Deep Convolutional Neural Networks for left ventricle segmentation.

Left ventricle (LV) segmentation is crucial for quantitative cardiac function analysis. Manual segmentation of the endocardium and epicardium is highly cumbersome; physicians limit delineation to the end-diastolic and end-systolic phases. A fully automated system could provide an analysis of cardiac morphology for all phases in a much shorter time. Most of the current LV segmentation methods are semi-automated and require error prone manual initialization. A fully-automated LV segmentation method would expedite the functional analysis of the LV, reduce subjectivity and improve patient experience. We automatically segment the LV wall in cardiac MRI images with a Deep Convolutional Neural Network (DCNN). This algorithm first calculates the probability of a pixel belonging to the LV wall or background and then generates a label based on those probabilities without manual initialization. We then compare these results to the results obtained with another DCNN initialization method using Gabor filters. With Gabor DCNN we obtain an accuracy of 0.97, specificity of 0.984, sensitivity of 0.841 and mean accuracy of 0.902. This shows that Gabor filters perform better than random filters in the DCNN for LV segmentation.

Prediction of regulatory pathways using mRNA expression and protein interaction data: application to identification of galactose regulatory pathway.

We propose a novel technique that constructs gene regulatory networks from DNA microarray data and gene-protein databases and then applies Mason rule to systematically search for the most dominant regulators of the network. The algorithm then recommends the identified dominant regulator genes as the best candidates for future knock-out experiments. Actively choosing the genes for knock-out experiments allows optimal perturbation of the pathway and therefore produces the most informative DNA microarray data for pathway identification purposes. This approach is more practically advantageous in analysis of large pathways where the time and cost of DNA microarray data experiments can be reduced using the proposed optimal experiment design. The proposed method was successfully tested on the galactose regulatory network.

Robust watermarking in non-ROI of medical images based on DCT-DWT.

Increasing demand and utilization of telemedicine require transmission of medical information and images over internet. Since authenticity of received images is crucial and patient's information should be included with minimum changes in images, robust watermarking schemes are needed. In this paper, we propose a robust watermark method that embeds patient's information outside the region of interest (ROI) in medical image. In order to find appropriate regions for embedding, we use saliency as a means of measuring importance of regions and find blocks having minimum overlap with the ROI. The algorithm employs wavelet transform and also discrete cosine transform (DCT) domains in the embedding stage and redundantly embeds watermark to increase robustness against possible alterations. Moreover, voting is utilized in the extraction phase. Experimental results show the efficiency of the proposed method and better results are obtained as compared to comparable methods with same size of the watermarked data.

Automatic detection of melanoma using broad extraction of features from digital images.

Automatic and reliable diagnosis of skin cancer, as a smartphone application, is of great interest. Among different types of skin cancers, melanoma is the most dangerous one which causes most deaths. Meanwhile, melanoma is curable if it were diagnosed in its early stages. In this paper we propose an efficient system for prescreening of pigmented skin lesions for malignancy using general-purpose digital cameras. These images can be captured by a smartphone or a digital camera. This could be beneficial in different applications, such as computer aided diagnosis and telemedicine applications. It could assist dermatologists, or smartphone users, evaluate risk of suspicious moles. The proposed method enhances borders and extracts a broad set of dermatologically important features. These discriminative features allow classification of lesions into two groups of melanoma and benign. This method is computationally appropriate as a smartphone application. Experimental results show that our proposed method is superior in diagnosis accuracy compared to state-of-the-art methods.

Hardware image assessment for wireless endoscopy capsules.

Wireless capsule endoscopy is a new technology in the realm of telemedicine that has many advantages over the traditional endoscopy systems. Transmitted images should help diagnosis of diseases of the gastrointestinal tract. Two important technical challenges for the manufacturers of these capsules are power consumption and size of the circuitry. Also, the system must be fast enough for real-time processing of image or video data. To solve this problem, many hardware designs have been proposed for implementation of the image processing unit. In this paper we propose an architecture that could be used for the assessment of endoscopy images. The assessment allows avoidance of transmission of medically useless images. Hence, volume of data is reduced for more efficient transmission of images by the endoscopy capsule. This is done by color space conversion and moment calculation of images captured by the capsule. The inputs of the proposed architecture are RGB image frames and the outputs are images with converted colors and calculated image moments. Experimental results indicate that the proposed architecture has low complexity and is appropriate for a real-time application.

