Hsiao Piau Ng

Medical Image Segmentation Using K-Means Clustering and Improved Watershed Algorithm

We propose a methodology that incorporates k-means and improved watershed segmentation algorithm for medical image segmentation. The use of the conventional watershed algorithm for medical image analysis is widespread because of its advantages, such as always being able to produce a complete division of the image. However, its drawbacks include over-segmentation and sensitivity to false edges. We address the drawbacks of the conventional watershed algorithm when it is applied to medical images by using k-means clustering to produce a primary segmentation of the image before we apply our improved watershed segmentation algorithm to it. The k-means clustering is an unsupervised learning algorithm, while the improved watershed segmentation algorithm makes use of automated thresholding on the gradient magnitude map and post-segmentation merging on the initial partitions to reduce the number of false edges and over-segmentation. By comparing the number of partitions in the segmentation maps of 50 images, we showed that our proposed methodology produced segmentation maps which have 92% fewer partitions than the segmentation maps produced by the conventional watershed algorithm

Medical image segmentation using watershed segmentation with texture-based region merging

The use of the watershed algorithm for image segmentation is widespread because it is able to produce a complete division of the image. However, it is susceptible to over-segmentation and in medical image segmentation, this meant that that we do not have good representations of the anatomy. We address this issue by thresholding the gradient magnitude image and performing post-segmentation merging on the initial segmentation map. The automated thresholding technique is based on the histogram of the gradient magnitude map while the post-segmentation merging is based on the similarity in textural features (namely angular second moment, contrast, entropy and inverse difference moment) belonging to two neighboring partitions. When applied to the segmentation of various facial anatomical structures from magnetic resonance (MR) images, the proposed method achieved an overlap index of 92.6% compared to manual contour tracings. It is able to merge more than 80% of the initial partitions, which indicates that a large amount of over-segmentation has been reduced. Results produced using watershed algorithm with and without the proposed and proposed post-segmentation merging are presented for comparisons.

3D Segmentation and Quantification of a Masticatory Muscle from MR Data Using Patient-Specific Models and Matching Distributions

A method is proposed for 3D segmentation and quantification of the masseter muscle from magnetic resonance (MR) images, which is often performed in pre-surgical planning and diagnosis. Because of a lack of suitable automatic techniques, a common practice is for clinicians to manually trace out all relevant regions from the image slices which is extremely time-consuming. The proposed method allows significant time savings. In the proposed method, a patient-specific masseter model is built from a test dataset after determining the dominant slices that represent the salient features of the 3D muscle shape from training datasets. Segmentation is carried out only on these slices in the test dataset, with shape-based interpolation then applied to build the patient-specific model, which serves as a coarse segmentation of the masseter. This is first refined by matching the intensity distribution within the masseter volume against the distribution estimated from the segmentations in the dominant slices, and further refined through boundary analysis where the homogeneity of the intensities of the boundary pixels is analyzed and outliers removed. It was observed that the left and right masseter muscles’ volumes in young adults (28.54 and 27.72cm3) are higher than those of older (ethnic group removed) adults (23.16 and 22.13cm3). Evaluation indicates good agreement between the segmentations and manual tracings, with average overlap indexes for the left and right masseters at 86.6% and 87.5% respectively.

Medical Image Segmentation Using Feature-based GVF Snake

We propose a feature-based GVF snake for medical image segmentation here. Feature-based criteria are introduced for the GVF snake to stop its iterations. Without these criteria, the GVF snake might continue its iterations even though it has converged at the targeted object and result in longer computational time. The feature here is the area of the targeted object. Our proposed method comprises of two stages, namely the training stage and the segmentation stage. In the training stage, we acquire prior knowledge on the relative area of the targeted object from training data. In the segmentation stage, the proposed feature-based GVF snake is applied to segment the object from the image after computing the estimated area of the targeted object. In our proposed method, the GVF snake stops its iterations when the area bounded by its propagation is approximately equal to the estimated area and when it undergoes little change over two consecutive iterations. To illustrate the effectiveness of our proposed method, we applied it to the segmentation of the masseter muscle, which is the strongest jaw muscle, from 2-D magnetic resonance (MR) images. Numerical evaluation done indicates good agreement between the computerized and manual segmentations, with mean overlap of 92%.

Quantitative analysis of human masticatory muscles using magnetic resonance imaging

OBJECTIVES The objective of this study was to quantitatively evaluate the correlation between left and right masticatory muscle volumes in normal subjects. METHODS Contiguous 1 mm MR scans were obtained of 12 normal adult subjects aged 20-25 years using a Siemens 1.5 T MR scanner. The volumes of the human masticatory muscles (masseter, lateral and medial pterygoid) were measured from the scans using our previously proposed method. To test for inter- and intraobserver reproducibility, measurements were performed by two users on two separate occasions, with a span of 2 weeks between them and with the previous results blinded. Good inter- and intraobserver reproducibility was achieved in our study. RESULTS The mean volumes for left and right masseters, and lateral and medial pterygoids were 29.54 cm3, 29.65 cm3, 9.47 cm3, 10.23 cm3, 8.69 cm3 and 8.92 cm3, respectively. The Pearson correlation coefficients between the volumes of the left and right masseters, lateral and medial pterygoids are 0.969, 0.906 and 0.924, respectively. CONCLUSIONS The computed volumes of the masticatory muscles show a strong correlation between the volumes of the left and right masseters, and lateral and medial pterygoids for normal adult subjects. The total masticatory muscle volume on the left and right sides of normal subjects is similar.

