Description:FIELD OF THE INVENTION
The present invention relates to medical image processing and, more specifically, to an improved segmentation technique for spinal cord images obtained through computed tomography (CT). The invention enhances the accuracy of spinal cord segmentation by utilizing a Kernel-based Fuzzy C-Means (KFCM) clustering algorithm, effectively detecting abnormalities, and aiding in medical diagnosis.
BACKGROUND OF THE INVENTION
The lumbar spine from computed tomography images is segmented in axial view and sagittal view. Low back pain in human being is due to the abnormality in lumbar spine.
Clustering type segmentation was used in the existing systems. Fuzzy C means (FCM) can specifically be used for medical images to detect abnormality.
Existing models segment the spine either in axial view or sagittal view. The systems will not be applicable for both views.
The segmentation method segments and highlights the edges and internal organs of the spine. Accuracy is proved to be better.
Hard clustering divides an image into a specified number of subdivisions. Therefore the pixel may or may not belong to a subset. FCM is included in soft clustering which allows the pixel to belong to many subdivisions of the image simultaneously.
SUMMARY OF THE INVENTION
This summary is provided to introduce a selection of concepts, in a simplified format, that are further described in the detailed description of the invention.
This summary is neither intended to identify key or essential inventive concepts of the invention and nor is it intended for determining the scope of the invention.
The invention presents a novel segmentation framework that integrates kernel-based clustering with fuzzy logic for spinal cord segmentation on CT images. The method classifies spinal tissues, including the vertebral body, intervertebral discs, and spinal canal, by computing fuzzy membership values for each pixel. The iterative calculation ensures the convergence of the objective function, leading to precise segmentation results.
Unlike traditional FCM, the proposed KFCM incorporates spatial contextual information by introducing a controlling parameter. This parameter adjusts the influence of neighboring pixels, reducing the impact of noise and enhancing segmentation accuracy. The KFCM algorithm dynamically modifies membership functions based on local pixel variations, ensuring robustness in differentiating spinal structures.
The invention also integrates Gaussian Radial Basis Function (GRBF) as the kernel function, improving the ability to handle non-linear structures within spinal images. The selection of kernel width plays a crucial role in segmentation performance, balancing sensitivity and boundary detection efficiency.
Furthermore, the proposed segmentation method is applicable to both healthy and abnormal spinal structures, facilitating automated identification of degenerative changes. Comparative evaluation with existing methods demonstrates superior accuracy, particularly in delineating vertebrae and intervertebral discs.
The KFCM algorithm is computationally optimized to enhance processing efficiency, making it suitable for real-time medical applications. By integrating this segmentation framework into clinical imaging systems, the invention provides an effective tool for spinal cord analysis, improving diagnostic precision and patient outcomes.
To further clarify advantages and features of the present invention, a more particular description of the invention will be rendered by reference to specific embodiments thereof, which is illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the invention and are therefore not to be considered limiting of its scope. The invention will be described and explained with additional specificity and detail with the accompanying drawings.
The tissue types are classified on the basis of calculating membership values of all pixels in an image. Tissue classification in medical image processing is the important process in the quantification from tissue volumes, detection of pathology and surgery. The FCM clustering algorithm is taken to segment the three types of tissues i.e., vertebral body, disc and spinal canal present in the spinal cord. The membership values are calculated iteratively for getting the smallest value or convergence of objective function of FCM in consideration with membership values and cluster centroids.
BRIEF DESCRIPTION OF THE DRAWINGS
The illustrated embodiments of the subject matter will be understood by reference to the drawings, wherein like parts are designated by like numerals throughout. The following description is intended only by way of example, and simply illustrates certain selected embodiments of devices, systems, and methods that are consistent with the subject matter as claimed herein, wherein:
Figure 1 KFCM segmentation on CT healthy human vertebra in L1 from axial view
Figure 2 KFCM method of segmentation of CT healthy human disc lying between L1-L2 vertebra from axial view
Figure 3: KFCM method of segmentation of CT healthy human image taken in sagittal view
Figure 4: Segmentation on CT abnormal vertebra L5 (axial view) using KFCM
Figure 5: KFCM method of segmentation of CT abnormal disc L4-L5 taken in axial view
Figure 6: Segmentation of CT abnormal image in sagittal view using KFCM
The figures depict embodiments of the present subject matter for the purposes of illustration only. A person skilled in the art will easily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles of the disclosure described herein.
DETAILED DESCRIPTION OF THE INVENTION
The detailed description of various exemplary embodiments of the disclosure is described herein with reference to the accompanying drawings. It should be noted that the embodiments are described herein in such details as to clearly communicate the disclosure. However, the amount of details provided herein is not intended to limit the anticipated variations of embodiments; on the contrary, the intention is to cover all modifications, equivalents, and alternatives falling within the scope of the present disclosure as defined by the appended claims.
