Journal of Medical Physics
 Home | Search | Ahead of print | Current Issue | Archives | Instructions | Subscription | Login  The official journal of Association of Medical Physicists of India      
 Users online: 336  Home  EMail this page Print this page Decrease font size Default font size Increase font size 


 
REVIEW ARTICLE
Year : 2010  |  Volume : 35  |  Issue : 1  |  Page : 3-14
 

Automated medical image segmentation techniques


1 School of Biomedical Engineering, Institute of Technology, Institute of Medical Sciences, Banaras Hindu University, Varanasi-221 005, UP, India
2 Department of Radiotherapy and Radiation Medicine, Institute of Medical Sciences, Banaras Hindu University, Varanasi-221 005, UP, India

Date of Submission25-Apr-2009
Date of Decision15-Jul-2009
Date of Acceptance24-Aug-2009
Date of Web Publication7-Jan-2010

Correspondence Address:
Neeraj Sharma
School of Biomedical Engineering, Institute of Technology, Banaras Hindu University, Varanasi-221 005, UP
India
Login to access the Email id

Source of Support: None, Conflict of Interest: None


DOI: 10.4103/0971-6203.58777

Rights and Permissions

 

   Abstract 

Accurate segmentation of medical images is a key step in contouring during radiotherapy planning. Computed topography (CT) and Magnetic resonance (MR) imaging are the most widely used radiographic techniques in diagnosis, clinical studies and treatment planning. This review provides details of automated segmentation methods, specifically discussed in the context of CT and MR images. The motive is to discuss the problems encountered in segmentation of CT and MR images, and the relative merits and limitations of methods currently available for segmentation of medical images.


Keywords: Artificial intelligence techniques, computed tomography, magnetic resonance imaging, medical images artifacts, segmentation


How to cite this article:
Sharma N, Aggarwal LM. Automated medical image segmentation techniques. J Med Phys 2010;35:3-14

How to cite this URL:
Sharma N, Aggarwal LM. Automated medical image segmentation techniques. J Med Phys [serial online] 2010 [cited 2017 Jul 22];35:3-14. Available from: http://www.jmp.org.in/text.asp?2010/35/1/3/58777



   Introduction Top


With increasing use of Computed topography (CT) and Magnetic resonance (MR) imaging for diagnosis, treatment planning and clinical studies, it has become almost compulsory to use computers to assist radiological experts in clinical diagnosis, treatment planning. Reliable algorithms are required for the delineation of anatomical structures and other regions of interest (ROI). The goals of computer-aided diagnosis (CAD) are:

  • To automate the process so that large number of cases can be handled with the same accuracy i.e. the results are not affected as a result of fatigue, data overload or missing manual steps.
  • To achieve fast and accurate results. Very high-speed computers are, now, available at modest costs, speeding up computer-based processing in the medical field.
  • To support faster communication, wherein patient care can be extended to remote areas using information technology.
The techniques available for segmentation of medical images are specific to application, imaging modality and type of body part to be studied. For example, requirements of brain segmentation are different from those of thorax. The artifacts, which affect the brain image, are different - partial volume effect is more prominent in brain while in the thorax region it is motion artifact which is more prominent. Thus while selecting a segmentation algorithm one is required to consider all these aspects. The problems common to both CT and MR medical images are:

  • Partial volume effect
  • Different artifacts: example motion artifacts, ring artifacts, etc and
  • Noise due to sensors and related electronic system.
There is no universal algorithm for segmentation of every medical image. Each imaging system has its own specific limitations. For example, in MR imaging (MRI) one has to take care of bias field noise (intensity in-homogeneities in the RF field). Of course, some methods are more general as compared to specialized algorithms and can be applied to a wider range of data. A brief survey of three generations of medical image segmentation techniques has been provided by D.J. Withey and Z.J. Koles. [1]


   Medical Imaging Modalities Top


Medical imaging is performed in various modalities, such as MRI, CT, ultrasound, positron emission tomography (PET), etc. In the present review, we are focusing primarily on the segmentation of MR and CT images only.

Magnetic Resonance Imaging : MR imaging is the most widely used technique in the field of radio imaging. [2],[3] MR is a dynamic and flexible technology that allows achieving variable image contrast by using different pulse sequences and by changing the imaging parameters corresponding to longitudinal relaxation time (T1), and transverse relaxation time (T2), and signal intensities on T1 and T2 weighted images relate to specific tissue characteristics. The contrast on MR image is a factor dependent on pulse sequence parameters. The most common pulse sequences are T1- weighted and T2-weighted spin-echo sequences. MR imaging of the body is performed to get the structural details of brain, liver, chest, abdomen and pelvis which helps in diagnosis or monitoring the treatment.

Brain MR Imaging: MR is generally more sensitive in detecting brain abnormalities during the early stages of disease, and is excellent in early detection of cases of cerebral infarction, brain tumors, or infections. MR is particularly useful in detecting white matter disease, such as multiple sclerosis, progressive multifocal leukoencephalopathy, leukodystrophy, and post-infectious encephalitis. In contrast, CT scan fails to detect white matter abnormalities. In case of MR images of the brain, the primary determinants of signal intensity and contrast are the T1 and T2 relaxation times. The contrast is distinctly different on T1 and T2-weighted images. Also, brain pathologies have some common signal characteristics. Pathologic lesions can be separated into five major groups by their specific signal characteristics on the two basic images: T2- weighted, and T1-weighted [Table 1].

MR Liver Imaging: MR provides outstanding intrinsic soft contrast that can enhance subtle differences between normal and pathologic tissues and tissues of different histologic subtypes. Nonionizing radiation is used and MRI contrast agents are not nephro-toxic. MRI images may be acquired with multiplanar capabilities which are especially useful in depicting various anatomic relationships. MRI system is specifically used in characterization of metastases and primary liver tumors e.g., benign lesions such as focal nodular hyperplasia (FNH), adenoma, hemangioma and malignant lesions (cancer) such as hepatocellular carcinomas (HCC).

Chest MRI is used to detect following disorders: thymus tumor, lung masses, esophageal mass, other masses (aggregations of cells) or tumors of the chest, abnormal lymph nodes, swollen glands and enlarged lymph nodes in any location of the chest, staging of tumors including invasion of blood vessels, alveolar bullae (COPD), bronchial abnormalities, bronchiectasis, cystic lung lesions, pleural abnormalities, including thickening or pleural effusion, abnormal pulmonary vessels, aortic stenosis, etc.

Abdominal MRI may reveal many medical conditions, including: abscess, acute tubular necrosis, adrenal masses, cancer, enlarged spleen or liver, gallbladder or bile duct problems, gallstones, bile duct stones, hemangiomas, kidney infection, kidney damage, lymphadenopathy, obstructed venacava, pancreatic cancer, tumor of the gallbladder, abdominal aortic aneurysm, ovarian cancer, etc.

MR Angiography is used to detect blockages or enlargements of blood vessels, including the aorta, renal arteries, and arteries in the legs, renal arterial obstruction, renal vein thrombosis, etc.

The following artifacts are present in MR imaging:

  1. Partial Volume
  2. RF Noise
  3. Intensity inhomogeneity
  4. Gradient
  5. Motion
  6. Wrap Around
  7. Gibbs Ringing
  8. Susceptibility
These MR imaging artifacts are shown in [Figure 1] respectively.

