Full Research Paper
Biological Computer Modeling / Biological Computer Simulation
Fereshte Yousefi Rizi; Alireza Ahmadian; Javad Alirezaie; Emadoddin Fatemizadeh; Nader Rezaei
Volume 2, Issue 3 , June 2008, Pages 165-177
Abstract
Partial volume effect and image noise greatly decrease the visibility of the airway wall. Another dilemma with airway segmentation methods, which significantly influences their accuracy, is the leakage into the extra-luminal regions due to thinness of the airway wall during the process of segmentation. ...
Read More
Partial volume effect and image noise greatly decrease the visibility of the airway wall. Another dilemma with airway segmentation methods, which significantly influences their accuracy, is the leakage into the extra-luminal regions due to thinness of the airway wall during the process of segmentation. A solution to this problem in the previous methods was based on leak detection and reduction by adjusting the segmentation parameters and performing the whole segmentation process, which is very time consuming and demands user interaction. The new strategy presented here is to prevent the leakage by taking the advantage of the fact that the airway branches are cylindrically shaped objects. This has been achieved by introducing a new mathematical shape optimization approach embedded in FC-FCM algorithm to retain the cylindrical properties of the airway branches during the segmentation process. The main role of this optimization approach is to detect and correct the underlying voxels which belonging to the airway by satisfying both conditions of the fuzzy connectivity and shape features. The proposed FC-FCM algorithm was first applied on four data sets each containing 430 CT images of CT images of airway tree. The result showed an accuracy of 93% obtained for segmentation of the airway tree up to the fourth generation. We then applied OPT-FC-FCM algorithm to segment the airway tree with optimization process up to the sixth generation of airway. The result proves the ability of our proposed method to complete a visually acceptable segmentation of airway trees with no leakage. The number of detected branches was found 65 (4 times of those obtained by using just the FC-FCM method).
Full Research Paper
Biomedical Image Processing / Medical Image Processing
Azar Tolouee; Hamid Abrishami Moghaddam; Masoume Giti
Volume 2, Issue 3 , June 2008, Pages 179-189
Abstract
Automatic classification of lung tissue patterns in high-resolution computed tomography (HRCT) images of patients affected with interstitial lung diseases (ILD) is an important stage in the construction of a computer-aided diagnosis system. In this study, classification of Jung tissue patterns was conducted ...
Read More
Automatic classification of lung tissue patterns in high-resolution computed tomography (HRCT) images of patients affected with interstitial lung diseases (ILD) is an important stage in the construction of a computer-aided diagnosis system. In this study, classification of Jung tissue patterns was conducted using a new machine learning approach. The proposed system comprises three stages. In the first stage, the parenchyma region in HRCT lung images is separated using a set of thresholding, filtering and morphological operators. In the second stage, two sets of overcomplete wavelet filters, namely discrete wavelet frames and rotated wavelet frames are utilized to extract the features from the defined regions of interest (ROJs) within parenchyma. Then, in the third stage, the fuzzy k-nearest neighbor algorithm is employed to perform the pattern classification. Our experiments in lung pattern classification were rendered on four different lung tissue patterns (ground glass, honey combing, reticular, and normal) selected from a database of 340 images from 17 subjects. After applying the technique to classify these patterns in small ROis, we extended the classification scheme to the whole lung in order to produce the quantitative scores of abnormalities in lung parenchyma of the patients. The performance of the proposed method was compared with two state-of-the-art computer based methods for lung tissue characterization. It was also validated against the experienced observers. The average kappa statistic of agreement between two radiologists and the computer was found to be 0.6543 where as the average kappa statistic for the interobserver agreement was 0.6848. This computer system can approach the performance of the expert observers in the diagnosing regions of interest and can help to produce objective measures of abnormal patterns in lung HRCT images.
