Magnetic Resonance Imaging
305
spontaneous neural fluctuations during diagnostic measurements. Fat tissue can be
a source of shear and stress, which momentarily result in a change in magnetic spin
orientation. This phenomenon is referred to as shear and strain noise. Rapid imaging
will put a strain on the Nyquist theorem resulting in image distortions resulting from
slow magnetic orientation response and slow sampling rate.
The main methods used for denoising of the MR images are soft and hard thresholding of the wavelet coefficients as introduced in Part I of the book. Another popular filtering method of the wavelet coefficients applies statistical methods. Filtering
of the wavelet coefficients using statistical analysis is frequently performed in two
stages. The first stage detects the presence of the signal by applying a χ 2 -test. The
second stage entails thresholding of the individual coefficients of the remaining
subbands by applying a two-tailed Z-test. Wavelet transform (WT) of f MRI signals in the time domain can selectively remove unwanted introduction of temporal
autocorrelations.
15.7.2 FEATURE EXTRACTION
Geometrical features of an object in MR images can be easily captured using measures such as area, eccentricity, compactness, and so on. Since MR images have
texture details of objects such as tumors, several texture measures as the variance
of the gray levels of the pixels within the object can be used to represent texture.
Geometrical features together with texture measures are often sufficient to identify
the majority of malignant and benign tumors from MR images.
All DFT, DCT, and DWT decomposition methods discussed in the previous chapters are used for feature extraction from MR images. DFT coefficients in high frequencies are used to capture the texture of the objects under study. Wavelet methods
are heavily used for both MRI filtering and feature extraction. The wavelet features
are often the wavelet coefficients (or the second power of the coefficients) at different
scales. These coefficients, however, need to be selected to represent the true information in the image and not the additive noise.
15.8 COMPARISON OF MRI WITH OTHER IMAGING MODALITIES
As mentioned earlier, the resolution of MRI is relatively higher than other image
modalities. In addition, MR technology allows both anatomical imaging (regular
MRI) and functional imaging (f MRI) of the biological tissues. Other technologies
often are used either for anatomical imaging or for functional imaging. Moreover,
unlike some other technologies, including x-ray CT, MRI is known to pose to harm
on the biological tissues.
Another difference between MRI and CT imaging is the fact that MRI can be performed in any arbitrary slice orientations, while CT imaging is performed in axial
direction only. This difference has an impact on image registration. For registration
purposes, the CT and MRI scan will need to be in the same field of view to assure
the best possible matches. Frequently, this can only be accomplished if the decision
to perform both CT and MRI is made in advance. MRI does not image bone, while
x-ray CT will image predominantly bone tissues.
305
spontaneous neural fluctuations during diagnostic measurements. Fat tissue can be
a source of shear and stress, which momentarily result in a change in magnetic spin
orientation. This phenomenon is referred to as shear and strain noise. Rapid imaging
will put a strain on the Nyquist theorem resulting in image distortions resulting from
slow magnetic orientation response and slow sampling rate.
The main methods used for denoising of the MR images are soft and hard thresholding of the wavelet coefficients as introduced in Part I of the book. Another popular filtering method of the wavelet coefficients applies statistical methods. Filtering
of the wavelet coefficients using statistical analysis is frequently performed in two
stages. The first stage detects the presence of the signal by applying a χ 2 -test. The
second stage entails thresholding of the individual coefficients of the remaining
subbands by applying a two-tailed Z-test. Wavelet transform (WT) of f MRI signals in the time domain can selectively remove unwanted introduction of temporal
autocorrelations.
15.7.2 FEATURE EXTRACTION
Geometrical features of an object in MR images can be easily captured using measures such as area, eccentricity, compactness, and so on. Since MR images have
texture details of objects such as tumors, several texture measures as the variance
of the gray levels of the pixels within the object can be used to represent texture.
Geometrical features together with texture measures are often sufficient to identify
the majority of malignant and benign tumors from MR images.
All DFT, DCT, and DWT decomposition methods discussed in the previous chapters are used for feature extraction from MR images. DFT coefficients in high frequencies are used to capture the texture of the objects under study. Wavelet methods
are heavily used for both MRI filtering and feature extraction. The wavelet features
are often the wavelet coefficients (or the second power of the coefficients) at different
scales. These coefficients, however, need to be selected to represent the true information in the image and not the additive noise.
15.8 COMPARISON OF MRI WITH OTHER IMAGING MODALITIES
As mentioned earlier, the resolution of MRI is relatively higher than other image
modalities. In addition, MR technology allows both anatomical imaging (regular
MRI) and functional imaging (f MRI) of the biological tissues. Other technologies
often are used either for anatomical imaging or for functional imaging. Moreover,
unlike some other technologies, including x-ray CT, MRI is known to pose to harm
on the biological tissues.
Another difference between MRI and CT imaging is the fact that MRI can be performed in any arbitrary slice orientations, while CT imaging is performed in axial
direction only. This difference has an impact on image registration. For registration
purposes, the CT and MRI scan will need to be in the same field of view to assure
the best possible matches. Frequently, this can only be accomplished if the decision
to perform both CT and MRI is made in advance. MRI does not image bone, while
x-ray CT will image predominantly bone tissues.