Page 53 - ARNM-1-2

Advances in Radiotherapy
& Nuclear Medicine Image fusion’s significance in medicine

Section 2 mainly introduces the application of traditional They employed two different fusion rules based on a non-
image fusion methods in the field of NMMI. The studies parametric density model and variable weighting theory
reporting on the application of traditional image fusion for the fusion of low-frequency and high-frequency
methods in NMMI are provided in Table 1. A comparison coefficients. The fusion image is constructed by applying the
of different image fusion methods is provided in Table S1. inverse non-subsampled contourlet transform operation
to all composite coefficients. Haddadpour et al. used
[10]
2.1. Spatial fusion
MRI and PET as input images and fused them based on the
Spatial fusion methods mainly rely on pixel-level fusion, combination of two-dimensional Hilbert transform (2-D
directly merging pixels from different images. These HT) and intensity-hue-saturation (IHS) method, which
methods include simple average, weighted average, preserves both spatial and spectral features of input images.
maximum value, minimum value, and high-pass filtering . Stokking et al. proposed a hue-saturation-value (HSV)
[8]
[11]
Although these methods are simple and easy to implement, model for fusing anatomical and functional information
they usually require pre-processing and post-processing obtained from MRI and SPECT modes using a color
and can easily cause spatial distortion in the fused image. coding scheme. This model outperforms the RGB model
Liu et al. proposed a new image fusion method based and allows for quick, simple, and intuitive retrospective
[9]
on a multi-resolution and non-parametric density model. determination of color coding in the fused image. Chen
[12]
Table 1. Traditional image fusion methods in NMMI
Authors Year Fusion methods Multimodal Fusion Techniques Organ
images
Liu et al. [9] 2019 Spatial fusion PET-MRI Non-parametric density model and variable Brain
weighting theory
Haddadpour 2017 Spatial fusion PET-MRI Two-dimensional Hilbert transform (2-D HT) Brain
et al. [10] and intensity-hue-saturation (IHS) method
Chen [12] 2017 Spatial fusion PET-MRI Combined the IHS model with Log-Gabor Brain
transform
He et al. [41] 2010 Spatial fusion PET-MRI IHS and principal component analysis (PCA) Brain
Liu et al. [14] 2010 Frequency fusion PET-CT Multiwavelet transform Lung
Xiong et al. [42] 2017 Frequency fusion PET-CT Shift-invariant Shearlet Transform (SIST) Brain
and adaptive Pulse coupled neural network
(PCNN)
Bhavana and 2015 Frequency fusion PET-MRI Discrete Wavelet Transform (DWT) Brain
Krishnappa [43]
Wang et al. [44] 2006 Frequency fusion PET-MRI DWT Brain
Du et al. [45] 2018 Frequency fusion PET-CT Parallel significant features Brain
Shabanzade et al. [18] 2019 Decision-level PET-MRI Non-parametric Bayesian technique Brain
fusion
Zong and Qiu [20] 2017 sparse SPECT-MRI, Ssparse representation of classified image Brain, lung
representation PET-CT, patches
fusion MRI-CT
Shahdoosti and 2018 Sparse SPECT-MRI, Tetrolet transform Brain
Mehrabi [46] representation CT-MRI,
fusion PET-MRI
Zhu et al. [21] 2018 Hybrid fusion PET-MRI, Spatial method for cartoon component and Brain
SPECT-MRI sparse representation method for texture
components
Chaitanya et al. [47] 2017 Hybrid fusion PET-MRI Shearlet transformation and discrete cosine Brain
transform
Daneshvar and 2010 Hybrid fusion PET-MRI IHS and the retina-inspired model (RIM) Brain
Ghassemian [22] fusion technique
Abbreviations: CT: Computed tomography; MRI: Magnetic resonance imaging; PET: Positron emission tomography; SPECT: Single-photon emission
computed tomography.

Volume 1 Issue 2 (2023) 3 https://doi.org/10.36922/arnm.0870