It is also to be understood that various arrangements may be devised that, although not explicitly described or shown herein, embody the principles of the present disclosure. Moreover, all statements herein reciting principles, aspects, and embodiments of the present disclosure, as well as specific examples, are intended to encompass equivalents thereof.
The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of example embodiments. As used herein, the singular forms “a",” “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises,” “comprising,” “includes” and/or “including,” when used herein, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components and/or groups thereof.
It should also be noted that in some alternative implementations, the functions/acts noted may occur out of the order noted in the figures. For example, two figures shown in succession may, in fact, be executed concurrently or may sometimes be executed in the reverse order, depending upon the functionality/acts involved.
In addition, the descriptions of "first", "second", “third”, and the like in the present invention are used for the purpose of description only, and are not to be construed as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Thus, features defining "first" and "second" may include at least one of the features, either explicitly or implicitly.
Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which example embodiments belong. It will be further understood that terms, e.g., those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.
The invention utilizes a robust KFCM-based segmentation technique that improves the classification of spinal tissues in CT images. The methodology involves multiple processing stages, including image preprocessing, feature extraction, fuzzy clustering, and post-segmentation refinement.
Image preprocessing involves contrast enhancement and noise reduction to improve segmentation accuracy. Gaussian smoothing is applied to suppress high-frequency noise while preserving anatomical details. Normalization techniques standardize intensity variations across different CT scans, ensuring consistency in feature extraction.
The segmentation process begins by computing fuzzy membership values for each pixel based on grayscale intensities and neighborhood information. The objective function of FCM is iteratively optimized to minimize segmentation errors. The kernel function replaces conventional Euclidean distance, enhancing cluster separation in complex spinal structures.
A novel spatial parameter is introduced, dynamically adjusting cluster membership based on pixel neighborhood variations. This approach reduces noise sensitivity and improves boundary delineation. The parameter is computed using the local variation coefficient (LVC), which quantifies grayscale heterogeneity within a specified neighborhood.
To ensure robustness, the segmentation framework integrates adaptive kernel width selection, preventing excessive smoothing of fine details. The Gaussian Radial Basis Function (GRBF) is employed to improve non-linear class separability, effectively distinguishing between vertebrae, intervertebral discs, and spinal canal regions.
The proposed method is evaluated using a dataset of healthy and abnormal spinal images, demonstrating high segmentation accuracy compared to existing methods. Qualitative assessments confirm the effectiveness of KFCM in detecting vertebral boundaries and degenerative changes, supporting automated clinical diagnosis.
The computational efficiency of the algorithm is optimized through parallel processing techniques, enabling real-time segmentation. The invention is designed for integration into clinical imaging platforms, enhancing radiological assessments and aiding in pre-surgical planning.
The tissue types are classified based on calculating membership values of all pixels in an image. Tissue classification in medical image processing is the important process in the quantification from tissue volumes, detection of pathology and surgery. The FCM clustering algorithm is taken to segment the three types of tissues i.e., vertebral body, disc, and spinal canal present in the spinal cord. The membership values are calculated iteratively for getting the smallest value or convergence of objective function of FCM in consideration with membership values and cluster centroids.
KFCM segmentation algorithms for spinal cord images has been proposed. Segmentation is the image processing step to divide an image into several meaningful regions of similar characteristics. Segmentation step for spinal cord CT images is necessary for differentiating lamina and spinal cord. The segmentation process is necessary to identify the changes in the structure of internal organs.
Normally the belongingness to the subset is represented in terms of fuzzy membership function. The membership function provides an idea about the presence of noise and the partial volume effect which often occurs in tissues. The degree of fuzziness for any pixel is greater than one if it belongs to a subset completely. The membership value not equal to two is not considered in FCM. For the pixel lying on the boundary between two subsets, the membership value varies from zero to one. This is partial belongingness and this would give the best result for the overlapping pixel sets. The tissue types are classified based on calculating membership values of all pixels in an image. Tissue classification in medical image processing is the important process in the quantification from tissue volumes, detection of pathology and surgery. The FCM clustering algorithm is taken to segment the three types of tissues i.e., vertebral body, disc and spinal canal present in the spinal cord. The membership values are calculated iteratively for getting the smallest value or convergence of objective function of FCM in consideration with membership values and cluster centroids.
FCM algorithm does not take the spatial information into consideration. Therefore, spatial FCM (SFCM) algorithm is adopted and the controlling parameter is added with the objective function equation of FCM.
The controlling parameter 𝛼 is used to control the spatial information from the neighboring pixels. This parameter is set fixed and ranges from zero to one. The SFCM algorithm is complex as calculations are done iteratively. So, in the proposed algorithm, the controlling parameter is appended with the term replacing the pixels' intensity with weighted image.
The value of 𝛼 is set manually in advance with care for every pixel to have a control on the amount of information in the presence of noise. But it requires the knowledge of noise information. The amount of noise present in windows will be different. So, depending on the pixel the calculation of 𝛼 varies and a regularization parameter pi is introduced. The parameter is made variable to the noise amount present in the pixels which are to be processed.
