Comparing Conventional and Deep Feature Models for Classifying Fundus Photography of Hemorrhages

TL;DR

This study compares conventional and deep feature extraction methods for classifying hemorrhages in fundus photographs, finding deep models generally outperform traditional features in accuracy.

Contribution

It introduces a novel hemorrhage segmentation technique and evaluates the effectiveness of deep versus conventional features for classification.

Findings

Deep models outperform conventional features in classification accuracy.

Image preprocessing improves hemorrhage detection.

Deep features show higher robustness in challenging cases.

Abstract

Diabetic retinopathy is an eye-related pathology creating abnormalities and causing visual impairment, proper treatment of which requires identifying irregularities. This research uses a hemorrhage detection method and compares classification of conventional and deep features. Especially, method identifies hemorrhage connected with blood vessels or reside at retinal border and reported challenging. Initially, adaptive brightness adjustment and contrast enhancement rectify degraded images. Prospective locations of hemorrhages are estimated by a Gaussian matched filter, entropy thresholding, and morphological operation. Hemorrhages are segmented by a novel technique based on regional variance of intensities. Features are then extracted by conventional methods and deep models for training support vector machines, and results evaluated. Evaluation metrics for each model are promising, but findings suggest that comparatively, deep models are more effective than conventional features.