The main advantages of MR imaging system are:.

  • It has an excellent capability for soft tissue imaging
  • It has very high resolution of the order of 1mm cubic voxels
  • It has high signal to noise ratio
  • Multi channel images with variable contrast can be achieved by using different pulse sequences; this can be further utilized for segmenting and classifying different structures.
Disadvantages of MR imaging:

  • MR acquisition takes considerably longer time as compared to CT and
  • In case of MR it is more difficult to obtain uniform image quality.
Computed Tomography Imaging: The word tomography is derived from two Greek words; tomos, which means slice or section, and graphia, which means description. CT scan is an imaging modality which uses X-rays to obtain structural and functional information about the human body. The CT image is the reconstructed image and is reconstructed on the basis of X-ray absorption profile. X-rays are electromagnetic waves and used in diagnosis based on its property that all matters and tissues differ in their ability to absorb X-rays. [2] Dense tissues such as the bones appear white on a CT film while soft tissues such as the brain or liver appear gray. The cavities filled with air such as lungs appear black. CT performs better in cases of trauma and emergent situations. It provides better bone detail and has high sensitivity for acute hemorrhage. CT has become an important tool in medical imaging to supplement X-rays, medical ultrasonography (USG) and MR imaging. Although it is still quite expensive, it is the gold standard in the diagnosis of a large number of different disease entities. It is, more recently, being used in early screening of diseases, for example CT colonography for patients with a high risk of colon cancer.

CT scans are particularly used in imaging and the diagnosis of following body parts: brain, liver, chest, abdomen and pelvis, spine and also for CT based angiography.

Brain CT Imaging : In case of brain imaging, CT scans are typically used to detect: bleeding, brain damage and skull fracture in patients with head injuries; bleeding caused by a ruptured or leaking aneurysm in a patient with a sudden severe headache, blood clot or bleeding within the brain shortly after a patient exhibits symptoms of a stroke, brain tumors, cyst, diseases related to malformations of the skull, enlarged brain cavities (ventricles).

CT scanning is fast and simple, provides more detailed information on head injuries, and stroke; can reveal internal injuries and bleeding quick enough to help save lives in emergency cases.

Liver CT Imaging: In case of liver imaging, CT is the most commonly used imaging technique for evaluation of hepatic lesions. Large hepatocellular carcinomas tend to be heterogeneous, and may demonstrate a typical mosaic appearance on CT. There are relative advantages and disadvantages of both hepatic MR and CT. In general, CT is less costly than MR, more readily available, and most radiologists and many referring physicians have a relatively high degree of confidence in looking at CT images. Some studies, however, have found that CT is less sensitive and specific than MR for detection and characterization of focal hepatic disease.

Chest Imaging: Chest CT is used to detect: tumors in lungs, pneumonia, tuberculosis, emphysema, diffuse interstitial lung diseases, inflammation or other diseases of pleura, the membrane covering the lungs.

Abdomen and Pelvis CT Imaging: It is used to detect: abscesses in abdomen, inflamed colon, cancers of (i) colon, (ii) liver and (iii) pancreas, pancreatitis, lymphoma, diverticulitis, appendicitis.

Spine CT Imaging: It is used to detect various types of tumors in vertebral column, herniated inter-vertebral disk, fractures and other injuries and measure bone density and level of osteoporosis.

CT based angiography is used to identify a small aneurysm or arteriovenous malformation inside the brain, detect thrombosis in veins, atherosclerotic disease in the carotid artery and indicate disease in renal artery.

The artifacts present in CT imaging are: [4]

1) partial volume effect [Figure 1]a, 2) streak artifacts, 3) motion artifacts, 4) beam hardening artifacts, 5) ring artifacts, and 6) bloom artifacts.

These artifacts are shown in [Figure 2].

Advantages of CT imaging include:

  • Less expense and wide availablity
  • High spatial resolution with modern multi-slice scanners
  • Short scan time
  • Higher sensitivity than MR for sub-arachnoids hemorrhage
  • Higher sensitivity in detecting intra-cranial calcifications
Disadvantages of CT imaging system are:

  • Inferior soft tissue contrast compared to MRI as it is X-ray-based
  • Radiation exposure
Despite the disadvantages, CT scans are extensively used in the radiographic study of brain, liver and thorax.

Representation of Medical Images: Images are presented in 2-D as well as in 3-D domain. In the 2-D domain each element is called pixel, while in 3-D domain it is called voxel. In some cases we represent 3-D images as a sequential series of 2-D slices. The advantages associated with this type of representation include requirement of lower computational complexity and lesser memory. [5],[6]


   Segmentation Top


Segmentation is the process dividing an image into regions with similar properties such as gray level, color, texture, brightness, and contrast. [7],[8],[9] The role of segmentation is to subdivide the objects in an image; in case of medical image segmentation the aim is to:

  • Study anatomical structure
  • Identify Region of Interest i.e. locate tumor, lesion and other abnormalities
  • Measure tissue volume to measure growth of tumor (also decrease in size of tumor with treatment)
  • Help in treatment planning prior to radiation therapy; in radiation dose calculation
Automatic segmentation of medical images is a difficult task as medical images are complex in nature and rarely have any simple linear feature. Further, the output of segmentation algorithm is affected due to

  • partial volume effect.
  • intensity inhomogeneity
  • presence of artifacts
  • closeness in gray level of different soft tissue
Artifacts present in MR and CT images can be divided into three categories on the basis of image processing technique needed to rectify them: (i) artifacts needing appropriate filtering technique. For example, noise artifact, susceptibility artifact and presence of nonsharp edges in the image (ii) artifact needing appropriate image restoration techniques for example motion artifacts and (iii) artifact needing specific algorithm are; partial volume, intensity inhomogeneity.

Although a number of algorithms have been proposed in the field of medical image segmentation, medical image segmentation continues to be a complex and challenging problem. Different researchers have done the classification of segmentation techniques in one or another way. [1],[10] At present, from the medical image processing point of view we have done the classification of segmentation techniques on the basis of gray level based and textural feature based techniques. Further, we consider artificial intelligence as tools to optimize these basic techniques to achieve accurate segmentation results. Thus, the broad classification of techniques available for segmentation of an image classified into two classes is as follows:

Methods based on gray Level features

  1. Amplitude segmentation based on histogram features [11]
  2. Edge based segmentation
  3. Region based segmentation [12]
Methods based on texture features [13],[14]

Method based on gray level features


Amplitude segmentation based on histogram features

This includes segmentation of an image based on thresholding of histogram features and gray level thresholding is perhaps the simplest example of this technique. This is particularly suitable for an image with region or object of uniform brightness placed against a back ground of different gray level, A threshold can be applied to segment the object and background. Mathematically the threshold can be defined as follows.



Where r i, j is the resulting pixel at co-ordinate (i, j), p i, j is the pixel of input image and T is the value of threshold.

Equation 1 gives good results for segmentation of image with bi-modal histogram and fails in the case of an image with multi-modal histogram. Thresholding operation, defined by equation 1 is very basic and simple, and works well only when the object and background have uniform brightness of distinct gray level values respectively. This simple threshold operation does not work well at segmentation of images with multiple objects each having distinct gray level value varying over a band of values. To overcome this limitation, band thresholding based multiple thresholding operation is applied as follows:



Here, the K th band is corresponding to object/region having pixel values in the range of T k to T k+1 where T k is the lower limit of gray level and T k+1 is the upper limit of Gray level band.

