RESULTS:Compared with manual analysis, the sensitivity, specificity, missed diagnosis rate and misdiagnosis rate of AI were 91.84%, 99.06%, 8.16% and 0.94% respectively. The Kappa value of consistency analysis of the two diagnosis results was 0.817(P<0.001). Compared with manual analysis, the sensitivity and specificity of AI group to diagnose non-DR were 99.06% and 91.84% respectively. The sensitivity and specificity of mild NPDR were 85.36% and 98.52% respectively. The sensitivity and specificity of moderate NPDR were 81.53% and 98.55% respectively. The sensitivity and specificity of severe NPDR were 70% and 99.51% respectively. The sensitivity and specificity of PDR were 86.67% and 99.63% respectively. The Kappa value of the consistency analysis of DR staging diagnosis was 0.878(P<0.01).

CONCLUSION: The AI remote screening system adopted in this study showed good consistency with the results of manual analysis, which can meet the needs of DR screening and provide a new effective prevention and treatment mode for DR patients in the community.]]> 2022/7/27 16:29:06 Zhen Li, Jun-Feng Piao, Xiao-Ting Li, Ning Wang, Xue Xiao and Wei-Ning Rong 21true METHODS:Infrared meibomian gland images were collected and 193 of them were picked out for establishing the database. The images were manually labeled by three clinicians. UNet++ network and automatic data expansion strategy were introduced to construct the automatic meibomian glands segmentation model. The feasibility and effectiveness of the proposed segmentation model were analyzed by precision, sensitivity, specificity, accuracy and intersection over union.

RESULTS: Taking manual labeling as the gold standard, the presented method segment the glands effectively and steadily with accuracy, sensitivity and specificity of 94.31%, 82.15% and 96.13% respectively. On the average, only 0.11s was taken for glands segmentation of single image.

CONCLUSIONS: In this paper, deep learning technology is introduced to realize automatic segmentation of meibomian glands, achieving high accuracy, good stability and efficiency. It would be quite useful for calculation of gland morphological parameters, the clinical diagnosis and screening of related diseases, improving the diagnostic efficiency.]]> 2022/6/28 10:46:47 Jia-Wen Lin, Zhi-Ming Lin, Tai-Chen Lai, Lin-Ling Guo, Jing Zou and Li Li 20true METHODS: The network structure of Phase-fusion PSPNet for the segmentation of pterygium lesions is proposed based on the PSPNet model structure. In our network, the up-sampling module is connected behind the pyramid pooling module, which gradually increase the sampling based on the principle of phased increase. Therefore, the information loss is reduced, it is suitable for segmentation tasks with fuzzy edges. The experiments conducted on the dataset provided by the Affiliated Eye Hospital of Nanjing Medical University, which includes 517 ocular surface photographic images of pterygium were divided into training set(330 images), validation set(37 images)and test set(150 images), which the training set and the validation set images are used for training, and the test set images are only used for testing. Comparing results of intelligent segmentation and expert annotation of pterygium lesions.

RESULTS: Phase-fusion PSPNet network structure for pterygium mean intersection over union(MIOU)and mean average precision(MPA)were 86.31% and 91.91%, respectively, and pterygium intersection over union(IOU)and average precision(PA)were 77.64% and 86.10%, respectively.

CONCLUSION: Convolutional neural networks can segment pterygium lesions with high precision, which is helpful to provide an important reference for doctors' further diagnosis of disease and surgical recommendations, and can also visualize the pterygium intelligent diagnosis.]]> 2022/5/30 15:27:46 Shao-Jun Zhu, Xin-Wen Fang, Bo Zheng, Mao-Nian Wu and Wei-Hua Yang 19true 2022/5/30 15:27:46 Yan-Xi Wang, Cheng-Hu Wang, Jing-Yue Zhang, Man-Hua Xu, Zheng-Hong Peng, Ye Nie and Gang-Jin Kang 18true METHODS:Twenty ophthalmology residents with equal seniority who had completed 3a standardized residency training in the Affiliated Eye Hospital of Nanjing Medical University from 2019 to 2021 were prospectively enrolled. After passing the theoretical examination, residents were randomly divided into virtual surgery exercise(Dry-lab)group(n=10)and real animal surgery exercise(Wet-lab)group(n=10). Dry-lab and Wet-lab group residents performed training using the Eye SI surgical simulator and pig eye respectively. At the end of the training, the overall training effects of the residents in both groups were rated using the Eye SI surgical simulator and the real pig eye operation, and the module training effects of the residents in both groups were rated using the virtual surgical simulator. Furthermore, a questionnaire survey was used to objectively evaluate the two training methods.