Full Research Paper
Biomedical Image Processing / Medical Image Processing
Emadoddin Fatemizadeh; Parisa Shooshtari
Volume 2, Issue 3 , June 2008, Pages 191-201
Abstract
Nowadays due to the huge capacity and bandwidth essentials for medical images, communications and storage purposes, medical images compression is one of most important concepts in this area. Error free compression techniques have the weakness of low compression ratio. On the other hand, lossy techniques ...
Read More
Nowadays due to the huge capacity and bandwidth essentials for medical images, communications and storage purposes, medical images compression is one of most important concepts in this area. Error free compression techniques have the weakness of low compression ratio. On the other hand, lossy techniques with high compression ratio result in low quality of the images. In recent years, some special compression schemes have been suggested by splitting the original image into two regions: Region of Interest (ROI) with lossless compression and the Region of Background (ROB) with lossy compression and a lower quality. In this paper, we proposed a novel selective compression approach to compress 3D brain MR images. For this purpose, an adaptive mesh for the first slice was designed and estimation of the gray levels of the next slices was performed through deformations of the mesh elements. After residual image determination, the error between the original image and the approximated image was transformed to the wavelet domain using a region-based discrete wavelet transform (RBDWT). Finally, the wavelet coefficients were coded by an object-based SPIHT coder.
Full Research Paper
Biomedical Image Processing / Medical Image Processing
Mehdi Marsousi; Javad Alirezaie; Armen Kocharian
Volume 2, Issue 3 , June 2008, Pages 203-214
Abstract
In this paper, a new method for boundary detection of left ventricle in echocardiography images is proposed. We have modified B-Spline Snake algorithm to achieve much faster convergence and more reliability toward noises in echocardiography images. A novel approach for inserting new node points during ...
Read More
In this paper, a new method for boundary detection of left ventricle in echocardiography images is proposed. We have modified B-Spline Snake algorithm to achieve much faster convergence and more reliability toward noises in echocardiography images. A novel approach for inserting new node points during iterations is applied to maintain a maximum distance between two adjacent nodes. This strategy is applied in order to simultaneously increase the smoothness of the contour and optimize the computational time. A multi-resolution strategy is also adapted to provide further robustness toward noises in the images. In addition, morphological operators are utilized to specify the initial contour automatically within the left ventricle chamber in echocardiography images. The parameters of node points are determined during each transition from coarser to finer resolution according to the average intensity of the sample points on the contour near each node point. The volumes of left ventricle in the end of both systolic and diastolic frames are calculated using modified Simpson method. The ejection fraction ratio is also calculated; this is frequently used by specialist before each surgery. Moreover, a method is introduced to draw the 3D model of left ventricle with the aid of basis function of B-Spline. The proposed method is assessed by comparison between the obtained results and clinical observations by expert radiologists and demonstrates a high accuracy.
Full Research Paper
Biomedical Image Processing / Medical Image Processing
Saeed Kermani; Hamid Abrishami Moghaddam; Mohammad Hasan Moradi
Volume 2, Issue 3 , June 2008, Pages 215-231
Abstract
This paper presents a new method for quantification analysis of left ventricular performance from the sequences of cardiac magnetic resonance imaging using the three-dimension active mesh model (3DAMM). AMM is composed of topology and geometry of L V and associated elastic material properties. The ...
Read More
This paper presents a new method for quantification analysis of left ventricular performance from the sequences of cardiac magnetic resonance imaging using the three-dimension active mesh model (3DAMM). AMM is composed of topology and geometry of L V and associated elastic material properties. The LV deformation is estimated by fitting the model to the initial sparse displacements which is measured by a new establishing point correspondence procedure. To improve the model, a new shape-based interpolation algorithm was proposed for reconstruction of the intermediate slices. The proposed approach is capable of estimating the displacement field for every desired point of the myocardial wall. Then it leads to measure dense motion field and the local dynamic parameters such as Lagrangian strain. To evaluate the performance of the proposed algorithm, eight image sequences (six real and two synthetic sets) were used and the findings were compared with those reported by other researchers. For synthetic image sequence sets, the mean square error between the length of motion field estimated by the Algorithm and the analytical values was less than 0.5 mm. The results showed that the strain measurements of the normal cases were generally consistent with the previously published values. The results of analysis on a patient data set were also consistent with his clinical evidence. In conclusion, the results demonstrated the superiority of the novel strategy with respect to our formerly presented algorithm. Furthermore, the results are comparable to the current state-of-the-art methods.