For application of thresholding based segmentation technique, it is required to apply the correct threshold values in order to achieve proper segmentation results, otherwise results are poor. The histogram of an image is particularly used to determine the value of threshold. The histogram of abdomen CT image is shown in [Figure 3]. There are three peaks (maxima) separated by two minima. The values of these minima are selected as threshold for segmentation of image; the original Abdomen CT image and corresponding segmentation result are shown in [Figure 4]a and b respectively.

Object 1 belongs (0 to 55)

Object 2 belongs (55 to 200)

Object 3 belongs (200 to 255)

For histogram-based optimal segmentation of images numbers of methods have been proposed by different researchers. [11],[15],[16],[17] Frank et al, [17] have used optimal threshold selection method for segmentation of T1 weighted MR brain image to segment gray matter, white matter, and cerebro spinal fluid.

Limitations

Selection of proper values of threshold is quite difficult. Performance is affected in presence of artifacts.

Edge based segmentation

Edge based segmentation is the most common method based on detection of edges i.e. boundaries which separate distinct regions. Edge detection method is based on marking of discontinuities in gray level, color etc., and often these edges represent boundaries between objects. This method divides an image on the basis of boundaries.

Number of edge detecting operators based on gradient (derivative) function are available e.g. Prewitt, Sobel, Roberts (1 st derivative type) and Laplacian (2 nd derivative type), Canny, Marr-Hilclrath edge detector. Further, in edge based segmentation method, it is required to build the border by combining the detected edges into a edge chain in this process the spurious, or fake edges, weak edges are removed by thresholding operation. The different edge based segmentation algorithms are:

The generalized algorithm for edge based segmentation has the following steps.

  1. Apply the derivative operator to detect edges of the image
  2. Measure the strength of edges by measuring amplitude of the gradient
  3. Retain all edge having magnitude greater than threshold T (removal of weak edge)
  4. Find the position of crack edges; the crack edge is either retained or rejected based on the confidence it receives from it predecessor and successor edges
  5. Step 3 and 4 are repeated with different values of threshold so as to find out the closed boundaries; segmentation of an image is achieved
[Figure 5] shows the result of edge based segmentation of abdomen CT image [Figure 4]a, in the present result canny edge detector has been employed.

The limitations of edge based method are:

  • Performance is affected by the presence of noise
  • fake edges and weak edges may be present in the detected edge image which may have a negative influence on segmentation results
  • Edge detection techniques are required to be used in conjunction with region-based technique for complete segmentation.
Region based segmentation

Region based methods are based on the principle of homogeneity - pixels with similar properties are clustered together to form a homogenous region. The criteria for homogeneity is most of the time gray level of pixels [24] and this criteria can be specified by following conditions

R 1Ụ R 2 Ụ R 3 Ụ… Ụ R i = I

where R 1 , R 2 , R 3 ,…R i are the region in the image I,

and further, R 1∩ R 2 ∩ R 3 ∩… ∩ R i = 0

This is as per the set theory of homogeneity.

Region based segmentation is further divided into three types based on the principle of region growing:

  1. Region merging
  2. Region splitting
  3. Split and merge
Region merging

  • In this method some seeding points are required to initialize the process, the segmentation results are dependent on the choice of seeds.
  • Regions are grown iteratively by merging the neighboring pixels depending upon the merging criterion.
  • This process is continued until all pixels are assigned to their respective regions as per merging criterion.
Region splitting: Its principle is just opposite to region merging and whole image is continuously split until no further splitting of a region is possible.

Split and merge method: This is the combination of splits and merges utilizing the advantage of the two methods. This method is based on quad quadrant tree representation of data whereby image segment is split into four quadrants provided the original segment is non-uniform in properties. After this the four neighboring squares are merged depending on the uniformity of the region (segments). This split and merge process is continued until no further split and merge is possible.

The algorithm for split and merge follows the following steps.

  1. Define homogeneity criterion. Break image into four square quadrants
  2. If any resultant square is not homogeneous split it further into four quadrants
  3. At each level merge the two or more neighboring regions satisfying the condition of homogeneity
  4. Continue the split and merge until no further split and merge of region is possible
Apart from the above-mentioned techniques watershed segmentation based on the concept of topography and hydrography is also a region-based segmentation. [25],[26],[27],[28]

[Figure 6] shows the result of region based segmentation of abdomen CT image [Figure 4]a. K-mean clustering algorithm has been used in present case to cluster the pixels having similar gray levels.

The limitation of region based segmentation is that there are chances of under segmentation and over segmentation of regions in the image. However, this problem can be rectified in two ways

  • By optimally selecting the criterion for segmentation, for this several algorithm utilizing artificial intelligence techniques have been developed.
  • By combining region based approach with edge based approach. [12],[29]
Method based on the textural features

Textural features of image are important from image segmentation and classification point of view. Different researchers have used these features to achieve image segmentation, classification, and both segmentation as well as classification. The aim of texture based segmentation method is to subdivide the image into region having different texture properties, while in classification the aim is to classify the regions which have already been segmented by one or other method.

Definition of texture: Texture is defined as something consisting of mutually related elements. [24] A texture may be fine coarse, smooth, or grained depending upon its tone and structure. While tone is based on pixel intensity properties, structure is the spatial relationship between pixels. [30],[31] Further texture can be defined as the spatial arrangements of texture primitives or texture elements (also called as textone), arranged in more or less periodic manner. Texture primitive is a group of pixels representing the simplest or basic sub pattern and follows three main approaches for texture feature extraction based on the type of approach used.

  1. Statistical approach
  2. Syntactic or structural approach and
  3. Spectral approach
In case of statistical approach, texture is defined by a set of statistically extracted features represented as vector in multidimensional feature space. The statistical features could be based on first-order, second-order or higher-order statistics of gray level of an image. The feature vector so generated from patterns is assigned to their specific class by probabilistic or deterministic decision algorithm. [32] In case of syntactic approach, texture is defined by texture primitives which are spatially organized according to placement rules to generate complete pattern.

In syntactic feature based pattern recognition, a formal analogy is drawn between the structural pattern and the syntax of language. [33]

In spectral method, textures are defined by spatial frequencies and evaluated by autocorrelation function of a texture.

Some methods available for textural feature extraction and classification based on the above approaches are: co-occurrence matrix method based on statistical description of gray level of an image, [34],[35] gray level run length method, [36] fractal texture description method, [37] syntactic method [38] and Fourier filter method. [39]

Comparing the above-mentioned three approaches; spectral frequency-based methods are less efficient while statistical methods are particularly useful for random patterns/textures and for complex patterns, syntactic or structural methods give better results.

Texture based methods as best suited for segmentation of medical image, when compared to segmentation of medical image using simple gray level based methods. [40],[41],[42]

Other approaches of segmentation

Apart from the above methods, the following two methods of image segmentation are also available.

