4.5 Article

An Adaptive MRI-PET Image Fusion Model Based on Deep Residual Learning and Self-Adaptive Total Variation

Journal

ARABIAN JOURNAL FOR SCIENCE AND ENGINEERING
Volume 47, Issue 8, Pages 10025-10042

Publisher

SPRINGER HEIDELBERG
DOI: 10.1007/s13369-020-05201-2

Keywords

MRI-PET fusion; Self-adaptive total variation; ADMM; Deep residual network; Symmetric network fusion; Regularized auto-pooling

Ask authors/readers for more resources

In this study, an adaptive MRI-PET fusion framework based on deep learning and self-adaptive total variation (DRL-SATV) is proposed and evaluated. The proposed model achieves better fusion results compared to state-of-the-art models on benchmark datasets of MR_Gad-PET, MR_T1-PET, and MR_T2-PET modalities. Various objective metrics, time and design complexity evaluations, and behavioral analyses demonstrate the superiority of the proposed model, especially in terms of free parameters, advocating its usage in clinical settings.
Multi-modal medical image fusion facilitates construction of composite images capturing complementary features from discrete images of multiple modalities. Fusion of anatomical and functional images captures structural and functional details in the fused image, which enables an integral examination of both images for a through diagnosis. In this paper, we propose an adaptive MRI-PET fusion framework based on a deep learning framework and self-adaptive total variation called DRL-SATV. The proposed model tested on MR_Gad-PET, MR_T1-PET and MR_T2-PET modalities from a benchmark dataset achieves best mutual information values of 3.6964, 3.7170 and 3.5491, respectively, compared to state-of-the-art models. Further, we have established the superiority of the proposed model with other objective metrics, time and design complexity evaluations and behavioral analyses with respect to free parameters, advocating its usage in clinical settings.

Authors

I am an author on this paper
Click your name to claim this paper and add it to your profile.

Reviews

Primary Rating

4.5
Not enough ratings

Secondary Ratings

Novelty
-
Significance
-
Scientific rigor
-
Rate this paper

Recommended

No Data Available
No Data Available