No reference quality evaluation of medical image fusion

Abstract

Medical image fusion (MIF) attracts much attention in clinical use. Many MIF algorithms have been proposed over the past decade. Existing MIF algorithms create different fused image, however, current quality evaluation method is not designed for MIF images. So, we present a no reference quality evaluation of medical image fusion. Firstly, a MIF image database (MIFID) is built, and radiologist ratings are selected to conduct subjective test. Then an objective quality evaluation metric of medical image fusion is proposed via the phase congruency and standard deviation. Image salient features and image information are very important for visual quality of fused image. Based on this consideration, we combine the two existing quality evaluation metrics to assess MIF images. Finally, five comparative study experiments are implemented based on the MIFID. Experimental results reveal that the proposed quality evaluation metric is superior to the existing state-of-the-art metrics, which is more applicable to evaluate MIF images. K E Y W O R D S medical image fusion, no reference quality evaluation, phase congruency, standard deviation Int J Imaging Syst Technol. 2018;1-7. wileyonlinelibrary.com/journal/ima V C 2017 Wiley Periodicals, Inc. | 1

Exploiting Quality-Guided Adaptive Optimization for Fusing Multimodal Medical Images

“…The image objective evaluation index can measure the quality of the image and can be used as the basis for the fitness function selection. In addition, the Q MMIF is experimentally proved to have more accurately than existing evaluation strategies in evaluating the MMIF image [17]. To the best of our knowledge, this is the first time that the Q MMIF model is used as quality-guided adaptive optimization in the field of medical image fusion.…”

“…The choice of fitness function is determined by experiments. In implementation, the fitness function is constructed by the standard deviation (SD) [28], the normalized mutual information (MI) [33], Xydeas et al's gradient based metric Q G [34], Yang et al's metric (Q Y ) [35], and Q MMIF [17], respectively, in MMIF image database [17]. We find the results of Q MMIF in objective evaluation metrics are best.…”

“…We introduce a quality evaluation metric for MMIF (Q MMIF ) [17] as the quality-guided fitness function.…”

“…Many image quality assessment approaches have been presented [22]- [24], however these existing evaluation criterias are not designed for MMIF, which limit in evaluating multimodal fused images, and little has been done to compare them with subjective data that contains a wide variety of image modalities and fusion algorithms. In our previous work [17], a MMIF image database is built, 20 radiologists participated in the subjective test, the proposed quality evaluation metric for MMIF (Q MMIF ) can maintain good consistency with the subjective experimental results, and can more accurately evaluate the MMIF image. In this paper, the fitness function is constructed by the Q MMIF evaluation criterion, which advances fruit flies to fly from the original location to the best location with higher effectiveness.…”