Novel Uses and Challenges of Artificial Intelligence in Diagnosing and Managing Eyes with High Myopia and Pathologic Myopia

A study on the variability and correlation of ocular biological measurement parameters in adult myopic patients

Objective

This study aims to explore the differences in ocular parameters among adult myopic patients with different degrees of myopia and axial lengths, and to investigate the correlations between these ocular parameters.

Methods

This single-center observational study collected clinical data from myopic patients aged 18-45 years who visited the Eye Hospital of Nanjing Medical University between January and June 2023. The data included laterality, diopter of spherical power (DS), diopter of cylindrical power (DC), spherical equivalent (SE), axial length (AL), central corneal thickness (CCT), flat meridian keratometry (K1), steep meridian keratometry (K2), mean keratometry (Km), anterior chamber depth (ACD), corneal radius of curvature (CRC), and axial length/corneal radius of curvature ratio (AL/CRC). Following predefined inclusion and exclusion criteria, 1,026 eyes were included in the study. Patients were grouped based on SE and AL parameters into different degrees of myopia. Analysis of variance (ANOVA) and Welch ANOVA were used to compare intergroup differences. Spearman correlation coefficients were calculated to analyze the correlations between parameters, and linear regression and ROC curve analyses were performed.

Results

Significant differences (p < 0.05) were found among mild, moderate, and high myopia groups in parameters such as DS, DC, AL, K1, Km, ACD, CRC, and AL/CRC. Significant differences (p < 0.05) were also found in DS, DC, SE, CCT, K1, K2, Km, ACD, CRC, and AL/CRC among different axial length groups. Spearman correlation analysis showed a strong correlation between AL and DS, SE, and between AL/CRC and DS, SE, AL. Linear regression analysis revealed that the coefficient of determination (R2) for AL and SE was 0.699, and for AL/CRC and SE, it was 0.861. ROC curve analysis demonstrated high accuracy for both AL and AL/CRC in identifying high myopia, with an AUC of 0.952 for AL/CRC, which was superior to the AUC of 0.905 for AL (p < 0.05).

Conclusion

This study found significant differences in ocular parameters among patients with different degrees of myopia and axial lengths. There was a significant negative correlation between AL, AL/CRC, and SE. Compared to AL, AL/CRC had a stronger correlation with SE and higher accuracy in identifying high myopia.

Impacts of environments on school myopia by spatial analysis techniques in Wuhan

The prevalence of myopia in China has increased significantly in recent years, and the age of onset has become younger. Previous studies have indicated that various environmental factors can influence the development of school myopia. However, the environmental impacts on school myopia remains to be investigated. Discoveries in this field may contribute to better urban planning. This study involved 7,610 students (aged 6-12 years, 4084 boys and 3526 girls) from six primary schools in Wuhan, China. We evaluated the associations between school myopia and the environment by analyzing the geographical distribution of myopic children. We utilized the spatial statistical analysis model. The Normalized Difference Vegetation Index (NDVI) risk coefficient for a 5,000-m radius around target schools was 0.379 (p = 0.008), while the NDVI risk coefficient for a 100-m radius around target schools was 0.241 (p = 0.047). The sports area risk coefficient for a 5,000-m radius around target schools was 0.234 (p = 0.016). We found that the specific buffers of NDVI and sports area around schools were associated with the prevalence of school myopia in schools, which worth further research to guide future initiatives on school myopia from an environmental perspective.

A preliminary study of artificial intelligence to recognize tessellated fundus in visual function screening of 7-14 year old students

BACKGROUND

To evaluate the accuracy of artificial intelligence (AI)-based technology in recognizing tessellated fundus in students aged 7-14 years.

METHODS

A retrospective study was conducted to collect consecutive fundus photographs for visual function screening of students aged 7-14 years old in Haikou City from June 2018 to May 2019, and 1907 cases were included in the study. Among them, 949 cases were male and 958cases were female. The results were manually analyzed by two attending ophthalmologists to ensure the accuracy of the results. In case of discrepancies between the results analyzed by the two methods, the manual results were used as the standard. To assess the sensitivity and specificity of AI in recognizing tessellated fundus, a Kappa consistency test was performed comparing the results of manual recognition with those of AI recognition.

RESULTS

Among 1907 cases, 1782 cases, or 93.4%, were completely consistent with the recognition results of manual and AI; 125 cases, or 6.6%, were analyzed with differences. The diagnostic rates of manual and AI for tessellated fundus were 26.1% and 26.4%, respectively. The sensitivity, specificity and area of the ROC curve (AUC) of AI for recognizing tessellated fundus in students aged 7-14 years were 88.0%, 95.4% and 0.917, respectively. The results of test showed that that the manual and AI identification results were highly consistent (κ = 0.831, P = 0.000).