RESULTS:Residents in Dry-lab group had significantly higher total scores on both operation assessments,simulator assessment and real pig eye operation assessment than Wet-lab group(88.03±1.34 vs 80.35±2.87, 87.50±3.03 vs 77.60±5.62, 88.57±1.89 vs 83.10±3.22, all P<0.01). The simulator module assessment results showed that the residents in Dry-lab group scored significantly better than Wet-lab group in terms of scores and completion time on each module(P<0.01). The questionnaire results showed that Dry-lab group rated better than Wet-lab group in terms of the novelty of training, the proximity to the real surgical experience, the degree of help to the improvement of microsurgery skills, the confidence to perform real surgery, and the overall satisfaction of surgical training(P<0.05).

CONCLUSION:Applying virtual-reality surgery exercise to cataract surgery skills training for ophthalmology residents can significantly improve the cataract skills, increase overall training satisfaction, and help residents enhance their confidence, psychological quality, decision-making, and processing ability during real surgery at the initial stage of practice. This provides a new standard and model for establishing a formal and standardized cataract surgery training system for ophthalmology residents.]]> 2022/4/24 18:49:20 Jia-Jun Li, Ke-Ran Li and Wei-Hong Shang 17true 2022/4/24 18:49:21 Pei-Yu Liu and Xu Zhang 16true METHODS: A total of 750 images of normal, observational and operative anterior sections of pterygium were collected from the Affiliated Eye Hospital of Nanjing Medical University between May 2020 and April 2021. Seven triclassification models were respectively trained with original data set and enhanced data set. Totally 470 clinical images were tested, and the generalization ability of the model before and after data enhancement was compared to determine the best model for the automatic classification system of pterygium.

RESULTS:The average sensitivity, specificity and AUC of the best model trained on the original data set were 92.55%, 96.86% and 94.70% respectively. After data was enhanced, the sensitivity, specificity and AUC of different models were increased by 3.7%, 1.9% and 2.7% on average. The sensitivity, specificity and AUC of the EfficientNetB7 model trained on the enhanced data set were 93.63%, 97.34% and 95.47% respectively.

CONDLUSION: The EfficientNetB7 model, which was trained on the enhanced data set, achieves the best classification result and can be used in the automatic classification system of pterygium.This automatic classification system can diagnose diseases about pterygium better, and it is expected to be an effective screening tool for primary medical care. It also provides reference for the research on the refinement and grading of pterygium.]]> 2022/4/24 18:49:21 Kai He, Mao-Nian Wu, Bo Zheng, Wei-Hua Yang, Shao-Jun Zhu and Ling Jin 15true 2022/4/24 18:49:21 Shuai Yang, Jie Shao and Jun Zhang 14true 2022/4/24 18:49:21 Jing-Wen Wang and Wen Xu 13true 2022/4/24 18:49:21 Yi Yu, Yi-Wen Zhou and Yan-Ning Yang 12true 2022/4/24 18:49:21 Yu-Ke Ji, Nan Chen, Zhi-Peng Yan, Ke-Ran Li, Cheng-Hu Wang, Guo-Fan Cao, Qin Jiang and Wei-Hua Yang 11true METHODS: A retrospective study. In this study, 2 157 images of fungal keratitis patients who underwent in vivo confocal microscopy examination in the Department of Ophthalmology of the people's Hospital of Guangxi Zhuang Autonomous Region from March 2017 to January 2020 were included as the dataset, which was classified according to the results of microbial culture. The dataset was subsequently randomly divided into training set(1 380 images), validation set(345 images)and test set(432 images). We used the transfer learning Inception-ResNet V2 network to construct a diagnostic model, and to compare the performance of the model trained on different datasets. The performance of the diagnostic model evaluated with specificity, sensitivity, accuracy, and area under the receiver operating characteristics curve(AUC).

RESULTS: The model trained with the original training set had a specificity rate of 71.6%, a sensitivity rate of 72.0%, an accuracy rate of 71.8% and AUC of 0.785(95%CI: 0.742-0.828, P<0.0001). And the model trained with the augmented training set had a specificity rate of 76.6%, a sensitivity rate of 83.1%, an accuracy rate of 79.9% and AUC of 0.876(95%CI: 0.843-0.909, P<0.0001), which made the model's prediction performance boost.

