FIRE: Fundus Image Registration dataset

Abstract

Purpose: Retinal image registration is a useful tool for medical professionals. However, performance evaluation of registration methods has not been consistently assessed in the literature. To address that, a dataset comprised of retinal image pairs annotated with ground truth and an evaluation protocol for registration methods is proposed.Methods: The dataset is comprised by 134 retinal fundus image pairs. These pairs are classified into three categories, according to characteristics that are relevant to indicative registration applications. Such characteristics are the degree of overlap between images and the presence/absence of anatomical differences. Ground truth in the form of corresponding image points and a protocol to evaluate registration performance are provided.Results: The proposed protocol is shown to enable quantitative and comparative evaluation of retinal registration methods under a variety of conditions.Conclusion: This work enables the fair comparison of retinal registration methods. It also helps researchers to select the registration method that is most appropriate given a specific target use.

A lossless compression method for multi-component medical images based on big data mining

“…Table 2 illustrates the experimental results of soft compression algorithm for multi-component images and other classical systems on Malaria, BCCD, Melanoma and FIRE 54 datasets. The statistics include mean, minimum, maximum and variance about compression ratio.…”

Fundus Image Registration Technique Based on Local Feature of Retinal Vessels

“…The registration performance when utilizing CURVE feature points in the featurebased RIR technique was demonstrated on the FIRE dataset. CURVE was paired with the SIFT descriptor [41], and the registration performance of CURVE-SIFT was compared with five existing feature-based RIR techniques; GDB-ICP [13], Harris-PIIFD [10], Ghassabi's-SIFT [8], H-M 16 [16], H-M 17 [9] and D-Saddle-HOG [14]. Overall, CURVE-SIFT successfully registered 44.030% of the image pairs in the FIRE dataset, while the success rate of the existing feature-based RIR techniques is less than 27.612%.…”

“…s initial is set by referring to the size of the initial Gaussian image G 0,−1 . These values are determined by observing the retinal vessels with the thickest width on the fundus images from five datasets; CHASE_DB1 [33,34], DRIVE [35,36], HRF [37,38], STARE [39,40] and Fundus Image Registration (FIRE) dataset [41]. Furthermore, by considering scale or zoom less than 1.5 [8].…”

“…The details of CHASE_DB1, DRIVE, HRF and STARE datasets are described in Table 2. The registration performance of CURVE-SIFT is evaluated in the Fundus Image Registration dataset (FIRE) [41]. This dataset is the only public fundus image registration dataset with ground truth annotation.…”

DRMIME: Differentiable Mutual Information and Matrix Exponential for Multi-Resolution Image Registration

“…We refer to our proposed method as DRMIME (differentiable registration with mutual information and matrix exponential). DRMIME is able to achieve state-of-the-art accuracy on two benchmark data sets: FIRE [4] and ANHIR [5].…”