  • Model based segmentation and
  •  Atlas More Details based segmentation.
Model based segmentation: The basic approach is that the structure of organs has a repetitive form of geometry and can be modeled probabilistically for variation of shape and geometry. This can be used as constraint while segmenting the image and involves:

  1. registration of the training data.
  2. probabilistic representation of variation of registered data.
  3. statistical influence between model and image.
Model based methods of segmentation involve active shape and appearance model, deformable models and level-set based models.

Disadvantages

  1. They require manual interaction to place an initial model and choose appropriate parameters.
  2. Standard deformable models can also exhibit poor convergence to concave boundaries. [10]
Atlas based segmentation approaches are the most frequently used and powerful approaches in the field of medical image segmentation. In this, information on anatomy, shape, size, and features of different, organs, soft tissues is compiled in the form of atlas or look up table (LUT). Atlas guided approaches are similar to co-relation approaches and the plus point of atlas based approaches is - it performs segmentation and classification in one go. Atlas based segmentation approaches are among the third-generation algorithms. There are indications that certain atlas based methods can compete with manual segmentations although atlas selection, atlas registration procedure, and the manual tracing protocol used in atlas formation are factors that can affect performance. [1] However, they face limitations in segmenting complex structure with variable shape, size, and properties and expert knowledge is required in building the database.

Artificial Intelligence Tools for Segmentation and Classification

Automatic segmentation
methods have been based on artificial intelligence (AI) based techniques. AI techniques can be classified as supervised and unsupervised. Supervised segmentation requires operator interaction throughout the segmentation process whereas unsupervised methods generally require operator involvement only after segmentation is complete. Unsupervised methods are preferred to ensure a reproducible result [43] ; however, operator interaction is still required for error correction in the event of an inadequate result. [44]

Supervised methods

In the supervised category, we can place mostly Artificial Neural Network (ANN) based algorithms. ANN is composed of large number of interconnected processing elements (artificial neurons) working in unison to solve specific problems. The main advantages of ANN are:

  • ability to learn adaptively, using training data to solve complex problems.
  • capability of self-organization; it can create its own organization depending upon the information it receives during learning time
  • capability of performance in real time because of parallel configuration
In case of ANN, learning is achieved by the adaptation of weights and bias of the neurons with respect to the training procedure and training data. ANN has been widely used for segmentation and classification purposes in both supervised and unsupervised modes. [45],[46],[47] Although a variety of different neural network based algorithms have been developed for texture based segmentation and classification with good classification accuracy, [47],[48] most of these texture classifier algorithms require extensive supervision, training; their performance is sensitive to training parameters and is adversely affected in the presence of noise. At times supervised image segmentation and classification methods become very expensive, difficult and even impossible to correctly select and label the training data with its true category. [49] Training is the main requirement of many ANN based algorithms where the classifiers need to be trained before it can be applied to segmentation and classification problem. Further, for different data sets, analysis of different images of different type and format, the whole effort of selecting training data set and training is required to be redone.

Unsupervised methods

Most of the unsupervised algorithms are cluster based and not dependent on training and training data. The two commonly used algorithms for clustering are K-mean or Hard C-mean and Fuzzy C-means. [50] K-means algorithm produces results that correspond to hard segmentation while fuzzy C-mean produces soft segmentation which can be converted into hard segmentation by allowing the pixels to have membership of cluster in which they have maximum value of membership coefficients.

In clustering, the aim is to construct decision boundaries based on unlabeled training data. [49] Clustering is the process of finding natural grouping clusters in multidimensional feature space. It is difficult because clusters of different shapes and sizes can occur in multidimensional feature space. A number of functional definitions of clusters have been proposed: Patterns within a cluster are more similar to each other than patterns belonging to different clusters. [49] Image segmentation may be considered a clustering [51-53] process in which the pixels are classified into the attribute regions based on the texture feature vector calculated around the pixel local neighborhood. Fuzzy clustering is a good method of classifying collection of data point to reside in multiple clusters with different degrees of membership (fuzzy c mean algorithm). [50]

However, the main limitations of fuzzy clustering algorithm are: (a) sensitivity to initial partition matrix (b) stopping criterion (c) solution may get stuck at local minima. Hence, clustering techniques may not result in optimal solution and there is no best clustering algorithm for a particular application. A number of different algorithms are required to be tried to find the best one.

Segmentation of CT and MR Images

Segmentation of CT and MR images involves three main image related problems; noise that can alter the intensity of a pixel such that its classification becomes uncertain, intensity inhomogeneity where the intensity level of a single tissue class varies gradually over the extent of the image, and images have finite pixel size and are subject to partial volume averaging where individual pixel volumes contain a mixture of tissue classes so that the intensity of a pixel in the image may not be consistent with any one class.

Some methods available for CT image segmentation are:

Methods available for MR image segmentation: MR imaging is specifically used in brain imaging and thus lot of research work has been done particularly in the areas of MR brain image segmentation. [61],[62],[63],[64] The main goal in brain MR segmentation is to segment gray matter, white matter and cerebrospinal fluid. Segmentation is also used to find out the regions corresponding to lesions tumors, cyst, edema, and other pathologies and for this mostly T1- weighted images are used.

Most of the segmentation methods available for CT and MR images segmentation are intensity based i.e. gray level based hence; the segmentation results are affected by (1) intensity in-homogeneities and (2) partial volume effects. Accordingly, different researchers have proposed methods for correction of these problems.

Intensity In-homogeneity Correction: In MRI intensity, inhomogeneity artifacts cause shading effect to appear over the images. [65] As a result, they affect the segmentation result while using simple gray level based segmentation techniques. One way to rectify these intensity inhomogeneity artifacts is by using image enhancement techniques. [66],[67] Further, for this segmentation, techniques based on statistical methods [68],[69] and fuzzy methods, [58],[70],[71],[72],[73] which gives soft segmentation results are particularly useful in overcoming the above mentioned limitation. A review of methods for correction of intensity inhomogeneity artifact in MR images has been presented by Vovk et al. [74]

Partial Volume Effect Correction: When multiple tissues contribute to single pixel or voxel the resultant image is blurred at boundaries of the different region or object and this effect is called as partial volume effect. To deal with partial volume effect, soft segmentation is a good option. In soft segmentation we allow the region or classes to overlap i.e. pixels are allowed to have multiple memberships with varying degree of membership coefficient in different regions. In hard segmentation, we do not allow overlapping of the segmented region and the pixels are forced to reside in the region in which they are having maximum membership. Thus soft segmentation retains more information about the original image by allowing the pixel to have membership in the multiple regions.

For hard segmentation the pixel is having binary membership defined as



Where j is j th pixel belongs to image (I), and m k,j is the membership function of j th pixel in region R k . Similarly for soft segmentation the pixel has multiple memberships in different regions and the membership function must satisfy the following constraints:



where N = total number of segmented regions in the image (I)

From the value of membership of pixel j in R k (k th region) one can measure how strongly the pixel belongs to the region; more the membership value more strongly it is the member of region R k . At the border region the pixel can have different membership in different regions.

Fuzzy clustering is an excellent method for soft segmentation [58],[72] and the most widely used unsupervised algorithm for segmentation of both CT and MR images. Soft segmentation based membership can be converted to hard segmentation by allowing the pixel to have the value of membership function one (1) corresponding to the region for which it has highest membership value. [Figure 7]A shows brain CT image; individual segments of this image obtained using simulated annealing based fuzzy-c-means algorithm [58] are shown in [Figure 7]B-E and [Figure 7]F shows segmented image in pseudo color.


