Improving the Generalizability of Infantile Cataracts Detection via Deep Learning-Based Lens Partition Strategy and Multicenter Datasets

Abstract

Infantile cataract is the main cause of infant blindness worldwide. Although previous studies developed artificial intelligence (AI) diagnostic systems for detecting infantile cataracts in a single center, its generalizability is not ideal because of the complicated noises and heterogeneity of multicenter slit-lamp images, which impedes the application of these AI systems in real-world clinics. In this study, we developed two lens partition strategies (LPSs) based on deep learning Faster R-CNN and Hough transform for improving the generalizability of infantile cataracts detection. A total of 1,643 multicenter slit-lamp images collected from five ophthalmic clinics were used to evaluate the performance of LPSs. The generalizability of Faster R-CNN for screening and grading was explored by sequentially adding multicenter images to the training dataset. For the normal and abnormal lenses partition, the Faster R-CNN achieved the average intersection over union of 0.9419 and 0.9107, respectively, and their average precisions are both > 95%. Compared with the Hough transform, the accuracy, specificity, and sensitivity of Faster R-CNN for opacity area grading were improved by 5.31, 8.09, and 3.29%, respectively. Similar improvements were presented on the other grading of opacity density and location. The minimal training sample size required by Faster R-CNN is determined on multicenter slit-lamp images. Furthermore, the Faster R-CNN achieved real-time lens partition with only 0.25 s for a single image, whereas the Hough transform needs 34.46 s. Finally, using Grad-Cam and t-SNE techniques, the most relevant lesion regions were highlighted in heatmaps, and the high-level features were discriminated. This study provides an effective LPS for improving the generalizability of infantile cataracts detection. This system has the potential to be applied to multicenter slit-lamp images.

Influence of short-term hypoxia exposure on dynamic visual acuity

“…DVA, or dynamic visual acuity, refers to the ability to capture, decompose, and perceive images of moving objects while observing (Jiang et al, 2021 ). For high-altitude flight, pilots must constantly observe the display and control interfaces and the external environment in a relatively non-stationary state in order to rapidly and accurately obtain real-time information related to flight safety, which is crucial to ensure the stable operation of the aircraft.…”

AI-based diagnosis of nuclear cataract from slit-lamp videos

“…Wu et al 16 demonstrated robust performance in cataract diagnostic AI using slit-lamp photographs with equal outcomes for ophthalmologists, with a sensitivity and specificity of 92.00% and 83.85%, respectively (Table 1 ). Jiang et al 22 introduced a deep learning-based lens partition model, employing multicenter datasets, which yielded accuracy, sensitivity, and specificity rates ranging from 92.57 to 97.96%, 91.95–97.04%, and 93.08–98.74%, respectively (Table 1 ). Focusing of different disease, Li et al 15 revealed a high performance in the keratitis diagnostic AI model slit-lamp and smartphone photograph (sensitivity: 81.5–98.7%, specificity: 95.0–99.8%, and accuracy: 95.4–99.3%).…”

Cataract Risk Factors Prediction Using Deep Leaning Models

“…This model has the ability to transform cataracts on the basis of monitoring and computation of the intraocular lens (IOL). Jiang et al [18] proposed the Hough transform and R-CNN to improve the generalization of IC (Infantile Cataracts) diagnosis. There were 1643 slit-lamp images were collected from five ophthalmic clinics and they were calculated using the LPS (lens partition strategies).…”