Study on Correlation Between Subjective and Objective Metrics for Multimodal Retinal Image Registration

Abstract

Retinal imaging is crucial in diagnosing and treating retinal diseases, and multimodal retinal image registration constitutes a major advance in understanding retinal diseases. Despite the fact that many methods have been proposed for the registration task, the evaluation metrics for successful registration have not been thoroughly studied. In this paper, we present a comprehensive overview of the existing evaluation metrics for multimodal retinal image registration, and compare the similarity between the subjective grade of ophthalmologists and various objective metrics. The Pearson's correlation coefficient and the corresponding confidence interval are used to evaluate metrics similarity. It is found that the binary and soft Dice coefficient on the segmented vessel can achieve the highest correlation with the subjective grades compared to other keypoint-supervised or unsupervised metrics. The paper established an objective metric that is highly correlated with the subjective evaluation of the ophthalmologists, which has never been studied before. The experimental results would build a connection between ophthalmology and image processing literature, and the findings may provide a good insight for researchers who investigate retinal image registration, retinal image segmentation and image domain transformation.

Retinal Image Registration Using Partial Intensity Invariant Feature Descriptor and Redundant Keypoints Elimination Techniques

“…The Root Mean Square Error (RMSE) is a widely used statistical measure for assessing the accuracy of a given registration method [46]. The RMSE value is determined by quantifying the difference between the experimental matching points obtained by the registration method and the reference matching points obtained through manual calibration.…”

An Unsupervised Learning Approach to 3D Rectal MRI Volume Registration

“…Recently, deep learning methods have achieved significant improvement in runtime without sacrificing registration performance [25], [26], [28], [24], [31], [29], [9]. Vos et al proposed a coarse-to-fine pipeline that cascades an affine estimation network with B-spline estimation network based on deep learning [24].…”

“…In subjective tests, we use chessboard images, which alternatively show the patches from fixed and moving images, to evaluate the continuity between fixed and registered image [9]. We also manually label binary feature maps of Femur head and Internal obturator muscle, whose shape and size are invariant over treatment, from MRI data.…”