CONCLUSION: In this study, we constructed an intelligent diagnosis system for fungal keratitis fusarium through transfer learning, which has higher accuracy, and realized the intelligent diagnosis of fungal keratitis pathogen fusarium. Furthermore, we verified that the data augmentation strategy can improve the performance of the intelligent diagnosis system when the original dataset is limited, and this method can be used for intelligent diagnosis and identification of fungal keratitis pathogen fusarium.]]> 2022/4/24 18:49:22 Guang-Yi Huang, Ning-Ning Tang, Qi Chen, Qian-Qian Lan, Li Jiang, Yi-Yi Hong, Jian Lyu, Min Li, Si-Ming Zeng and Fan Xu 10true METHODS: From November 2019 to November 2021, a total of 1 579 OCT images were collected in the outpatient department, and the imaging results of ophthalmologist and AI were collected. The Kappa consistency test of classification results without ARMD and with ARMD were analyzed.

RESULTS: The Kappa coefficients of the judgement of ophthalmologists about ARMD was 0.934. The Kappa coefficients between AI and ophthalmologists was 0.738. The sensitivity, specificity and area under curve(AUC)of AI to ARMD were 73.08%, 95.07% and 0.841 respectively.

CONCLUSION: AI has a high consistency with ophthalmologists in the recognition of ARMD based on OCT images, which is suitable for primary hospitals to complete the early screening and early referral of ARMD.]]> 2022/4/24 18:49:22 Yan Jiang, Fei-Ping Xu, Jing-Cheng Wang, Sha-Sha Wang, Rui Liu, Ting-Yi Cao, Wen Yuan, Xin-Jian Chen and Ji-Li Chen 9true METHODS:The right eyes of 145 subjects who were treated at the Hangzhou Branch of the Eye Hospital of Wenzhou Medical University from January to November 2021 were selected for inclusion in the study. Meibography images of 60 of these subjects were randomly selected for AI training. The meibomian region and each MG in meibography were annotated and formed into datasets. The datasets were used for training and obtaining an AI system based on residual neural network(ResNet)combined with the U-Net model. The AI system was used to automatically analyze the MG morphological parameters in 85 subjects, including 53 patients with obstructive MGD and 32 volunteers with normal meibomian glands. The clinical indices including ocular surface disease index(OSDI), tear meniscus height, tear film break-up time(TBUT), corneal fluorescein staining, lid margin score, meiboscore, and meibomian gland expressibility score were also observed. The correlation between MG morphological parameters and clinical indices were analyzed.

RESULTS: After several iterations, we finally obtained an AI system with Intersection over Union of 92.0%. Using this AI system, we found that there was a significant correlation between the MG density in the upper eyelid with OSDI(r_s=-0.320), TBUT(r_s=0.484), lid margin score(r_s=-0.350), meiboscore(r_s=-0.749), and meibum expressibility score(r_s=0.425)(all P<0.05). The MG density in the lower eyelid was significantly correlated with OSDI(r_s=-0.420), TBUT(r_s= 0.598), lid margin score(r_s=-0.396), meiboscore(r_s=-0.720), and meibum expressibility score(r_s=0.438)(all P<0.05). The MG density in the total eyelid was significantly correlated with OSDI(r_s=-0.404), TBUT(r_s=0.601), lid margin score(r_s=-0.416), meiboscore(r_s=-0.805), and meibum expressibility score(r_s=0.480)(all P<0.05).

CONCLUSION:The AI system based on CNN in this study is an accurate and efficient MG morphological evaluation system, which can be conveniently used to evaluate the MG morphology of MGD patients quickly and accurately by using the MG density index established by us. MG density is a new quantitative index to evaluate meibomian gland atrophy, which is more accurate than meiboscore.]]> 2022/4/24 18:49:22 Zu-Hui Zhang, Xin-Xin Yu, Xiao-Lei Lin, Ya-Na Fu and Qi Dai 8true 2022/11/29 14:58:55 Xue Han, Jing-Juan Ding, Shu-Ting Lu, Qin Jiang, Wei-Hua Yang and Jin-Song Xue 7true METHODS:In this retrospective case series study, the 210 preoperative and postoperative corneal topographies from 87 age-related cataract patients(105 eyes)undergoing femtosecond laser arcuate keratotomy at Shanxi Eye Hospital between March 2018 and March 2020 were selected and divided into a training set(180)and a test set(30)for model training and testing. The peak signal-to-noise ratio(PSNR), structural similarity(SSIM)and Alpins astigmatism vector analysis were used to compare the accuracy of postoperative corneal topography prediction under different attention mechanisms.

