An Outperforming Artificial Intelligence Model to Identify Referable Blepharoptosis for General Practitioners (Preprint)
BACKGROUND Accurate identification and prompt referral for blepharoptosis can be challenging for general practitioners. An artificial intelligence-aided diagnostic tool could underpin decision-making. OBJECTIVE To develop an AI model which accurately identifies referable blepharoptosis automatically and to compare the AI model’s performance to a group of non-ophthalmic physicians. METHODS Retrospective 1,000 single-eye images from tertiary oculoplastic clinics were labeled by three oculoplastic surgeons with ptosis, including true and pseudoptosis, versus healthy eyelid. The VGG (Visual Geometry Group)-16 model was trained for binary classification. The same dataset was used in testing three non-ophthalmic physicians. The Gradient-weighted Class Activation Mapping (Grad-CAM) was applied to visualize the AI model RESULTS The VGG16-based AI model achieved a sensitivity of 92% and a specificity of 88%, compared with the non-ophthalmic physician group, who achieved a mean sensitivity of 72% [Range: 68% - 76%] and a mean specificity of 82.67% [Range: 72% - 88%]. The area under the curve (AUC) of the AI model was 0.987. The Grad-CAM results for ptosis predictions highlighted the area between the upper eyelid margin and central corneal light reflex. CONCLUSIONS The AI model shows better performance than the non-ophthalmic physician group in identifying referable blepharoptosis, including true and pseudoptosis, correctly. Therefore, artificial intelligence-aided tools have the potential to assist in the diagnosis and referral of blepharoptosis for general practitioners.