Abstract:Objective Automated brain magnetic resonance (MR) image segmentation is a complex problem especially if accompanied
by intensity inhomogeneity and noise. This paper proposed a modified fuzzy c-mean (MFCM) method which is used for the automatic
segmentation of brain MR imaging. Methods The algorithm begined with a preprocessing step where we implemented automatic bias
removal and contrast enhancement. This was followed by automated retrieval of mean intensity positions of various tissues detected.
The corrected image was then passed on to an MFCM. The segmented result was further passed through neighborhood-based
membership ambiguity correction which smooths the ambiguous boundaries and also removes pixel level noise between continuous
regions of intensities. Results Brain Web normal brain simulated database with noise ranging from 0–9% and the inhomogeneity
was 0 and 40% respectively. Qualitative evaluation results showed that the proposed method could provide clearer boundaries than
that without pre- and post-processing. Quantitative evaluation results indicated that the improved active contour algorithm generated
a higher degree of sensitivity, specificity and similarity than traditional FCM and FSL library tool-based software. Conclusion The
proposed algorithm is of fully automatic segmentation, with faster computation and faster convergence of the objective function,
which makes it be a feasible method for automatic brain MR segmentation.