Current Medical Imaging Reviews (v.11, #2)

Meet Our Editorial Board Member: by Ge Wang (69-69).

In medical image processing, detecting Magnetic Resonance Imaging (MRI) brain tumor is one of the most important task. A brain tumor is a collection of cells that have grown and multiplied uncontrollably. It is formed by undesired cells, either normally found in the disparate part of the brain, such as, lymphatic tissue, glial cells, neurons, blood vessels, skull or spread from cancers mainly located in other organs. Brain tumors are classified based on the type of tissue involved in the brain. Generally, it can be classified into two types such as, benign (non-cancerous) and malignant (cancerous). MRI brain tumor identification and detection is an important, but time consuming task performed by medical experts. This paper presents an automatic MRI brain detection and classification method based on K-Nearest Neighbor (KNN) classifier and Hidden Markov Model (HMM) classifier. The proposed method consists of three stages, such as, preprocessing, feature extraction and classification. Here, the Gaussian filtering technique is used to preprocess the given image by eliminating the noise and filtering the image. The feature extracting involve extracting the first order statistical features, second order statistical features and moment invariant features. Finally, the K-NN and HMM classifiers are employed to classify the given image as normal or abnormal. The experimental results evaluate the performance of the proposed algorithm in terms of sensitivity, specificity and classification rate.

Computer Aided System for Detection and Classification of Breast Cancer by I. Selvamani, G. Tholkappia Arasu (77-84).
The success for treatment of breast cancer patients depends on the early detection of breast cancer. In this paper, computer aided system for the detection and classification of breast cancer using mammogram images. The proposed system consists of the following three stages as mammogram image enhancement, feature extraction and Classification. The Shift invariant non sub sampled Contourlet transform is used for mammogram image enhancement. The transform coefficients are extracted as features for both training and classification of mammogram images. The mammogram images classification are performed using Support vector machine (SVM) and feed forward back propagation neural network classifier. The neural network classifier achieved 100% classification rate over the images in publicly available dataset. The proposed method achieved 83% of sensitivity, 99% of specificity and 98% of accuracy in Mammogram Image Analysis Society (MIAS) dataset.

Ultrasound Thyroid Gland Volume Estimation: A Review by Maher Faik Esmaile, Mohammad Hamiruce Marhaban, Rozi Mahmud, M. Iqbal Saripan (85-90).
It is important to evaluate thyroid size before any thyroid surgical involvement to introduce the minimal invasive surgical procedures and the accuracy of the therapy dosage is directly based on the accuracy of thyroid volume estimation. This paper surveys the volume estimation of thyroid gland with Ultrasound (US). The techniques that have been used in the Ultrasound thyroid volume calculations have been discussed in addition to the data type, situations of application of each medical device with thyroid gland disorder or disease. Besides that, the comparison between the Ultrasound and some other medical devices that are used in the thyroid volume estimation has been made. We conclude by selecting several papers which have presented original ideas that: first, MRI and CT have the highest accuracy in the thyroid gland volume estimation but the Ultrasound still widely and robust procedure that are used for thyroid volume measurements in the routine clinical setting. Second, the automated calculation for the thyroid size improved the estimation results especially for the Ultrasound; third, there are no studies that applied these calculations on the large multinodular goiters because of the US probe foot pad limitation.

The strength of positron emission tomography (PET) lies in the application of its underlying technological and molecular biology advancements to clinical practice and pharmacological development. Currently, PET is used to identify malignant pulmonary nodules, evaluate mediastinal disease, detect distant metastasis, assess response to therapy, and identify novel oncologic drug targets. Over the last decade, PET has also increasingly influenced the management of locally advanced nonsmall cell lung cancer (NSCLC) by directing chemoradiation and surgical treatment. Although the majority of clinical trials and practice have used 2-18F- fluoro-2-deoxy-D-glucose (FDG) as the radiopharmaceutical agent for evaluating NSCLC, novel radiolabeled tracers, radiopharmaceutical agents, and targeted immunotherapy drugs are actively being investigated to improve the treatment of NSCLC. This review focuses on the utility of PET for diagnosing malignant pulmonary nodules, staging the extent of disease, and evaluating immuno-oncology therapies in locally advanced NSCLC.

Can Virtual Bronchoscopy be a Complementary Method for Fiberoptic Bronchoscopy? by Mehtap Beker-Acay, Sevinc Sarinc-Ulasli, Ebru Unlu, Emre Kacar, Ersin Gunay, Nazan Okur, Cigdem Ozdemir, Aylin Yucel (99-104).
Objectives: Fiberoptic bronchoscopy is considered to be the 'gold standart' technique that allows direct visualization of the airway lumen and mucosa. It was found that virtual bronchoscopy is highly accurate in the detection of central airway stenosis and correlated closely with FOB in grading tracheobronchial stenosis. This study explores the utility of VB to evaluate various tracheobronchial lesions and grading stenosis using FOB as the standart of reference.

Methods: The population of this prospective study was 42 patients examined in our department between November 2013 and February 2014. The presence or absence of the endoluminal lesions, obstructive lesions, external compressions and mucosal changes were recorded. Also anatomical variations and diverticulas were noted on both VB and FOB. Sensitivity, specificity, positive and negative predictive values were calculated with 2x2 contingency tables.

Results: A sum of 1115 airway segment of 42 patients were evaluated. The sensitivity and specifity values for diagnosis of obstructive lesions was 88,8% and 96,7%, for detection of endoluminal lesions 66,6% and 92,3, % and for mucosal changes they were 41,1% and 96%. 2 of the patients had diverticula according to VB and couldn't be seen by FOB. All of the anatomical variations were depicted on both VB and FOB.
Conclusions: In selected patients whom cannot tolerate or have contraindications for FOB, VB can be used as an initial method. Future investigation by developing large scale studies will likely lead to wider acceptance of this method, including to its use as a complementary method with FOB.