Full Research Paper
Biomedical Image Processing / Medical Image Processing
Raheleh Kafieh; Alireza Mehri Dehnavi; Saeed Sadri; Seyed Hamid Raji
Volume 2, Issue 3 , June 2008, Pages 233-246
Abstract
Cephalometry is the scientific measurement of head dimensions to predict craniofacial growth, plan treatment and compare different cases. There have been many attempts to automate cephalometric analysis with the aim of reducing the time required to obtain an analysis, improve the accuracy of landmark ...
Read More
Cephalometry is the scientific measurement of head dimensions to predict craniofacial growth, plan treatment and compare different cases. There have been many attempts to automate cephalometric analysis with the aim of reducing the time required to obtain an analysis, improve the accuracy of landmark identification and reduce the errors due to clinician subjectivity. This paper introduces a method for automatic landmark detection on cephalograms. We introduced a combination of model-based methods and neural networks on cephalograms. For this purpose, first some feature points were extracted using a nonlinear diffusion filter and Susan Edge Detector to model the size, rotation, and translation of skull. A neural network was used to classify the images according to their geometrical specifications. Using learning vector quantization (L VQ) for every new image, the possible coordinates of landmarks were estimated. Then a modified active shape model (ASM) was applied and a local search to find the best match to the intensity profile was used and every point was moved to get the best location. Finally, a sub-image matching procedure was applied to pinpoint the exact location of each landmark. In order to evaluate the results of this method, 20 randomly selected images were used with a drop-one-out method. Each image had a dimension of about 170x200 mm, digitized in 100 dpi (4 pixel == 1mm). On average, 24% of the 16% landmarks were within 1mm of correct coordinates, 61 percent within 2 mm, and 93 percent within 5 mm. the proposed method in this study has had a distinct improvement over the other proposed methods of automatic landmark detection.
Full Research Paper
Biomedical Image Processing / Medical Image Processing
Poune Roshani Tabrizi; Reza Aghaeizade Zoroofi
Volume 2, Issue 3 , June 2008, Pages 247-266
Abstract
Drowsiness detection is vital in preventing traffic accidents. In this project, we propose three new algorithms for pupil and iris detection, lips localization and eyes state analysis, which we incorporate into a four step system for drowsiness detection: face detection, drowsiness parameters extraction ...
Read More
Drowsiness detection is vital in preventing traffic accidents. In this project, we propose three new algorithms for pupil and iris detection, lips localization and eyes state analysis, which we incorporate into a four step system for drowsiness detection: face detection, drowsiness parameters extraction from eyes, drowsiness parameter extraction from mouth and drowsiness level determination. Many current efforts, which are based on face analysis, focus only on using a single visual cue to characterize driver's state of alertness. This approach that relies on a single visual cue may encounter difficulty when the required visual features cannot be acquired accurately or reliably. There are few systems that use several visual cues to characterize driver's state of alertness. These systems are based on IR illuminators or training data. IR illuminators can be hazardous to eye health. Thus, our proposed system determines drowsiness level using a combination of several visual cues and contextual information. Also, it requires no training data at any step or IR illuminators. We analyzed and compared different parts of the systems with other methods using IMM, HCE, CVL database and 30 video sequences in two drowsy and active states from 15 persons. Finally, we achieved excellent drowsiness level results from the study population. We determined drowsiness level as follows: 1. The eyes and mouth state (detecting whether they were open or closed) was analyzed as 94.3% and 95.1 %, respectively; 2. Drowsiness level was determined in different situations such as normal blinking, fast blinking, normal speaking, yawning and long eye closure and 3. The participants were given a warning message when the drowsiness level reached over the threshold of 0.95.
