , Macular Degeneration/drug therapyABSTRACT
This study investigated spectral-domain optical coherence tomography (OCT) biomarkers to predict short-term anatomical improvement for diabetic macular edema (DME) after dexamethasone (DEX) injection in intravitreal anti-vascular endothelial growth factor (anti-VEGF) non-responders. This retrospective comparative study included 31 eyes of 31 patients with DME unresponsive to anti-VEGF, divided into better and lesser responder groups. OCT prior to DEX injection was used to evaluate the morphological features including optical density (ODN) and optical density ratio (ODR) of the outer nuclear layer (ONL) cysts. Correlations between baseline OCT parameters and mean central macular thickness (CMT) changes at 1 month were analyzed. There were no between-group differences in age, sex, number of previous anti-VEGF injections, duration of diabetes, or HbA1c level. However, ODN and ODR values in ONL cysts were lower in the better responder group (p = 0.020 and p < 0.001, respectively). ODN and ODR showed negative correlations with CMT changes (R = −0.546, p = 0.002 and R = −0.436, p = 0.014, respectively). Higher OCT reflectivity in the foveal cystoid space was associated with a lesser decrease in CMT after DEX injection in anti-VEGF non-responders, suggesting the usefulness of this parameter in predicting short-term CMT responses after DEX injection.
ABSTRACT
The study aimed to evaluate the long-term changes in circumpapillary retinal nerve fiber layer (RNFL) thickness after vitrectomy for rhegmatogenous retinal detachment (RRD) repair. A total of 33 eyes of 33 patients were enrolled. By using optical coherence tomography, the circumpapillary RNFL thickness was measured before surgery and 1, 3, 6 months and 1, 2, 3 years after surgery and compared with the preoperative value. The effect of duration, location, and extent of RRD on RNFL thickness change was evaluated. There was a significant increase of circumpapillary RNFL thickness at the 1-month, 3-month [except in the nasal superior sector (P = 0.627)], and only in the nasal inferior sector at 6-month (P = 0.010) follow-up compared with the baseline value (all Ps < 0.05). No significant differences were observed 1, 2, and 3 years after the surgery (P > 0.05). The duration, location, and extent of detachment did not reveal significant correlations with RNFL parameters (P > 0.05). Circumpapillary RNFL thickness in eyes with RRD after vitrectomy demonstrated a transient increase during the early postoperative period. This increase was not associated with duration, location, and extent of RRD. At 3 years following surgery, no RNFL thinning or thickening was observed.
Subject(s)
Retinal Detachment , Vitrectomy , Humans , Nerve Fibers , Retinal Detachment/surgery , Retinal Ganglion Cells , Tomography, Optical Coherence/methods , Vitrectomy/methodsABSTRACT
We investigated the combined effect of silicone tamponade and the internal limiting membrane (ILM) peeling and investigated whether timing of peeling of the ILM affects the outcomes of vitrectomy with silicone oil tamponade in eyes with proliferative diabetic retinopathy (PDR). Here, we examined 63 eyes (58 patients) with PDR, which underwent vitrectomy with silicone oil tamponade and stepwise removal of silicone oil. ILM peeling was performed just before oil injection (group 1; 33 eyes, 30 patients) or after oil removal (group 2; 30 eyes, 28 patients). Visual acuity and retinal and choroidal thicknesses were compared between the groups. Thinning of the inner retina, including the ganglion cell-inner plexiform layer and macular retinal nerve fiber layer, was evident at 1 year after surgery in both groups. Thinning of the total retina (P = 0.019) and inner retina (P = 0.008) was significantly correlated with final visual acuity. There was no considerable between-group difference observed in final visual acuity, intraocular pressure, or retinal or choroidal thickness at 1 year after surgery. The incidence of epiretinal membrane was higher during silicone endo-tamponade in group 2 (P = 0.033). Visual recovery and macular configuration in eyes with PDR are not affected by whether the ILM is peeled before or after silicone oil tamponade.
Subject(s)
Diabetes Mellitus , Diabetic Retinopathy , Epiretinal Membrane , Retinal Detachment , Basement Membrane , Diabetes Mellitus/surgery , Diabetic Retinopathy/complications , Diabetic Retinopathy/surgery , Humans , Retina/surgery , Retrospective Studies , Silicone Oils , Tomography, Optical Coherence/adverse effects , Vitrectomy/adverse effectsABSTRACT
We aimed to evaluate the association between the prevalence of diabetic retinopathy (DR) and coffee consumption in a Korean population. This cross-sectional study was based on data from the 2008-2011 Korean National Health and Nutrition Survey. Among 37,753 survey participants, the data of 1350 subjects with type 2 diabetes who underwent DR examination were analyzed. DR was graded using the modified Airlie House classification system. Coffee consumption data were obtained through food frequency questionnaires and categorized into four groups: almost none, < 1 cup/day, 1 cup/day, and ≥ 2 cups/day. The relationship between DR and coffee consumption was evaluated using multivariable logistic regression models adjusted for age, sex, education, occupation, income, smoking, alcohol intake, body mass index, physical activity, hypertension, dyslipidemia, diabetes duration, and glycated hemoglobin. The prevalence of DR was 20.0%. Non-proliferative DR was observed in 87.8% of all DR patients, and proliferative DR in 12.2%. The prevalence of DR and vision-threatening DR showed a significantly decreasing tendency according to daily coffee consumption (P for trend 0.025 and 0.005, respectively) after adjustment for possible confounders. This tendency was more prominent in those aged < 65 years (P for trend 0.005 and 0.003, respectively). Our findings suggest coffee consumption might be associated with DR reduction especially in Koreans with diabetes mellitus aged < 65 years.