An improved shape determinative slice determination method for patient-specific modeling of facial anatomical structure

ObjectiveIn planning of maxillofacial surgeries, analysis and quantification of facial anatomical structures are carried out. At CARS 2007, we proposed a method to determine shape determinative slices, which captures the salient features of the shape of the 3D anatomical structure, to facilitate building of patient-specific models and rapid quantification. The accuracy of the built models was satisfactory. Here we propose an improved method that improves the accuracy of the built models through automatic refinement on the choice of the shape determinative slices by incorporating information from test dataset.Materials and methodsTwelve magnetic resonance imaging (MRI) datasets from adult volunteers, whose identities are anonymized, are used in this study. The earlier proposed method is used to determine the initial normalized locations of the shape determinative slices from training datasets. Given a test data, 2D automatic segmentations were performed on these initial locations and their neighboring slices. An area-based criterion is then used to refine the choice of the shape determinative slice.Results and conclusionsA total of 24 (12 left and 12 right muscles) patient-specific models were built from the shape determinative slices determined by our proposed method. The average overlap index achieved is about 87%. The models built from the shape determinative slices determined using the improved method have improvement in accuracy of up to 4.2%. The process of selecting the new shape determinative slices is automatic and the results indicate that the proposed method is effective.

Muscles of mastication model-based MR image segmentation

AbstractObjective The muscles of mastication play a major role in the orodigestive system as the principal motive force for the mandible. An algorithm for segmenting these muscles from magnetic resonance (MR) images was developed and tested. Materials and methods Anatomical information about the muscles of mastication in MR images is used to obtain the spatial relationships relating the muscle region of interest (ROI) and head ROI. A model-based technique that involves the spatial relationships between head and muscle ROIs as well as muscle templates is developed. In the segmentation stage, the muscle ROI is derived from the model. Within the muscle ROI, anisotropic diffusion is applied to smooth the texture, followed by thresholding to exclude bone and fat. The muscle template and morphological operators are employed to obtain an initial estimate of the muscle boundary, which then serves as the input contour to the gradient vector flow snake that iterates to the final segmentation. Results The method was applied to segmentation of the masseter, lateral pterygoid and medial pterygoid in 75 images. The overlap indices (κ) achieved are 91.4, 92.1 and 91.2%, respectively. Conclusion A model-based method for segmenting the muscles of mastication from MR images was developed and tested. The results show good agreement between manual and automatic segmentations.

FUZZY C-MEANS ALGORITHM WITH LOCAL THRESHOLDING FOR GRAY-SCALE IMAGES

An improved fuzzy C-means (FCM) clustering method is proposed. It incorporates Otsu thresholding with conventional FCM to reduce FCMs susceptibility to local minima, as well as its tendency to derive a threshold that is biased towards the component with larger probability, and derive threshold values with greater accuracy. Thresholding is performed at the cluster boundary region in feature space. A comparison of the results produced by improved and conventional algorithms confirms the superior performance of the former.

Template-based Automatic Segmentation of Masseter Using Prior Knowledge

In this paper, we propose a knowledge-based, fully automatic methodology for segmenting the masseter, which is a muscle of mastication, from 2-D magnetic resonance (MR) images for clinical purposes. To our knowledge, there is currently no methodology which automatically segments the masseter from MR images. Our methodology uses five ground truths, where the masseter has been manually segmented and verified by medical experts, to serve as the reference and provide prior knowledge. The prior knowledge involved is the spatial relationship between the region of interest (ROI) of the head and ROI of the masseter. In the segmentation process, anisotropic diffusion first smoothens the ROI of the latter, and thresholding removes unwanted neighboring regions of the masseter. A template of the masseter is then used to obtain an initial segmentation of the muscle, which serves as the initialization to the gradient vector flow (GVF) snake for refining the initial segmentation. We performed 2-D segmentation of the masseter on a total of 25 MR images, which belong to the mid-facial region through the mandible from five data sets. Validation was done by comparing the segmentation results obtained by using our proposed methodology against manual segmentations done by medical experts, obtaining an average accuracy of 92%

Knowledge-driven 3-D Extraction of the Masseter from MR Data

In this paper, we propose a knowledge-driven highly automatic methodology for extracting the masseter from magnetic resonance (MR) data sets for clinical purposes. The masseter is a muscle of mastication which acts to raise the jaw and clench the teeth. In our initial work, we designed a process which allowed us to perform 2-D segmentation of the masseter on 2-D MR images. In the methodology proposed here, we make use of ground truth to first determine the index of the MR slice in which we will carry out 2-D segmentation of the masseter. Having obtained the 2-D segmentation, we will make use of it to determine the region of interest (ROI) of the masseter in the other MR slices belonging to the same data set. The upper and lower thresholds applied to these MR slices, for extraction of the masseter, are determined through the histogram of the 2-D segmented masseter. Visualization of the 3-D masseter is achieved via volume rendering. Our methodology has been applied to five MR data sets. Validation was done by comparing the segmentation results obtained by using our proposed methodology against manual contour tracings, obtaining an average accuracy of 83.5%