Image Processing Reference
FIGURE 1 From left to right: (a) Original MR image with severe bias field, (b) Estimated bias
field and (c) Corrected image.
Mentioned algorithm estimated bias field, corrected it, and improved the image contrast
and quality dramatically. Figure 2 indicates other examples of breast MRI.
FIGURE 2 From left to right: (a) Original MR image, (b) corrected image.
We can see that the bias field in our clinical database is 30-40% and the algorithm has re-
moved the bias field successfully.
ages are blurred because of bias field existence and breast compositions are not clearly shown.
In the images, which are normalized through mentioned algorithm, these regions are presen-
faty tissues became brighter than before and they can be separated from ibroglandular tis-
The bias field is the challenging problem of MR images. It changes the intensity values of
pixels in MRI image and corrupts these images. The correction of this problem is necessary for
subsequent computerized quantitative analysis. The high importance of bias field correction,
motivated us to use a fast, reliable, and robust algorithm to solve this problem. It is expected
that a fully automated algorithm for bias field correction in breast MRI may have great poten-
tial in clinical MRI applications.
In this chapter, an effective, robust, and accurate method is used for bias field correction
in breast MR images. We proposed a new combination of locally normalization, N3, and SFA
methods. The proposed technique takes advantage of N3 and SFA methods and not only pre-
serves simplicity, but also has the potential to be generalized to multivariate versions adapted
for segmentation applying multimodality images (e.g., T1, T2, and PD images). N3 method is
an iterative algorithm and does not need any model assumption. This method does not rely
on any prior knowledge of pathological data as a result can be applied at early stage and it is
a substantial advantage in automated data analysis. Breast MRI image as an input enters into
the proposed package and after different processing steps, the output is an image which indic-
ates breast compositions and density measurement of the breast. The efficacy of the algorithm
is presented on clinical breast MRIs and the results show the potential of method to extract
useful information for breast disease detection. Extension of this method for tumor and dis-
ease detection is the next challenging task for the future.