JMIR Publications

ABSTRACT

Background:

Retinopathy pf prematurity is the leading preventable cause of childhood blind-ness, timely intravitreal injection of anti-vascular endothelial growth factor is re-quired to prevent retinal detachment with consequent vision impairment and loss. However, anti- vascular endothelial growth factor has been reported to be associated with ROP reactivation. Therefore, prediction of reactivation after treatment is urgent need.

Objective:

To develop and validate prediction models for reactivation after anti-vascular endothelial growth factor intravitreal injection in infants with retinopathy of prema-turity using multimodal machine learning algorithms.

Methods:

Infants with ROP undergoing anti-vascular endothelial growth factor treatment were recruited from three hospitals, conventional machine learning, deep learning and fusion models were constructed. The areas under the curve, accurancy, sensitivity and specificity were used to show the performances of the prediction models.

Results:

239 cases with anti-vascular endothelial growth factor treatment were recruit-ed, including 90 with reactivation and 149 non-reactivation cases. The area under the curve for the conventional machine learning model was 0.806 and 0.805 in the inter-nal and external validation groups, respectively. The average area under the curve, sensitivity, and specificity in the external validation for the deep learning model were 0.787, 0.800 and 0.570, respectively. The specificity, area under the curve, and sensitivity for the fusion model were 0.686, 0.822, and 0.800 in external validation, separately.

Conclusions:

We constructed three prediction models for the reactivation of retinopathy of prematurity, fusion model achieved the best performance. Using this prediction model, we may optimize strategies for treating retinopathy of prematurity infants and developing better screening plans after treatment.