Current Medical Imaging - Volume 14, Issue 2, 2018

Background: Liver Stiffness (LS) assessed by Sonoelastography (SE), has been demonstrated as reliable non-invasive indicator of liver fibrosis stage in patients with Chronic Liver Diseases (CLD). Sonoelastography performs best in ruling-out cirrhosis (F=4) and ruling-in signifficant fibrosis (F≥2). However, it is insufficiently accurate to replace endoscopy for detection of Esophageal Varices (EV), being able to only ruling-out large EV. LS ≥ 25 kPa by Transient Elastography (TE) is considered highly suggestive for the presence of Clinically Significant Portal Hypertension (CSPH). Higher liver and spleen stiffness have been asociated with adverse clinical outcomes in CLD. Two-dimensional shear wave elastography (2D-SWE), the latest developed SE method, allows both visualisation and quantification of liver elasticity in real time superimposed over B-mode ultrasound image. Discussion: Meta-analysis of studies with Supersonic Shear Imaging (SSI) revealed comparable performance of this 2D-SWE to TE in fibrosis staging, with AUROCs 0.85 for F≥2 (LS cut-off 8.04 kPa) and 0.93 for F=4 (LS cut-off 11.12 kPa). Few studies reported very good performance of 2DSWE (SSI) to rule-in CSPH (AUROCs 0.79-0.95; LS cut-offs 15-25 kPa). While conflicting data exist with respect to its performance in predicting the presence of EV, prognostic utility of 2D-SWE (SSI) was demonstrated in a single study that reported 3.4-fold (P=0.026) higher risk of adverse outcome in patients with baseline LS≥21.5 kPa followed over 28 months. Conclusion: 2D-SWE (SSI) might be used to stage liver fibrosis in CLD, identify patients with compensated cirrhosis under risk of adverse outcomes and potentially stratify risk of having CSPH and EV.

Background: As a non-invasive MR technique, Diffusion weighted magnetic resonance imaging (DWI) can provide functional information by measuring Apparent Diffusion Coefficient (ADC) value. It is sensitive to the small changes of water molecules diffusion in the MR scanning and the ADC value reflexes the water molecules diffusion in the local tissue. Discussion: In recent years, DWI is widely used in the diagnosis and differential diagnosis of hepatic tumors (including monoexponential and biexponential model DWI), especially for benign and malignant hepatic tumors diagnosis and differential diagnosis. In addition, the research of ADC value before and after treatment in hepatic tumors can be applied to predict the efficacy and prognosis. In clinical, not only conventional MRI, MR spectroscopy or dynamic enhancement, but also DWI was needed to acquire for receiving reliable results. However, during the malignant and benign hepatic neoplasm, the mean ADC value was overlapped and it make the diagnosis value of DWI limited. Also, relatively low spatial resolution and difficulty in detecting small lesions are the shortages of DWI. Therefore, the conventional MRI sequences, dynamic contrast enhancement and DWI combined for the assessment of hepatic tumors diagnosis and treatment will be more valuable. Conclusion: The focus of this review is to apply the value of DWI in the diagnosis and treatment of hepatic tumors assessment.

Background: Skeletal muscle morphology plays an essential role in determining function. Ultrasonography has been widely adopted to evaluate the morphological parameters of different muscles in both static and dynamic circumstances. In recent years, numerous algorithms, especially automated algorithms, have been developed to extract morphological parameters from musculoskeletal ultrasound images. It is now possible to analyze muscle contraction dynamics not only using traditional techniques, such as Mechanomyography (MMG) and Electromyography (EMG), but also using ultrasonography. Conclusion: This review summarizes the important progress in these processing algorithms for different morphological parameters and concludes with recent results and forward-looking issues.

Background: RGD peptides are small peptides containing arginine-glycine-aspartic (Arg- Gly-Asp) acid triple-peptide motif and can specifically bind to integrin receptor on the cell surface. Discussion: The integrin receptors, especially alpha(v) beta(3), are highly expressed on the surface of activated endothelial cells and tumor cells while displaying low expression on mature endothelial cells. Molecular imaging can achieve the imaging of the biological processes at cellular and molecular levels in vivo and help in qualitative and quantitative research, making the evaluation of the expression levels of integrin possible in vivo. Conclusion: Therefore, RGD peptides show great potential in the study of specific imaging of tumor- induced angiogenesis. This study reviewed the recent research progress of RGD peptide– mediated imaging of tumor-induced angiogenesis, which can benefit the RGD peptide–targeted molecular imaging researches.

Background: Neonatal and paediatric encephalopathy can be a diagnostic challenge for clinicians. Potential aetiologies include hypoxic brain injury, stroke, infection, trauma, metabolic and electrolyte abnormalities, autoimmune conditions and drug ingestion. Radiology plays a key role in determining aetiology and, even when normal, directing further assessment. Conclusion: We present a review of the neuroradiological manifestations of neonatal and paediatric encephalopathies which will aide paediatricians and radiologists in their assessment of children with this condition.