Subject(s)
Diabetes Mellitus, Type 2 , Diabetic Retinopathy , Adult , Coffee/adverse effects , Cross-Sectional Studies , Diabetes Mellitus, Type 2/complications , Diabetes Mellitus, Type 2/epidemiology , Diabetic Retinopathy/epidemiology , Diabetic Retinopathy/etiology , Humans , Prevalence , Risk FactorsABSTRACT
Neovascular age-related macular degeneration (nAMD) is among the main causes of visual impairment worldwide. We built a deep learning model to distinguish the subtypes of nAMD using spectral domain optical coherence tomography (SD-OCT) images. Data from SD-OCT images of nAMD (polypoidal choroidal vasculopathy, retinal angiomatous proliferation, and typical nAMD) and normal healthy patients were analyzed using a convolutional neural network (CNN). The model was trained and validated based on 4749 SD-OCT images from 347 patients and 50 healthy controls. To adopt an accurate and robust image classification architecture, we evaluated three well-known CNN structures (VGG-16, VGG-19, and ResNet) and two customized classification layers (fully connected layer with dropout vs. global average pooling). Following the test set performance, the model with the highest classification accuracy was used. Transfer learning and data augmentation were applied to improve the robustness and accuracy of the model. Our proposed model showed an accuracy of 87.4% on the test data (920 images), scoring higher than ten ophthalmologists, for the same data. Additionally, the part that our model judged to be important in classification was confirmed through Grad-CAM images, and consequently, it has a similar judgment criteria to that of ophthalmologists. Thus, we believe that our model can be used as an auxiliary tool in clinical practice.
Subject(s)
Deep Learning , Macular Degeneration/classification , Macular Degeneration/diagnosis , Neovascularization, Pathologic/classification , Neovascularization, Pathologic/diagnosis , Pattern Recognition, Automated/methods , Tomography, Optical Coherence/methods , Aged , Aged, 80 and over , Computer Simulation , Diagnostic Techniques, Ophthalmological/standards , Female , Humans , Macular Degeneration/diagnostic imaging , Male , Middle Aged , Neovascularization, Pathologic/diagnostic imaging , OphthalmologistsABSTRACT
Central serous chorioretinopathy (CSC) is one of the most common macular diseases that can reduce the quality of life of patients. This study aimed to build a deep learning-based classification model using multiple spectral domain optical coherence tomography (SD-OCT) images together to diagnose CSC. Our proposed system contains two modules: single-image prediction (SIP) and a final decision (FD) classifier. A total of 7425 SD-OCT images from 297 participants (109 acute CSC, 106 chronic CSC, 82 normal) were included. In the fivefold cross validation test, our model showed an average accuracy of 94.2%. Compared to other end-to-end models, for example, a 3D convolutional neural network (CNN) model and a CNN-long short-term memory (CNN-LSTM) model, the proposed system showed more than 10% higher accuracy. In the experiments comparing the proposed model and ophthalmologists, our model showed higher accuracy than experts in distinguishing between acute, chronic, and normal cases. Our results show that an automated deep learning-based model could play a supplementary role alongside ophthalmologists in the diagnosis and management of CSC. In particular, the proposed model seems clinically applicable because it can classify CSCs using multiple OCT images simultaneously.
Subject(s)
Central Serous Chorioretinopathy/classification , Central Serous Chorioretinopathy/diagnostic imaging , Deep Learning , Acute Disease , Adult , Aged , Case-Control Studies , Central Serous Chorioretinopathy/pathology , Chronic Disease , Diagnosis, Differential , Female , Humans , Image Interpretation, Computer-Assisted/statistics & numerical data , Male , Middle Aged , Retina , Tomography, Optical CoherenceABSTRACT
Central serous chorioretinopathy (CSC) is the fourth most common retinopathy and can reduce quality of life. CSC is assessed using optical coherence tomography (OCT), but deep learning systems have not been used to classify CSC subtypes. This study aimed to build a deep learning system model to distinguish CSC subtypes using a convolutional neural network (CNN). We enrolled 435 patients with CSC from a single tertiary center between January 2015 and January 2020. Data from spectral domain OCT (SD-OCT) images of the patients were analyzed using a deep CNN. Five-fold cross-validation was employed to evaluate the model's ability to discriminate acute, non-resolving, inactive, and chronic atrophic CSC. We compared the performances of the proposed model, Resnet-50, Inception-V3, and eight ophthalmologists. Overall, 3209 SD-OCT images were included. The proposed model showed an average cross-validation accuracy of 70.0% (95% confidence interval [CI], 0.676-0.718) and the highest test accuracy was 73.5%. Additional evaluation in an independent set of 104 patients demonstrated the reliable performance of the proposed model (accuracy: 76.8%). Our model could classify CSC subtypes with high accuracy. Thus, automated deep learning systems could be useful in the classification and management of CSC.