Evaluation of Image Interpolations for AMD Detection on OCT Images Using Mobile-OCT CNN Model

This paper employed various interpolation techniques for the optical coherence tomography (OCT) image scaling of age-related macular degeneration (AMD) detection using the MobileOCT convolutional neural network (CNN) model. The AMD detection accuracy was mainly used to evaluate the distinct interpolation techniques in the experiments. The simulation revealed that the bicubic interpolation method exhibited the highest accuracy of 99.09% on MobileOCT CNN model. Regarding to the computational complexity of interpolation methods, however, the nearest neighbor interpolation method with a lower computational complexity sacrificed a small amount of accuracy by 0.23%. This evaluation result made the nearest neighbor interpolation method more advantageous for the perspective of hardware implementation.