The local variation coefficient (LVC) is determined for knowing the deviations of grayscales within all windows which is are required be normalized in accordance with the average value of local grayscale. LVC will increase for more heterogeneity among central pixel and the neighbors if noise is present.
The pixels of high LVC will get higher values by the parameter pi and the pixels of low LVC will get lower values by pi. piwill be equal to 2+ wi when ith pixel is bright comparing to the neighbors’ average. wi is high for larger LVC. When the average grayscale is same as central pixel grayscale, pi becomes zero and the method will be equivalent to the FCM method.
pi is relevant to the grayscales within a specified neighborhood. pi is irrelevant to clustering measurements. So, pi is calculated before the clustering starts. But, in spatial FCM, the contextual weights are updated at each iteration.
If a segmentation type does not consider the inhomogeneous size and gray level of the region, it may have problems and fail. Therefore, a segmentation type is developed to consider the inhomogeneous size and gray level. At the same time the segmentation is not hampered by the inhomogeneities.
pi gives the contextual information depending on the grayscale heterogeneity of neighborhood. FCM gives the contextual information depending on the grayscale heterogeneity of neighborhood and cluster centers.
Homogeneous clustering is obtained by the proposed parameter pi. But FCM method makes the clustering to have homogeneous labels.
The calculation of Euclidean distance is simple. It is sensitive to perturbations and boundary. The kernel functions can place the data in higher dimensional space for easy separation. Nonlinear algorithm is obtained through the transformation from linear one using kernel with the help of dot product. The Euclidean distance can be replaced by kernel function K.
The kernel width for Gaussian Radial Basis Function (GRBF) is written as σ. GRBF is considered for increasing the accuracy of segmentation.
Choice of the kernel width σ requires great care. For a larger width, linear exponential effect will be reached and for a smaller width, the cluster boundaries will become sensitive to data lying outside the edges. σ was equal to 150 while in other experiments sample variance was used to estimate 𝜎. σ is calculated using the distance variances between pixels.
This method is kernel based FCM or KFCM.
Figure 2 narrates the segmented result on CT healthy human disc lying between L1-L2 taken in axial view from KFCM. This method detects the lamina.
Figure 3 depicts the segmented result on CT healthy human vertebrae (sagittal view) from KFCM. The KFCM method detects the vertebrae L1-L5. The smoothness of vertebra’s shape is guaranteed. While comparing the image obtained with the proposed KFCM algorithm displayed in Figure 3 with the original image, the visual improvement will be noticed.
Figure 4 shows the segmented result of CT abnormal vertebra L5 (axial view) from KFCM. The KFCM method detects the vertebra boundary with edge details. It is clearly identified that the details are well preserved and the perceived quality is improved especially in the edges in the case of KFCM.
Figure 5 explains the segmented result on CT abnormal disc L4-L5 (axial view) from KFCM. This method detects the lamina in the presence of discontinuity.
The segmented result of lumbar CT abnormal vertebrae (sagittal view) using the proposed KFCM method is presented in Figure 6. The KFCM method detects the vertebrae. The abnormal disc spaces are seen brighter. It is inferred that the visual quality is improved in the edges of the image.
The algorithm clearly detects the internal organs effectively so that accuracy is more.
FCM performs well when applied in spinal vertebra (in axial view), spinal disc (in axial view) and the full spine (in sagittal view).
, Claims:1. A method for spinal cord segmentation in computed tomography images comprising:
a) Preprocessing of CT images to enhance contrast and reduce noise;
b) Applying a Kernel-based Fuzzy C-Means (KFCM) clustering algorithm for tissue classification;
c) Integrating a spatial controlling parameter to refine segmentation accuracy;
d) Utilizing Gaussian Radial Basis Function (GRBF) as the kernel function for improved class separability;
e) Computing local variation coefficients (LVC) to adjust cluster memberships dynamically.
2. The method as claimed in claim 1, wherein the segmentation process is applicable to both axial and sagittal views of the spinal cord.
3. The method as claimed in claim 1, wherein the spatial controlling parameter adjusts membership functions based on neighborhood pixel variations.
4. The method as claimed in claim 1, wherein adaptive kernel width selection is employed to optimize segmentation performance.
5. The method as claimed in claim 1, wherein the system is optimized for real-time segmentation in clinical imaging platforms.
6. The method as claimed in claim 1, wherein post-segmentation refinement techniques enhance boundary delineation of spinal structures.
7. The method as claimed in claim 1, wherein the segmentation framework improves diagnostic accuracy for detecting degenerative spinal conditions.
8. The method as claimed in claim 1, wherein the algorithm integrates parallel processing for enhanced computational efficiency.
9. The method as claimed in claim 1, wherein quantitative evaluation demonstrates superior segmentation performance compared to existing techniques.
10. The method as claimed in claim 1, wherein the framework is adaptable for automated radiological assessments and pre-surgical planning.