CONCLUSION

AI analysis has high specificity and sensitivity for tessellated fundus identification in students aged 7-14 years, and it is feasible to apply artificial intelligence to visual function screening in students aged 7-14 years.

Multimodal deep transfer learning to predict retinal vein occlusion macular edema recurrence after anti-VEGF therapy

Purpose

To develop a multimodal deep transfer learning (DTL) fusion model using optical coherence tomography angiography (OCTA) images to predict the recurrence of retinal vein occlusion (RVO) and macular edema (ME) after three consecutive anti-VEGF therapies.

Methods

This retrospective cross-sectional study consisted of 2800 B-scan OCTA macular images collected from 140 patients with RVO-ME. The central macular thickness (CMT) > 250 μm was used as a criterion for recurrence in the three-month follow-up after three injections of anti-VEGF therapy. The qualified OCTA image preprocessing and the lesion area segmentation were performed by senior ophthalmologists. We developed and validated the clinical, DTL, and multimodal fusion models based on clinical and extracted OCTA imaging features. The performance of the models and experts predictions were evaluated using several performance metrics, including the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, and specificity.

Results

The DTL models exhibited higher prediction efficacy than the clinical models and experts' predictions. Among the DTL models, the Vgg19 performed better than that of the other models, with an AUC of 0.968 (95 % CI, 0.943-0.994), accuracy of 0.913, sensitivity of 0.922, and specificity of 0.902 in the validation cohort. Moreover, the fusion Vgg19 model showed the highest prediction efficacy among all the models, with an AUC of 0.972 (95 % CI, 0.946-0.997), accuracy of 0.935, sensitivity of 0.935, and specificity of 0.934 in the validation cohort.

Conclusions

Multimodal fusion DTL models showed robust performance in predicting RVO-ME recurrence and may be applied to assist clinicians in determining patients' follow-up time after anti-VEGF therapy.

Insights into artificial intelligence in myopia management: from a data perspective

Given the high incidence and prevalence of myopia, the current healthcare system is struggling to handle the task of myopia management, which is worsened by home quarantine during the ongoing COVID-19 pandemic. The utilization of artificial intelligence (AI) in ophthalmology is thriving, yet not enough in myopia. AI can serve as a solution for the myopia pandemic, with application potential in early identification, risk stratification, progression prediction, and timely intervention. The datasets used for developing AI models are the foundation and determine the upper limit of performance. Data generated from clinical practice in managing myopia can be categorized into clinical data and imaging data, and different AI methods can be used for analysis. In this review, we comprehensively review the current application status of AI in myopia with an emphasis on data modalities used for developing AI models. We propose that establishing large public datasets with high quality, enhancing the model's capability of handling multimodal input, and exploring novel data modalities could be of great significance for the further application of AI for myopia.

DeepDrRVO: A GAN-auxiliary two-step masked transformer framework benefits early recognition and differential diagnosis of retinal vascular occlusion from color fundus photographs

Retinal vascular occlusion (RVO) are common causes of visual impairment. Accurate recognition and differential diagnosis of RVO are unmet medical needs for determining appropriate treatments and health care to properly manage the ocular condition and minimize the damaging effects. To leverage deep learning as a potential solution to detect RVO reliably, we developed a deep learning model on color fundus photographs (CFPs) using a two-step masked SwinTransformer with a Few-Sample Generator (FSG)-auxiliary training framework (called DeepDrRVO) for early and differential RVO diagnosis. The DeepDrRVO was trained on the training set from the in-house cohort and achieved consistently high performance in early recognition and differential diagnosis of RVO in the validation set from the in-house cohort with an accuracy of 86.3%, and other three independent multi-center cohorts with the accuracy of 92.6%, 90.8%, and 100%. Further comparative analysis showed that the proposed DeepDrRVO outperforms conventional state-of-the-art classification models, such as ResNet18, ResNet50d, MobileNetv3, and EfficientNetb1. These results highlight the potential benefits of the deep learning model in automatic early RVO detection and differential diagnosis for improving clinical outcomes and providing insights into diagnosing other ocular diseases with a few-shot learning challenge. The DeepDrRVO is publicly available on https://github.com/ZhouSunLab-Workshops/DeepDrRVO.

Artificial intelligence technology for myopia challenges: A review

Myopia is a significant global health concern and affects human visual function, resulting in blurred vision at a distance. There are still many unsolved challenges in this field that require the help of new technologies. Currently, artificial intelligence (AI) technology is dominating medical image and data analysis and has been introduced to address challenges in the clinical practice of many ocular diseases. AI research in myopia is still in its early stages. Understanding the strengths and limitations of each AI method in specific tasks of myopia could be of great value and might help us to choose appropriate approaches for different tasks. This article reviews and elaborates on the technical details of AI methods applied for myopia risk prediction, screening and diagnosis, pathogenesis, and treatment.