Super-Resolution Based Enhancement of Cardiac MR Images by Salah-ud-Din Ayubi, Usama Ijaz Bajwa, Muhammad Waqas Anwar (105-113).
Medical imaging is one of the important and challenging areas of research in the field of image processing. Medical images play an important role in the field of medical science by giving support to the diagnostic process of a disease and in suggesting the treatment. Medical images produced by different acquisition devices are not of very high resolution and they require a very deep and critical analysis to diagnose a disease. In our work we have addressed the problem of very low resolution by improving the spatial quality of the images by applying super resolution (SR). Process of SR is further composed of two steps, namely image registration and image reconstruction. We have targeted Medical Resonance (MR) images and improved their spatial resolution, because MR is more capable of capturing the details of soft tissues. Our proposed algorithm consists of phases which deal with the image both in the spatial domain as well as the frequency domain. We have used the demons deformable image registration algorithm in the image registration phase and wavelet based method for image reconstruction. After reconstruction various qualitative and quantitative measures have been applied on the images. These measures clearly support the claim that the proposed technique has improved the visibility of 58% of the images, such that, reconstructed high resolution images have more resolving power as compared to the low resolution input images.

Typical and Atypical Imaging Findings of Abdominal Teratomas by Neslin Sahin, Mine Genc, Aynur Solak, Esin Kasap, Seyhan Yalaz, Ilhami Solak, Berhan Genc, Serap Karaarslan (114-123).
Teratomas are most commonly observed as lesions of ovarian origin. They can also be detected in extragonadal regions such as brain, face, neck, mediastinum, retroperitoneum, and sacrococcygeal region. Ovarian teratomas are usually in mature cystic form as benign, well-differentiated, and cystic lesions. Immature teratomas and monodermal teratomas (struma ovarii, carcinoid tumors and neural tumors) are rare forms. Mature cystic teratomas are usually diagnosed by ultrasound (US) and magnetic resonance (MR) imaging. On US, a variety of appearances including echogenic sebaceous material and calcification are observed. MR imaging can specifically demonstrate fat component by fat-saturation sequences. On the other hand, teratomas are usually incidentally detected on computed tomography (CT) and fat attenuation within a cyst is diagnostic. It may be difficult to characterize immature teratomas due to nonspesicific findings on US. However, CT and MR can provide diagnosis by identifying small foci of fat within a mass with irregular solid component containing coarse calcifications. A small proportion of mature cystic teratomas can undergo malignant transformation (carcinomas or sarcomas). The purpose of this paper is to review the imaging features of various types of abdominally located teratomas for differentiation and diagnosis.

The abdominal aortic aneurysm (AAA) associated with congenital pelvic kidney is a rare clinical finding. We present a case of an infrarenal AAA with associated congenital left pelvic kidney followed up for 5 years, which was managed by regular surveillance. We describe this case to assist physicians and radiologists to recognize small aneurysms by computed tomography angiography (CTA) with low radiation dose and low iodine dose. To the best of our knowledge, this case is the first report by using CTA with the combination of low-concentration contrast medium, low radiation dose and iterative reconstruction.

3D Reconstruction of a Primary Aortoenteric Fistula – Centerline Calculation and Measurements by Bojan V. Stimec, Jean H.D. Fasel, Dejan Ignjatovic (127-131).
Background: Flow dynamics are a significant factor in the development and possible rupture of the aneurysms. Therefore, it is close to reason to present detailed data on the realistic 3D surface and internal geometry of a primary aortoenteric fistula, i.e. of its fistulous tract.
Methods: A primary aortoenteric fistula in a patient with an abdominal aorta aneurysm underwent quantitative analysis by means of multimodal image processing, based on multi-detector computed tomographic angiography dataset. The medical computer-aided design presented a 3D model with the fistula as a branch of the abdominal aorta. A full array of centerline calculations was performed on this fistula, three levels of the aneurysm and on the superior mesenteric artery.
Results: The hydraulic properties of the fistulous tract were comparable to the superior mesenteric artery and even to the abdominal aorta aneurysm.
Conclusion: 2D and 3D geometry measurements provide insight into the flow dynamics of the primary aortoenteric fistula, relevant for surgical treatment.


A Modified Expectation of Maximization Method and its Application to Image Segmentation by Lahouaoui Lalaoui, Tayeb Mohamadi, Abdelhak Djaalab, Harag Abdelghani (132-137).
In this paper we describe a modified segmentation method applied to image. An EM algorithm is developed to estimate parameters of the Gaussian mixtures. Recently, researchers are focusing more on the study of expectation of maximization (EM) due to its useful applications in a number of areas, such as multimedia, image processing, pattern recognition and bioinformatics. The human visual system can often correctly interpret images that are of quality that they contain insufficient explicit information to do so. The difficulty is mainly due to variable brain structures, various MRI artifacts and restrictive body scanning methods. The IBSR image segmentation data set is used to compare and evaluate the proposed methods. In this paper, we propose a modified expectation of maximization (MEM) based on the properties of likelihood, while reducing number of iteration for a sick of fast converge to the center of cluster and your application to image segmentation. The experiments on real images show that: (1) our proposed approach can reduce the number of iterations, which leads to a significant reduction in the computational cost while attaining similar levels of accuracy. (2)The approach also works well when applied to image segmentation. A methodology for calculate is presented for making use the error between the ground truth, human-segmented image data sets to compare, develop and optimize image segmentation algorithms. This error measure is based on object-by-object comparisons of a segmented image and a ground-truth (reference) image. Experimental results for segmented images demonstrate the good segmentation performance of the proposed approach.