A Segmentation-Oriented Inter-Class Transfer Method: Application to Retinal Vessel Segmentation

TL;DR

This paper introduces a patch-based transfer method for retinal vessel segmentation that leverages semi-supervised clustering and information bottleneck theory, achieving high accuracy with limited labeled data.

Contribution

It proposes a novel two-stage transfer framework with a task-specific feature space and instance selection, improving segmentation performance over existing methods.

Findings

Achieved over 97% accuracy on DRIVE dataset.

Outperformed human observers on DRIVE and STARE datasets.

Demonstrated that cross-class similar images enhance segmentation performance.

Abstract

Retinal vessel segmentation, as a principal nonintrusive diagnose method for ophthalmology diseases or diabetics, suffers from data scarcity due to requiring pixel-wise labels. In this paper, we proposed a convenient patch-based two-stage transfer method. First, based on the information bottleneck theory, we insert one dimensionality-reduced layer for task-specific feature space. Next, the semi-supervised clustering is conducted to select instances, from different sources databases, possessing similarities in the feature space. Surprisingly, we empirically demonstrate that images from different classes possessing similarities contribute to better performance than some same-class instances. The proposed framework achieved an accuracy of 97%, 96.8%, and 96.77% on DRIVE, STARE, and HRF respectively, outperforming current methods and independent human observers (DRIVE (96.37%) and STARE (93.39%)).