A research study led by Shanghai Jiao Tong University School of Medicine shows that a combination of artificial intelligence (AI) and smartphone photography can be used to diagnose common childhood eye conditions.
As reported in JAMA Network Open, the investigators developed a deep-learning based model that was able to identify shortsightedness (myopia), squint (strabismus) and ptosis, a condition affecting the eyelids, from smartphone photos.
“Myopia, strabismus, and ptosis are common eye problems in children that can greatly damage their visual health, overall well-being, and development,” write the authors.
“Early screening and identification of these diseases are essential for successful management and therapy. Screening for various eye illnesses is primarily conducted in hospitals by expert ophthalmologists, causing delays in screening, diagnosis, and treatment. A useful screening approach is needed to allow parents to do early screening at home.”
To build the AI model, Lin Li and Jie Xu, both based at the Shanghai Jiao Tong University School of Medicine, and colleagues used 1419 images from 479 children aged six to 12 years diagnosed with the three eye conditions.
The results showed that the model correctly detected myopia, strabismus and ptosis with a sensitivity of 0.84, 0.73 and 0.85, respectively. The specificity, ability to identify those without the condition, was a respective 0.76, 0.85, and 0.95, for the same three conditions.
Overall, the accuracy of the model for detecting myopia, strabismus and ptosis was 80%, 80%, and 92%, respectively.
“This cross-sectional study found that the detection model using AI showed strong performance in accurately identifying myopia, strabismus, and ptosis using only smartphone images,” conclude the authors.
“These results suggest that it can assist families in screening children for myopia, strabismus, and ptosis, facilitating early identification and reducing the risk of visual function loss and severe problems due to delayed screening. Moreover, using such information can help achieve a more equitable allocation of limited medical resources.”
The authors are not alone in their efforts to diagnose eye conditions using smartphone photography and apps. A recent article assessing the field found more than 48 apps aimed at diagnosing or helping to treat eye diseases referenced in over 70 research papers, 34 of which were publicly available.
For any medical app, accuracy is key. The researchers want to increase the number of pictures of each condition fed into their model to increase the overall accuracy of the model further. They also acknowledge that “collecting patients’ images from various perspectives can enhance the algorithm’s performance,” compared with just one snapshot.
