AI App Detects Eye Cancer from Photos with High Accuracy

A new smartphone AI application has been developed to detect eye cancers from photographs, demonstrating high accuracy in identifying potential malignancies. This research, conducted in China, involved a clinical trial with 614 participants, with a median age of 46, recruited through television and social media.

The app uses a deep learning model trained on images collected over a decade from specialist clinics. It achieved an area under the curve (AUC) of 0.945 in distinguishing malignant from benign lesions, improving to 0.977 in real-world use.

The application generated 58 referrals, leading to the confirmation of 20 malignancies through histopathology. The average number of referrals required for definitive treatment dropped significantly from 3.69 to 1.02 after the introduction of the system.

However, the study acknowledged limitations, including a small number of non-Chinese participants and a focus on short-term outcomes rather than long-term clinical impact. The research was published in JAMA Ophthalmology and was led by Sun Yat-sen University.