   Conclusion Top


Computer-aided segmentation is a key step finding application in computer aided diagnosis, clinical studies, and treatment planning. In recent years a wide variety of approaches have been proposed to segment CT and MR images having their own merits and limitations. The present review provides the basics of segmentation approaches and their respective features.

The approaches for image segmentation discussed in this review can be ranked on the basis of applicability, suitability, performance, and computational cost. Segmentation techniques based on gray level techniques such as thresholding, and region based techniques are the simplest techniques and find limited applications. However, their performance can be improved by integrating them with artificial intelligence techniques. Techniques based on textural features utilizing atlas or look-up-table have excellent results on medical image segmentation. However, they need expert knowledge in building the atlas. The limitation of atlas based technique is that under certain circumstances it becomes difficult to correctly select and label data; has difficulties in segmenting complex structure with variable shape, size, and properties. In such situations it is better to use unsupervised methods such as fuzzy-c-means algorithm.

A variety of different neural network-based algorithms are also available for texture-based segmentation and classification having good accuracy. However, most of these neural network-based algorithms require extensive supervision and training and their performance depends upon the training method and data used in training. Finally, it is desired from medical image segmentation and classification algorithms that they must have the following features: a) accuracy, b) reliability, c) repeatability, d) robustness and e) least dependency on the operator.