RESULTS:The model based on attention mechanism and Pix2Pix network can predict postoperative corneal topography, among which the model based on Self-Attention mechanism has the best prediction effect, with PSNR and SSIM reaching 16.048 and 0.7661, respectively. There were no statistically significant differences in the difference vector, difference vector axis position, surgically induced astigmatism, and correction index between real and generated corneal topography on the 3mm and 5mm rings(all P>0.05).

CONCLUSION:Based on the Self-Attention mechanism and Pix2Pix network, the postoperative corneal topography can be well predicted, which can provide reference for the surgical planning and postoperative effects of ophthalmic clinicians.]]> 2023/5/29 16:33:23 Guang-Hua Zhang, Nan Cheng, Zhe Zhang, Xiao-Na Li, Jing Pan, En-Hui Li and Wei-Yi Chen 6true 2023/5/29 16:33:23 Yu Zhou, Min Zhang, Yu-Jie Zhu and Qiong Lu 5true METHODS: A total of 4 196 UBM images were obtained from 675 patients(1 130 eyes)at the Eye Center of Renmin Hospital of Wuhan University from January 2021 to June 2022 were collected to build an image dataset. Using Unet++network to automatically segment ACA tissue, a support vector machine(SVM)algorithm was developed to automatically classify opening and closing of chamber angle, and an algorithm to automatically locate the sclera spur and measure ACA parameters was developed. Furthermore, a total of 631 UBM images of 127 subjects(221 eyes)at Huangshi Aier Eye Hospital and 594 UBM images of 188 subjects(257 eyes)at Zhongnan Hospital of Wuhan University were selected to evaluate the performance of the system under different environments.

RESULTS: The accuracy of the analysis system constructed in this study for chamber angle opening and closing was 95.71%. The intra-class correlation coefficient(ICC)values of all ACA angle parameters were greater than 0.960. ICC values of all ACA thickness parameters were greater than 0.884. The accurate measurement of ACA parameters depended in part on the accurate location of the scleral spur.

CONCLUSION: The intelligent analysis system constructed in this study can accurately and effectively evaluate ACA images automatically and is a potential screening tool for the rapid identification of ACA structures.]]> 2023/4/27 14:33:12 Wei-Yan Jiang, Yu-Lin Yan, Si-Min Cheng and Yan-Ning Yang 4true 2023/4/27 14:33:12 De-Sheng Liao and Min Wu 3true METHODS: Diagnostic study. A total of 830 patients(830 eyes)were collected, including 338 male(338 eyes)and 492 female(492 eyes), with an average age of 14-36(23.19±5.71)years. Among them, 731 patients(731 eyes)had undergone corneal refractive surgery at Chongqing Nanping Aier Eye Hospital from January 2020 to March 2022, and 99 patients had a diagnosed keratoconus from January 2015 to March 2022. Corneal diameter ≤11.1 mm was measured by Pentacam in all patients. Two cornea specialists classified patients' data into normal corneas, suspect keratoconus, and keratoconus groups based on the Belin/Ambrósio enhanced ectasia display(BAD)system in Pentacam. The data of 665 patients were randomly selected as the training set and the other 165 patients as the validation set by computer random sampling method. Seven parametric corneal features were extracted by convolutional neural networks(CNN), and the models were built by Residual Network(ResNet), Vision Transformer(ViT), and CNN+Transformer, respectively. The diagnostic accuracy of models was verified by cross-entropy loss and cross-validation method. In addition, sensitivity and specificity were evaluated using receiver operating characteristic curve.

RESULTS: The accuracy of ResNet, ViT, and CNN+Transfermer for the diagnosis of normal cornea and suspect keratoconus was 85.57%, 86.11%, and 86.54% respectively, and the area under the receiver operating characteristic curve(AUC)was 0.823, 0.830 and 0.842 respectively. The accuracy of models for the diagnosis of suspect keratoconus and keratoconus was 97.22%, 95.83%, and 98.61%, respectively, and the AUC was 0.951, 0.939, and 0.988 respectively.

CONCLUSION: For corneas ≤11.1 mm in diameter, the data model established by CNN+Transformer has a high accuracy rate for classifying keratoconus, which provides real and effective guidance for early screening.]]> 2023/2/2 16:41:49 Di-Hua Ao, Xi-Rui Tian, Ming-Xun Ma, Bo Zhang, Min Chen and Yan-Li Peng 2true 2023/10/24 9:37:23 Xiao-Pei Zhang, Jian-Feng Huang, Tong-Yan Li, Wei-Hua Yang 1true