Pattern Recognition Letters
A parametric gradient descent MRI intensity inhomogeneity correction algorithm
Pattern Recognition Letters
| Hi-index | 0.10 |
The paper proposes an automatic algorithm for the fat- and muscle-tissue delineation in the Magnetic Resonance Image data of the patient's leg with FacioScapuloHumeral muscular Dystrophy. The algorithm corrects the tissue inhomogeneity with a novel method that produces good results with low computation time and complexity. The estimated bias field is modelled as a multiplicative noise and uses low-pass filtering to obtain smoothness of the form. To reduce the impact of the background low-level intensity on the object high-level intensity, the background is remodified. The inhomogeneity correction method is validated by comparing its results with those of a simulated ground-truth image. In the segmentation procedure, fuzzy c-mean clustering is used. The segmentation results of the automatic algorithm are comparable to the medical-specialist annotations with a similarity index above 0.91, indicating an excellent result of the proposed automatic processing.