 
   References Top

1.Withey DJ, Koles ZJ. Three generations of medical image segmentation: Methods and available software. Int J Bioelectromag 2007;9:67-8.  Back to cited text no. 1      
2.Prince JL, Links JM. Medical imaging signals and system. Pearson Education 2006.  Back to cited text no. 2      
3.Macovski A. Medical imaging systems. Prentice-Hall: 1983.  Back to cited text no. 3      
4.Popilock R, Sandrasagaren K, Harris L, Kaser KA. CT artifact recognition for the nuclear technologist. J Nucl Med Technol 2008;36:79-81.  Back to cited text no. 4      
5.Li H, Deklerck R, Cuyper BD, Hermanus A, Nyssen E, Cornelis J. Object recognition in brain CT-scans: Knowledge based fusion of data from multiple feature extractors. IEEE T Med Imaging 1995;14:212-29.   Back to cited text no. 5      
6.Pham DL, Prince JL, Dagher AP, Xu C. An automated technique for statistical characterization of brain tissues in magnetic resonance imaging. Int J Patt Rec Art Intel 1997;11:1189-211.  Back to cited text no. 6      
7.Gonzalez RC, Woods RE. Digital image processing. 2nd ed. Pearson Education 2004.  Back to cited text no. 7      
8.Pratt KW. Digital image processing. 3rd ed. Willey; 2001. p. 551-87 .  Back to cited text no. 8      
9.Pal NR, Pal SH. A review on image segmentation techniques. Pattern Recog 1993;26:1277-94.  Back to cited text no. 9      
10.Pham DL, Xu C, Prince JL. Current methods in medical image segmentation. Ann Rev Biomed Engg 2000;2:315-37.  Back to cited text no. 10      
11.Ramesh N, Yoo JH, Sethi IK. Thresholding based on histogram approximation. IEEE Proc Vision Image Signal Proc 1995;142:271-9.  Back to cited text no. 11      
12.Sharma N, Ray AK. Computer aided segmentation of medical images based on hybridized approach of edge and region based techniques. Proc. of Int. Conf. on Mathematical Biology', Mathematical Biology recent trends by ANAMAYA Publishers; 2006. p. 150-5 .   Back to cited text no. 12      
13.Wang Z, Gierriero A, Sario M. Comparison of several approaches for segmentation of texture images. Pattern Recog Lett 1996;17:509-21.  Back to cited text no. 13      
14.Sharma N, Ray AK, Sharma S, Shukla KK, Pradhan S, Aggarwal LM. Segmentation and classification of medical images using texture-primitive features: Application of BAM-type artificial neural network. J Med Physics 2008;33:119-26.  Back to cited text no. 14      
15.Lai CC. A novel image segmentation approach based on particle swarm optimization. IEICE Trans Fundamentals 2006;E89A:324-7.  Back to cited text no. 15      
16.Brink AD. Minimum spatial entropy threshold selection. IEE Proc Vision Image Signal Proc 1995;142:128-32.  Back to cited text no. 16      
17.Frank RJ, Grabowski TJ, Damasio H. Voxelvise percentage tissue segmentation of human brain Magnetic resonance images. In Abstracts, 25th Annual meeting, Society for Neuroscience, Society for Neuroscience. Washington, DC: 1995. p 694.  Back to cited text no. 17      
18.Hancock ER, Kittler J. Edge labeling using dictionary-based relaxation. IEEE Trans PAMI 1990;12:165-81.  Back to cited text no. 18      
19.Liow YT. A contour tracing algorithm that preserves common boundaries between regions. CVGIP Image Understanding 1991;53:313-21.  Back to cited text no. 19      
20.Law T, Itoh H, Seki H. Image filtering, edge detection, and edge tracing using fuzzy reasoning. IEEE Trans PAMI 1996;18:481-91.   Back to cited text no. 20      
21.Ney H. A comparative study of two search strategies for connected word recognition Dynamic programming and heuristic search. IEEE Trans PAMI 1992;14:586-95.  Back to cited text no. 21      
22.Xu L, Oja E. Randomized Hough transform (RHT): Basic mechanism and algorithms, and computational complexities. CVGIP Image Comput 1993;57:131-54.  Back to cited text no. 22      
23.Kalvian H, Hirvonen P, Xu L, Oja E. Probabilistic and non-probabilistic Hough transform: Overview and comparisons. Image Comput 1995;13:239-52.   Back to cited text no. 23      
24.Sonka M, Hlavac V, Boyle R. Image processing, analysis and machine vision. Singapore: Thomson Learning; 1999.   Back to cited text no. 24      
25.Ng HP, Ong SH, Foong KW, Nowinski WL. An improved watershed algorithm for medical image segmentation. Proceedings 12th International Conference on Biomedical Engineering, 2005.  Back to cited text no. 25      
26.Nguyen HT, Worring M, Boomgaard RV. Watersnakes: Energy-driven watershed segmentation. IEEE Trans Pattern Anal Mach Intellig 2003;25:330-42.  Back to cited text no. 26      
27.Najman L, Schmitti M. Geodesic saliency of watershed contours and hierarchical segmentation. IEEE Trans PAMI 1996;18:12.  Back to cited text no. 27      
28.Lopez AM. Evaluation of methods for ridge and valley detection. IEEE Trans PAMI 1999;21:327.  Back to cited text no. 28      
29.Gevers T, Smeulders AW. Combining region splitting and edge detection through guided Delaunay image subdivision. Comp Vision Pattern Recognition IEEE Computer Society, Los Almitos, CA: 1997. p. 1021-6.  Back to cited text no. 29      
30.Haralick RM. Statistical and structural approaches to texture. Proc IEEE 1979;67:786-804.  Back to cited text no. 30      
31.Julsez B. Textons, the element of texture perception and their interactions. Nature 1981;290:91-7.  Back to cited text no. 31      
32.Fukunaga K. Introduction to statistical pattern recognition. 2nd ed. Academic Press; 1990.  Back to cited text no. 32      
33.Pavilidis T. Structural description and graph grammar. In: Chang SK, Fu KS, editors. Pictorial information systems. Springer Verlag Berlin; 1980. p. 86-103.  Back to cited text no. 33      
34.Argenti F, Alparone L, Benelli G. Fast algorithm for texture analysis using co-occurrence matrices. IEE Proc Part F: Radar Signal Proc 1990;137:443-8.  Back to cited text no. 34      
35.Haralick RM, Shanmugan K, Dinstein I. Textural features for image classification. IEEE Trans Syst Man Cybern 1973;3:610-22.  Back to cited text no. 35      
36.Chu CC, Aggarwal JK. The integration of image segmentation maps using region and edge information. IEEE Trans Pattern Anal Mach Intellig 1993;15:1241-52.  Back to cited text no. 36      
37.Chaudhuri BB, Sarkar N. Texture segmentation using fractal dimension. IEEE Trans PAMI 1995;17:72-7.  Back to cited text no. 37      
38.Lu SY, Fu KS. A syntactic approach to texture analysis. Comput Graphics Image Proc 1978;7:303-30.  Back to cited text no. 38      
39.Duda RO, Hart PE, Stork DG. Pattern classification. Singapore: Wiley; 2001. p 350-93 .  Back to cited text no. 39      
40.Tesar L, Shimizu A, Smutek D, Kobatake H, Nawano S. Medical image analysis of 3D CT images based on extension of Haralick texture features. Comput Med Imaging Graphics 2008;32:513-20.  Back to cited text no. 40      
