Automated ensemble method for merging retinal photoreceptor identification results on adaptive optics scanning laser ophthalmoscope images

Early diagnosis of retinopathy is important for its effective treatment. Adaptive optics scanning laser ophthalmoscopy is a promising cellular-level imaging technique for detecting retinopathy at an early stage. We propose an automated ensemble method for merging identification results of retinal photoreceptors on ophthalmoscope images. The different identification results and corresponding ground truth are generated by simulation, and the identification results are merged to increase the accuracy of retinal photoreceptor identification. The effectiveness of our ensemble method is demonstrated by comparing the generated and ensemble retinal photoreceptor identification results with the ground truths in terms of precision, recall, and F1-score, which shows the proposed method can improve retinal photoreceptor identification.