Brain MR image segmentation and bias field estimation using coherent local and non-local spatial constraints

Novel fuzzy clustering‐based bias field correction technique for brain magnetic resonance images

“…Further, the proposed technique is evaluated with one volume of clinical brain MR images with 21 selected slices from the IBSR database [24]. The results are compared with our implementation of MFCM [12], CLIC [16], BFELGI [20] and N3KHM [21] models. The suggested technique is validated using qualitative analysis.…”

“…7 shows the coronal cross‐section of the PD‐weighted brain MR image at 1 mm slice thickness with 5% noise and 20% IIH. Here, it shows the tissue only region of the PD‐weighted brain MR image and segmentation results with bias field correction using the state‐of‐the‐art algorithms (MFCM [12], CLIC [16], BFELGI [20], N3KHM [21]) and the proposed BFC technique, respectively.…”

“…The results in Fig. 9 show the histograms of the input image and the bias field corrected brain MR images using MFCM [12], CLIC [16], BFELGI [20], N3KHM [21] and the proposed model, respectively. The histogram of each brain MR image should contain four distinct peak areas for three tissue regions and the background.…”

“…Such an approach completely avoids the need for modelling the intensity inhomogeneity function separately. In recent years, non‐parametric and FCM‐based clustering methods are proposed [4–22].…”

“…The spatial factor is estimated from the local intensity probabilities and is used for bias field estimation. Shi et al [16] introduced the non‐local spatial constraint into coherent local intensity clustering (CLIC) for bias field estimation of the tissue regions. The smoothness of the bias field estimation is ensured by coherent local constraints.…”