41.Prasad M, Brown MS, Abtin F, Vasunilsahorn S, Goldin JG. Comparison of scleroderma detection techniques in CT images using supervised and unsupervised methods. Int J Biomed Engg Technol 2008;1:4.  Back to cited text no. 41      
42.Koss JE, Newman FD, Johnson TK, Krich DL. Abdominal organ segmentation using texture transform and Hopfield neural network. IEEE Trans Med Image 1999;18:640-8.   Back to cited text no. 42      
43.Clarke LP, Velthuizen RP, Camacho MA, Heine JJ, Vaidyanathan M, Hall LO, et al. 'MRI segmentation: Methods and applications. Magn Reson Imaging 1995;13:343-68.  Back to cited text no. 43      
44.Olabarriaga SD, Smeulders AW. Interaction in the segmentation of medical images: A survey. Med Image Anal 2001;5:127-42.  Back to cited text no. 44      
45.Kumar CV, Damyanti G, Pant R, Sreedhar CM. Segmentation and grading of brain tumor an apparent diffusion coefficient image using self organizing maps. Comput Med Imaging Graphica 2007;31:473-84.  Back to cited text no. 45      
46.Chaplot S, Patnaik LM, Jagannathan N. Classification of magnetic resonance brain images using wavelets as input to support vector machine and neural network. Biomed Signal Proc Control 2006;1:86-92.  Back to cited text no. 46      
47.Gletsos M, Mougiakakou SG, Matsopoulos GK, Nikita KS, Nikita AS, Kelekis D. A computer-aided diagnostic system to characterize CT focal liver lesions: Design and optimization of a neural network classifier. IEEE Trans Inform Technol Biomed 2003;7:153-62.  Back to cited text no. 47      
48.Engeland SV, Timp S, Karssemeijer N. Finding corresponding regions of interest in mediolateral oblique and craniocaudal mamographic views. Med Phys 2006;33:3203-12.  Back to cited text no. 48      
49.Jain AK, Duin RP, Mao J. Statistical pattern recognition: A review. IEEE Trans PAMI 2000;22:4-37.  Back to cited text no. 49      
50.Bezdek JC, Hall LO, Clarke LP. Review of MR image segmentation techniques using patternrecognition. Med Phys 1993;20:1033-48.  Back to cited text no. 50      
51.Frigui H, Krishnapuram R. A robust competitive clustering algorithm with application in computer vision. IEEE Trans PAMI 1999;21:450-65.  Back to cited text no. 51      
52.Tseng LY, Yang SB. A genetic approach to the automatic clustering problem. Patten Recog 2001;34:415-24.   Back to cited text no. 52      
53.Bandyopadhyay S. Simulated annealing using a reversible jump markov chain monte carlo algorithm for fuzzy clustering. IEEE Trans Knowledge Data Engg 2005;17:479-90.  Back to cited text no. 53      
54.Aurdal L, Omholt-Jensen T, Mala T, Edwin B. Segmentation of liver vessels as seen in MR and CT images. Int Congr Ser 2004;1268:201-6.  Back to cited text no. 54      
55.Bert A, Dmitriev I, Agliozzo S, Pietrosemoli N, Mandelkern M, Gallo T, et al. An automatic method for colon segmentation in CT colonography. Comput Med Imaging Graphics 2009;33:325-31.  Back to cited text no. 55      
56.Poon CS, Braun M. Image segmentation by a deformable contour model incorporating region analysis. Physics Med Biol 1997;42:1833-41.  Back to cited text no. 56      
57.Klinder T, Ostermann J, Ehm M, Franz A, Kneser R, Lorenz C. Automated model-based vertebra detection, identification, and segmentation in CT images. Med Image Anal 2009;13:471-82.  Back to cited text no. 57      
58.Sharma N, Ray AK, Sharma S, Shukla KK, Pradhan S, Aggarwal LM. Segmentation of medical images using simulated annealing based fuzzy C Means algorithm. Int J Biomed Engg Technol 2009;2:260-78.  Back to cited text no. 58      
59.Tabakov M. A fuzzy segmentation method for computed tomography images. Int J Intellig Inform Database Syst Technol Appl 2007;1:234-46.   Back to cited text no. 59      
60.Chen EL, Chung PC, Chen CL, Tsai HM, Chang CI. An automatic diagnostic system for CT liver image classification. IEEE Trans Biomed Engg 1998;45:783-93.  Back to cited text no. 60      
61.Song T, Gasparovicc C, Andreasen N. A hybrid tissue segmentation approach for brain MR images. Med Biol Eng Comput 2006;44:242-9.  Back to cited text no. 61      
62.Liao L, Lin T, Li B. MRI brain image segmentation and bias field correction based on fast spatially constrained kernel clustering approach. Pattern Recog Lett 2008;29:1580-8.  Back to cited text no. 62      
63.Kuo WF, Lin CY, Yung-Nien Sun YN. Brain MR images segmentation using statistical ratio: Mapping between watershed and competitive Hopfield clustering network algorithms. Comput Met Prog Biomed 2008;9:191-8.  Back to cited text no. 63      
64.Cuadra MB, Craene MD, Duay V. Dense deformation field estimation for Atlas-based segmentation of pathological MR brain images. Comput Met Prog Biomed 2006;84:67-75.  Back to cited text no. 64      
65.Sled JG, Pike GB. Standing-wave and RF penetration artifacts caused by elliptic geometry: An electrodynamic analysis of MRI. IEEE T Med Imag 1998;17:653-62.  Back to cited text no. 65      
66.Garcia-Sebastian M, Hernandez C, Anjou A. Robustness of an adaptive MRI segmentation algorithm parametric intensityinhomogeneity modeling. Neurocomputing 2009;72:2146-52.  Back to cited text no. 66      
67.Chen Y, Zhang J, Macione J. An improved level set method for brain MR image segmentation and bias correction. Comput Med Imaging Graphics 2009;33:510-19.  Back to cited text no. 67      
68.Cuadra MB, Cammoun L, Butz T. Comparison and validation of tissue modelization and statistical classification methods in T1-weighted MR brain images. IEEE Trans Med Imaging 2005;24:1548-65.   Back to cited text no. 68      
69.Zaidi H, Ruest T, Schoenahl F, Montandon ML. Comparative assessment of statistical brain MR image segmentation algorithms and their impact on partial volume correction in PET. Neuroimage 2006;32:1591-1607.  Back to cited text no. 69      
70.Zhu C, Jiang T. Multicontext fuzzy clustering for separation of brain tissues in magnetic resonance images. Neuroimage 2003;18:685-96.  Back to cited text no. 70      
71.Ahmed MN, Yamany SM. A modified fuzzy C-means algorithm for bias field estimation and segmentation of MRI data. IEEE Trans Med Imaging 2002;21:193-9.  Back to cited text no. 71      
72.Pham DL, Prince JL. An adaptive fuzzy c-means algorithm for image segmentation in the presence of intensity in homogeneities. Pattern Rec Letters 1999;14:57-68.  Back to cited text no. 72      
73.Shen S, Sandham W, Ranat M, Sterr A. MRI fuzzy segmentation of brain tissue using neighborhood attraction with neural-network optimization. IEEE Trans Inform Technol Biomed 2005;9:459-67.  Back to cited text no. 73      
74.Vovk U, Pernuš, Franjo, Likar, Boštjan. A review of methods for correction of intensity inhomogeneity in MRI. IEEE Trans Med Imaging 2007;26:405-21.  Back to cited text no. 74      