Background: Craniocervical Dissections (CCD) are a crucial emergency state causing 20% of strokes in patients under the age of 45. Although DSA (digital substraction angiography) is regarded as the gold standard, noninvasive methods of CT, CTA and MRI, MRA are widely used for diagnosis. Aim: Our aim is to illustrate noninvasive imaging findings in CCD. Conclusion: Emphasizing on diagnostic pitfalls, limitations and mimicking diseases.

Background: Major reason of mortality all around the world is cardiac disorders. Eminent progress of medical physics for imaging and exceptional benefits of MRI makes it a frequently used tool for examination and diagnosis of suspected cardiac abnormalities. Discussion: This article discusses characteristics of cardiac MR short axis images. Secondly, this study explains existing semi-automatic and automatic techniques for segmenting left ventricle from short axis cardiac MR images. The survey deeply explains the performance of automatic techniques of left ventricle segmentation figuring their results and comparison with other techniques. The aim of automatic segmentation is to reproduce accurate images and highlight mortality and morbidity problems by exact assessment of cardiac left ventricle functions in the case of cardiac afflictions. Conclusion: The study focuses and combines the modern approaches and proposed techniques of CMR-LV segmentation which automatically analyze and evaluate segmented CMR images.

Background: Renal segmentation is a challenging procedure, and the difficulty of segmentation varies from one imaging modality to another. Various imaging modalities such as CT, MRI, US, etc. are currently utilized to evaluate and analyze kidney pathologies such as kidney cancer, and other related disorders. Techniques: Among all available imaging techniques, CT is usually preferred by radiologists because it provides high-resolution images with good anatomical details. Over the past decades, numerous kidney segmentation techniques have been proposed in the literature, and an updated survey intending to classifying and summarizing the presented studies related to kidney segmentation is required. Discussion: To the best of our knowledge, no survey paper was found in the literature that reviews the kidney segmentation techniques. This survey aims to provide an introduction to beginners to initiate their research in this field and provides a comprehensive review of the state of the art renal segmentation techniques that have been applied to CT imaging modality. The quantitative performance of existing techniques has also been evaluated. Conclusion: The research gaps found in the literature and possible avenues for future research have also been addressed.

Background: Cataract is a disease that occurs when alterations in the eye lens result in blurred vision. These alterations are caused due to protein accumulation in the lens of an eye. It can lead to the decrease in vision and some loss of eyesight. Methods: The detection of cataract is normally done using slit-lamp exam, retinal examination, refraction and visual acuity test. However, it is possible to detect cataract with image processing and machine learning techniques while making its clinical analysis quite easier. Discussion: This paper discusses different techniques for automatic categorization and classification of cataract. These techniques present considerable possibilities to lessen the load of qualified ophthalmologists. Conclusion: Cataract sufferers in underdeveloped areas can get benefit by automated cataract classification to recognize their cataract conditions at the right time.

Background: Alzheimer Disease (AD) represents a major threat to the lives of human beings. In fact, the disease should be detected at an early stage to maximize the chances of survival. Matarial and Methods: Hence, the use of computer means making the diagnostic procedure automatic called: Computer-Assisted Diagnosis (CAD). This procedure is used to assist radiologists in the analysis of the disease; the number of the affected persons continues to grow in recent decades. As to our work, we made a Computer-Assisted Diagnosis for detecting Alzheimer's disease in early step Mild Cognitive Impairment (MCI). Conclusion: Our system contains three parts: Preprocessing, segmentation and a classification step. For the pretreatment step we used the Non-Local Means Filter (NLMF), the deformable model Level Set in the segmentation step to extract the Cortex and Hippocampus. Our contribution is to improve the segmentation step: we determined a priori shape and an automatic position for the initialization. Also, we added a priori knowledge of the surface. For the classification, our method is based on Support Vector Machine (SVM). The proposed system yields 92.5% accuracy in the early diagnosis of the AD.

Background: Medical imaging plays a key role in detecting and diagnosing abnormal patterns from scanned images. The computer aided automatic detection of the brain tumor was proposed in this work using Adaptive Neuro Fuzzy Inference System (ANFIS) classifier. Methods: The proposed system has the following stages as noise reduction, Gabor transform, feature extraction and ANFIS classifier. The impulse noises in the brain images were detected and removed using directional filtering algorithm. Gabor transform transformed the spatial domain image into multi resolution image and further Pixel invariant, Local Binary Pattern (LBP) and Discrete Wavelet Transform (DWT) features were extracted from the Gabor transformed image and these features were given to the ANFIS classifier to classify the image as either normal and abnormal. Discussion: The morphological operations were then applied over the abnormal image to segment the tumor regions. Conclusion: The proposed system achieved 99.8%sensitivity, 99.7%specificity, and 99.8% accuracy for the brain tumor detection.

