Pathological Myopic Image Analysis with Transfer Learning

TL;DR

This paper reviews transfer learning techniques applied to challenging myopic fundus image analysis tasks, demonstrating state-of-the-art results in classification, localization, and segmentation at a major challenge.

Contribution

It adapts popular deep learning architectures for myopic fundus image analysis, achieving top rankings in a competitive challenge.

Findings

Achieved 1st and 2nd place in several tasks at ISBI2019

Demonstrated effectiveness of transfer learning in medical image analysis

Improved accuracy in classifying and segmenting myopic fundus images

Abstract

We present a summary of transfer learning based methods for several challenging myopic fundus image analysis tasks including classification of pathological and non-pathological myopia,localisation of fovea,and segmentation of optic disc.By adapting existing popular deep learning architectures,our proposed methods have achieved 1st and 2nd place in several tasks at the Pathologic Myopia Challenge held at ISBI2019.