    Figures

  [Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5], [Figure 6], [Figure 7]
 
 
    Tables

  [Table 1]


This article has been cited by
1 Automated cerebellar segmentation: Validation and application to detect smaller volumes in children prenatally exposed to alcohol
Valerie A. Cardenas,Mathew Price,M. Alejandra Infante,Eileen M. Moore,Sarah N. Mattson,Edward P. Riley,George Fein
NeuroImage: Clinical. 2014; 4: 295
[Pubmed] | [DOI]
2 Three-Dimensional Reconstruction Volume: A Novel Method for Volume Measurement in Kidney Cancer
Timothy A. Durso,Jonathan Carnell,Thomas T. Turk,Gopal N. Gupta
Journal of Endourology. 2014; 28(6): 745
[Pubmed] | [DOI]
3 A statistical shape model of the human second cervical vertebra
Marine Clogenson,John M. Duff,Marcel Luethi,Marc Levivier,Reto Meuli,Charles Baur,Simon Henein
International Journal of Computer Assisted Radiology and Surgery. 2014;
[Pubmed] | [DOI]
4 Object extraction from T2 weighted brain MR image using histogram based gradient calculation
Gilanie, G. and Attique, M. and Hafeez-Ullah and Naweed, S. and Ahmed, E. and Ikram, M.
Pattern Recognition Letters. 2013; 34(12): 1356-1363
[Pubmed]
5 Micro-computed tomography imaging and analysis in developmental biology and toxicology
Wise, L.D. and Winkelmann, C.T. and Dogdas, B. and Bagchi, A.
Birth Defects Research Part C - Embryo Today: Reviews. 2013; 99(2): 71-82
[Pubmed]
6 Comparison of two methods for the estimation of subcortical volume and asymmetry using magnetic resonance imaging: A methodological study
Ertekin, T. and Acer, N. and Içer, S. and Ilica, A.T.
Surgical and Radiologic Anatomy. 2013; 35(4): 301-309
[Pubmed]
7 Reverse engineering of human bones by using method of anatomical features
Majstorovic, V. and Trajanovic, M. and Vitkovic, N. and Stojkovic, M.
CIRP Annals - Manufacturing Technology. 2013; 62(1): 167-170
[Pubmed]
8 High quality machine-robust image features: Identification in nonsmall cell lung cancer computed tomography images
Luke A. Hunter,Shane Krafft,Francesco Stingo,Haesun Choi,Mary K. Martel,Stephen F. Kry,Laurence E. Court
Medical Physics. 2013; 40(12): 121916
[Pubmed] | [DOI]
9 Automatic image segmentation and classification based on direction texton technique for hemolytic anemia in thin blood smears
Hung-Ming Chen,Ya-Ting Tsao,Shin-Ni Tsai
Machine Vision and Applications. 2013;
[Pubmed] | [DOI]
10 Comparison of two methods for the estimation of subcortical volume and asymmetry using magnetic resonance imaging: a methodological study
Tolga Ertekin,Niyazi Acer,Semra Ier,Ahmet T. Ilica
Surgical and Radiologic Anatomy. 2013; 35(4): 301
[Pubmed] | [DOI]
11 Reverse engineering of human bones by using method of anatomical features
Vidosav Majstorovic,Miroslav Trajanovic,Nikola Vitkovic,Milos Stojkovic
CIRP Annals - Manufacturing Technology. 2013; 62(1): 167
[Pubmed] | [DOI]
12 Inaccuracies in Additive Manufactured Medical Skull Models Caused by the DICOM to STL Conversion Process
Eero Huotilainen,Risto Jaanimets,Jir Valek,Petr Marcin,Mika Salmi,Jukka Tuomi,Antti Mkitie,Jan Wolff
Journal of Cranio-Maxillofacial Surgery. 2013;
[Pubmed] | [DOI]
13 Survey of computer-aided diagnosis of thyroid nodules in medical ultrasound images
Koundal, D. and Gupta, S. and Singh, S.
Advances in Intelligent Systems and Computing. 2013; 177 AISC(VOL. 2): 459-467
[Pubmed]
14 Voxel classification and graph cuts for automated segmentation of pathological periprosthetic hip anatomy
Malan, D.F. and Botha, C.P. and Valstar, E.R.
International Journal of Computer Assisted Radiology and Surgery. 2013; 8(1): 63-74
[Pubmed]
15 Computer Based Performance Evaluation of Segmentation Methods for Chest X-Ray Image
Manoj R. Tarambale,Nitin S. Lingayat
International Journal of Bioscience, Biochemistry and Bioinformatics. 2013; : 545
[Pubmed] | [DOI]
16 Object extraction from T2 weighted brain MR image using histogram based gradient calculation
Ghulam Gilanie,Muhammad Attique,Muhammad Hafeez-Ullah,Shahid Naweed,Ejaz Ahmed,Masroor Ikram
Pattern Recognition Letters. 2013; 34(12): 1356
[Pubmed] | [DOI]
17 Micro-computed tomography imaging and analysis in developmental biology and toxicology
L. David Wise,Christopher T. Winkelmann,Belma Dogdas,Ansuman Bagchi
Birth Defects Research Part C: Embryo Today: Reviews. 2013; 99(2): 71
[Pubmed] | [DOI]
18 Radial differential interior tomography and its image reconstruction with differentiated backprojection and projection onto convex sets
Shaojie Tang,Xiangyang Tang
Medical Physics. 2013; 40(9): 091914
[Pubmed] | [DOI]
19 Voxel classification and graph cuts for automated segmentation of pathological periprosthetic hip anatomy
Daniel F. Malan,Charl P. Botha,Edward R. Valstar
International Journal of Computer Assisted Radiology and Surgery. 2013; 8(1): 63
[Pubmed] | [DOI]
20 Human cervical carcinoma detection and glucose monitoring in blood micro vasculatures with swept source OCT
H. Ullah,E. Ahmed,M. Ikram
JETP Letters. 2013; 97(12): 690
[Pubmed] | [DOI]
21 Obtaining pseudo-3D information from single-plane X-ray imaging
J. Hrd,P. Oberta
Nuclear Instruments and Methods in Physics Research Section A: Accelerators, Spectrometers, Detectors and Associated Equipment. 2012; 690: 7
[Pubmed] | [DOI]
22 Semiautomated spleen volumetry with diffusion-weighted MR imaging
Jeongjin Lee,Kyoung Won Kim,Ho Lee,So Jung Lee,Sanghyun Choi,Woo Kyoung Jeong,Heewon Kye,Gi-Won Song,Shin Hwang,Sung-Gyu Lee
Magnetic Resonance in Medicine. 2012; 68(1): 305
[Pubmed] | [DOI]
23 Obtaining pseudo-3D information from single-plane X-ray imaging
Hrdý, J. and Oberta, P.
Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment. 2012; 690(6245269): 7-9
[Pubmed]
24 Mean multiclass Type I and II errors for training multilayer perceptron with particle swarm in image segmentation
Dos Santos, M.M. and Valença, M.J.S. and Dos Santos, W.P.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2012; 7435 LNCS: 135-142
[Pubmed]
25 Semiautomated spleen volumetry with diffusion-weighted MR imaging
Lee, J. and Kim, K.W. and Lee, H. and Lee, S.J. and Choi, S. and Jeong, W.K. and Kye, H. and Song, G.-W. and Hwang, S. and Lee, S.-G.
Magnetic Resonance in Medicine. 2012; 68(1): 305-310
[Pubmed]
26 Colorization and automated segmentation of human T2 MR brain images for characterization of soft tissues
Attique, M. and Gilanie, G. and Hafeez-Ullah and Mehmood, M.S. and Naweed, M.S. and Ikram, M. and Kamran, J.A. and Vitkin, A.
PLoS ONE. 2012; 7(3)
[Pubmed]
27 Automatic 3D segmentation of individual facial muscles using unlabeled prior information
Rezaeitabar, Y. and Ulusoy, I.
International Journal of Computer Assisted Radiology and Surgery. 2012; 7(1): 35-41
[Pubmed]
28 A Semi-Automatic Algorithm for Determining the Demyelination Load in Metachromatic Leukodystrophy
Clas, P. and Groeschel, S. and Wilke, M.
Academic Radiology. 2012; 19(1): 26-34
[Pubmed]
29 Segmentation algorithms for ear image data towards biomechanical studies
Ana Ferreira,Fernanda Gentil,Joo Manuel R. S. Tavares
Computer Methods in Biomechanics and Biomedical Engineering. 2012; : 1
[Pubmed] | [DOI]
30 A Semi-Automatic Algorithm for Determining the Demyelination Load in Metachromatic Leukodystrophy
Philipp Clas,Samuel Groeschel,Marko Wilke
Academic Radiology. 2012; 19(1): 26
[Pubmed] | [DOI]
31 Automatic 3D segmentation of individual facial muscles using unlabeled prior information
Yousef Rezaeitabar,Ilkay Ulusoy
International Journal of Computer Assisted Radiology and Surgery. 2012; 7(1): 35
[Pubmed] | [DOI]
32 A PAR-1-dependent orientation gradient of dynamic microtubules directs posterior cargo transport in the Drosophila oocyte
R. M. Parton,R. S. Hamilton,G. Ball,L. Yang,C. F. Cullen,W. Lu,H. Ohkura,I. Davis
The Journal of Cell Biology. 2011; 194(1): 121
[Pubmed] | [DOI]
33 Automatic 3D segmentation of facial soft tissues using unlabelled prior information | [Yz yumuşak dokusunun eti̇ketlenmemi̇ş nsel bi̇lgi̇ kullanilarak 3b otomati̇k bl tlenmesi̇]
Rezaeitabar, Y., Ulusoy, I.
2011 IEEE 19th Signal Processing and Communications Applications Conference, SIU 2011. 2011; art(5929573, ): 1-4
[Pubmed]
34 Combined 7-T MRI and histopathologic study of normal and dysplastic samples from patients with TLE
Garbelli, R., Zucca, I., Milesi, G., Mastropietro, A., DIncerti, L., Tassi, L., Colombo, N., (...), Spreafico, R.
Neurology. 2011; 76(13): 1177-1185
[Pubmed]
35 A PAR-1-dependent orientation gradient of dynamic microtubules directs posterior cargo transport in the Drosophila oocyte
Parton, R.M., Hamilton, R.S., Ball, G., Yang, L., Cullen, C.F., Lu, W., Ohkura, H., Davis, I.
Journal of Cell Biology. 2011; 194(1): 121-135
[Pubmed]
36 A knowledge-based approach for segmenting cerebral vasculature in neuroimages
Luo, S., Jin, J.J., Li, J.
Proceedings - 3rd International Conference on Measuring Technology and Mechatronics Automation,. 2011; : 74-77
[Pubmed]
37 Automatic and semi-automatic analysis of the extension of myocardial infarction in an experimental murine model
Esteves, T., Valente, M., Nascimento, D.S., Pinto-Do-, P., Quelhas, P.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) 6669 LNCS,. 2011; : 151-158
[Pubmed]
38 Application of AI techniques in medical image segmentation and novel categorization of available methods and tools
Rastgarpour, M., Shanbehzadeh, J.
IMECS 2011 - International MultiConference of Engineers and Computer Scientists 2011. 2011; 1: 519-523
[Pubmed]
39 A modified edge-based region growing segmentation of geometric objects
Jamil, N. and Soh, H.C. and Tengku Sembok, T.M. and Bakar, Z.A.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). 2011; 7066 LNCS(PART 1): 99-112
[Pubmed]
40 Medical Image Classification Using an Optimal Feature Extraction Algorithm and a Supervised Classifier Technique
Ahmed Kharrat,Karim Gasmi,Mohamed Ben Messaoud,Nacra Benamrane,Mohamed Abid
International Journal of Software Science and Computational Intelligence. 2011; 3(2): 19
[Pubmed] | [DOI]



 

Top
Print this article  Email this article
Previous article Next article

    

 
   Search
 
   Next article
   Previous article 
   Table of Contents
  
    Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
    Article in PDF (238 KB)
    Citation Manager
    Access Statistics
    Reader Comments
    Email Alert *
    Add to My List *
* Registration required (free)  


    Abstract
    Introduction
    Medical Imaging ...
    Segmentation
    Conclusion
    References
    Article Figures
    Article Tables

 Article Access Statistics
    Viewed13524    
    Printed390    
    Emailed2    
    PDF Downloaded601    
    Comments [Add]    
    Cited by others 40    

Recommend this journal