Background: Automated delineation of exact boundaries of nodules in Thoracic Computed Tomography (CT) images is one of the difficult tasks owing to the weak edges and fuzzy boundaries of nodules. Generally, the juxta-pleura nodules get missed due to an inaccurate extraction of lung field region. Methods: In this work, an attempt has been made to address such issues by developing an Intuitionistic Fuzzy domain Region-based Level Set Method named as IFRbLSM to detect pleura attached lung nodules in Thoracic CT images. In the proposed method, Intuitionistic fuzzy energy is incorporated into length regularization term of Region-based Level Set Method (RbLSM) to solve the problem of boundary leakage and a simple Lung field extraction algorithm is used as a preprocessing step to extract the lung field region accurately. Result: Due to inherent capability of Intuitionistic fuzzy sets in handling fuzzy boundaries, the proposed method improves the detection of juxta-pleural nodules both in terms of accuracy and time. The method has been tested and evaluated on the standardized Thoracic CT image dataset provided by Lung Image Database Consortium (LIDC). Performance has been evaluated in terms of True Positive Ratio, False Positive Ratio, False Negative Ratio, Tannimoto Coefficient, Dice Similarity Coefficient and time complexity. From experimental results, it has been observed that proposed IFRbLSM outperforms the Fuzzy domain Region-based Level Set Method (FRbLSM) in terms of all quantitative metrics. Conclusion: Moreover, IFRbLSM is computationally more efficient than FRbLSM.

Background: Most pattern recognition and data mining problems such as classification, clustering and retrieval rely on distance metric or similarity measure. The metrics used in such problems should be capable of reflecting the relationship among the data efficiently. Different features contribute differently to the discrimination of different classes of images. This is especially true in medical image classification where it is difficult to relate the radiological signs to image features and hence to determine the relative importance of the different image features. Methods: In this paper, we have proposed novel metrics that apply different weights on the deviations of the features. The weights have been used together with Euclidean distance and city block distance to derive weighted Euclidean distance (WED) and weighted city block distance (WCBD) respectively. Result: The weights are obtained using genetic algorithm for finding the optimal features for classification. The weights are representative of the relative significance of the features in the classification of diseases. The proposed adaptive weighted deviation based metrics (AWDMs) have been tested by using it in a computer aided diagnosis (CAD) system for diagnosis of lung disorders. Discussion: The AWDMs proposed in this work are capable of being tuned in accordance with the dataset used for training. The accuracy of the CAD system has been found to increase from 80.73% to 84.31% with the use of WED against Euclidean Distance and from 81.43% to 81.98% with the use of WCBD against city block distance. Conclusion: This establishes the improvement in performance with the usage of AWDMs.

Background: Ultrasound imaging is widely used for tumor detection and diagnosis. Feature extraction plays a critical role in the ultrasound-based computer-aided diagnosis system. Deep Polynomial Network (DPN) is a newly proposed deep learning algorithm, which also has the potential to learn for excellent representation from small dataset. Discussion: However, the final feature representation of DPN is the simple concatenation of the learned hierarchical features from different network layers, which essentially loses some properties exhibited by different network layers, and depresses the representative performance. Since the hierarchical features in DPN can be regarded as heterogeneous multi-view features, they can be effectively integrated by Multiple Kernel Learning (MKL) methods. Conclusion: In this work, we proposed a DPN and MKL based feature learning and classification framework (DPN-MKL) for ultrasound image based tumor diagnosis. The experimental results on breast ultrasound image dataset and prostate ultrasound image dataset show that DPN algorithm has superior performance to the commonly used deep learning algorithms, while the proposed DPNMKL framework outperforms all the single-view feature based algorithms.

Background: The role of Technetium-99 m (Tc-99m) labeled tracers has been discussed for the detection of viable myocardium. One of the methods that is considered to be used in the diagnosis of viability is gated Myocardial Perfusion Imaging (gMPI) conducted under the influence of Beta Blocker Drugs (BBD). Objective: The aim of this study is twofold: First, to evaluate the role of the Beta Blocker Drugs (BBD) on fixed perfusion defects on gated Myocardial Perfusion Imaging (gMPI) (myocardial viability) and second, to assess the role of BBD on the functional parameters of the left ventricle. Methods: 31 patients (28 men, 3 women) who were on BBD treatment and had fixed perfusion defects on their gMPI and submitted to fluorodeoxyglucose positron emission tomography/computed tomography (FDG PET/CT) were prospectively enrolled to the study. One week after the completion of the stress-rest gMPI, rest gMPI was repeated for each patient while they were on BBD treatment. All systolic and diastolic functional measurements, perfusion defect extent and size were obtained during both rest gMPI studies with and without BBD. The parameters derived from FDG PET/CT were compared to those that were derived from gMPI without BBD. Results: A statistically significant difference between the scores of both groups was not observed, except the Time to Peak Filling Rate (TTPFR). End-systolic/diastolic volumes derived from gated FDG PET/CT were significantly lower than the ones derived from the gMPI presumably due to higher spatial resolution of the FDG PET/CT while the study did not reveal a statistically significant difference between the ejection fraction values. Conclusion: According to our results, metoprolol does not seem to change functional parameters of the left ventricle that were detected by gMPI except TTPFR. Besides, it seems that metoprolol does not affect the size of the perfusion defect, and therefore, it does not have a role to demonstrate the presence of viability that was confirmed via FDG PET/CT.

Background: The metabolism of [18F]-FDG, an analogue of glucose molecule, is often evaluated by kinetic analysis. Methods: In this study, we compare the performance of the Iterative Two-stage (ITS) algorithm against the standard Weighted Nonlinear Least Squares (WNLS) and the Patlak analysis for PET imaging quantification in control and diseased rats. A full kinetic modelling and graphical analyses were performed on [18F]-FDG PET data (7 controls and 7 Myocardial Infarct (MI) rats). The evaluation of the identifiability of parameters and the goodness of the fit in nonlinear regression wereconducted by statistic measures i.e. the Coefficient of Variation (COV), Akaike Information Criterion (AIC) and the Model Selection Criterion (MSC). Bland-Altman analysis was used to assess the reproducibility of the net influx rate (Ki) constants in the mid-inferoseptal region. The Patlak Ki outcome values were compared to ITS Ki by calculating the percentage changes between their outcomes estimated values. Results: The WNLS and ITS show an excellent identifiability of the parameters (resp. 5.15≤COV≤8.92; 4.12≤COV≤ 10.20) and showed a good reproducibility of the Ki constants in controls. In contrast, in the MI group, WNLS revealed a poor identifiability in the mid-inferoseptal region (17.62≤COV≤38.15) whereas ITS revealed a good identifiability (10.84 ≤COV≤23.54). The Bland Altman analysis showed a large bias in the estimation of the Ki constants in the region of interest e.g. mid-inferoseptal, 0.053 ± 0.017 mL/ min/mL. The Patlak Ki values were underestimated by 6.8% and 7.4% compared to ITS Ki in controls and MI respectively.

Objective: To compare the accuracy of linear bone measurements on Cone Beam Computed Tomography (CBCT) images obtained using different tube currents. Methods: Twelve reference points were prepared by using burs and heated gutta-percha on each of the ten dry mandibles for horizontal and vertical measurements on anterior and posterior segments. Images were obtained with Kodak 9000 3D cbct system at 5 different tube currents (2 mA, 3.2 mA, 6.3 mA, 10 mA and 15 mA). Three radiologists did total of 1200 linear measurements. True lengths were determined with a digital caliper. Height and width measurements from each of the mA settings were compared to the gold standard using Bland and Altman plots. Overall comparison of the measurements was done using repeated measures ANOVA. Correlation between individual measurements of each observer was assessed with Pearson's correlation analysis. Results: No differences were found in the height and width measurements for both anterior (p>0.05) and posterior (p>0.05) measurements for different mA settings. Increase in mA caused an increase in the geometric accuracy of cbct images (p>0.05). Mean deviations from the gold standard ranged between - 0.7 and +0.12 mm among the lowest and highest tube settings. The correlation between observers' measurements ranged between 0.97 and 0.98 for increasing tube settings. Conclusion: Tube current may be reduced as low as 2mA without jeopardizing linear measurement accuracy.

Background: Non-rigid registration of medical images with intensity distortions is a difficult problem due to the change in pixel intensity. It is caused by contrast agent or intensity bias field. Methods: In some cases, this problem can be solved using Residual Complexity (RC) method. However, relative modification of parameter in residual complexity would result in completely different experimental effect. Another drawback is sensitivity to noise. To handle this problem, a new intensity-based similarity measure, Weighted Residual Complexity (WRC) has been proposed for effective medical image registration in this paper. Specifically, the local entropy image of two images is computed to be aligned respectively. Then, a weighting function using a function of the local entropy difference is modeled. The weighting function is used to weight the residual image in residual complexity adaptively. The residual image is defined as the difference between reference image and warped floating image. Results: The weighting function assigns smaller weight to residual image if the corresponding pixel value is larger in local entropy difference. The proposed technique was applied to simulative and real medical images. The contrast experiments were made with mutual information, diffeomorphic demons and residual complexity. Conclusion: Also, the analysis of experimental results was made qualitatively and quantitatively, which indicates that this new approach gives a better performance than diffeomorphic demons, mutual information and residual complexity.