Vessel extraction in X-ray angiograms using deep learning.

Coronary artery disease (CAD) is the most common type of heart disease which is the leading cause of death all over the world. X-ray angiography is currently the gold standard imaging technique for CAD diagnosis. These images usually suffer from low quality and presence of noise. Therefore, vessel enhancement and vessel segmentation play important roles in CAD diagnosis. In this paper a deep learning approach using convolutional neural networks (CNN) is proposed for detecting vessel regions in angiography images. Initially, an input angiogram is preprocessed to enhance its contrast. Afterward, the image is evaluated using patches of pixels and the network determines the vessel and background regions. A set of 1,040,000 patches is used in order to train the deep CNN. Experimental results on angiography images of a dataset show that our proposed method has a superior performance in extraction of vessel regions.

Melanoma detection by analysis of clinical images using convolutional neural network.

Melanoma, most threatening type of skin cancer, is on the rise. In this paper an implementation of a deep-learning system on a computer server, equipped with graphic processing unit (GPU), is proposed for detection of melanoma lesions. Clinical (non-dermoscopic) images are used in the proposed system, which could assist a dermatologist in early diagnosis of this type of skin cancer. In the proposed system, input clinical images, which could contain illumination and noise effects, are preprocessed in order to reduce such artifacts. Afterward, the enhanced images are fed to a pre-trained convolutional neural network (CNN) which is a member of deep learning models. The CNN classifier, which is trained by large number of training samples, distinguishes between melanoma and benign cases. Experimental results show that the proposed method is superior in terms of diagnostic accuracy in comparison with the state-of-the-art methods.

Automatic segmentation of multimodal brain tumor images based on classification of super-voxels.

Despite the rapid growth in brain tumor segmentation approaches, there are still many challenges in this field. Automatic segmentation of brain images has a critical role in decreasing the burden of manual labeling and increasing robustness of brain tumor diagnosis. We consider segmentation of glioma tumors, which have a wide variation in size, shape and appearance properties. In this paper images are enhanced and normalized to same scale in a preprocessing step. The enhanced images are then segmented based on their intensities using 3D super-voxels. Usually in images a tumor region can be regarded as a salient object. Inspired by this observation, we propose a new feature which uses a saliency detection algorithm. An edge-aware filtering technique is employed to align edges of the original image to the saliency map which enhances the boundaries of the tumor. Then, for classification of tumors in brain images, a set of robust texture features are extracted from super-voxels. Experimental results indicate that our proposed method outperforms a comparable state-of-the-art algorithm in term of dice score.

CT image segmentation in traumatic brain injury.

Traumatic brain injury (TBI) is a major cause of disability and death. Speed and accuracy are vital in diagnosing TBI for which computer-aided imaging analysis may speedup and improve the efficiency of diagnosis and help reduce mortality, long-term complications, and the associated costs. However, developing such a system is challenging due to some factors such as the inherent noise associated with obtaining the images, artifacts and quality of the images. An automated system that can preliminary identify, localize and quantify the imaging features of TBI would be beneficial in guiding real-time clinical diagnosis as well as for quality assurance. In this paper we propose an automated system to segment the hematoma region from CT images. The proposed method first performs denoising and image enhancement and then by developing a Gaussian mixture model, segmentation is carried out. We show the performance of the system by comparing the results with ground truth generated by specialists.

Region of interest extraction for lossless compression of bone X-ray images.

For few decades digital X-ray imaging has been one of the most important tools for medical diagnosis. With the advent of distance medicine and the use of big data in this respect, the need for efficient storage and online transmission of these images is becoming an essential feature. Limited storage space and limited transmission bandwidth are the main challenges. Efficient image compression methods are lossy while the information of medical images should be preserved with no change. Hence, lossless compression methods are necessary for this purpose. In this paper, a novel method has been proposed to eliminate the non-ROI data from bone X-ray images. Background pixels do not contain any valuable medical information. The proposed method is based on the histogram dispersion method. ROI is separated from the background and it is compressed with a lossless compression method to preserve medical information of the image. Compression ratios of the implemented results show that the proposed algorithm is capable of effective reduction of the statistical and spatial redundancies.

Adaptive stereo medical image watermarking using non-corresponding blocks.

Today with the advent of technology in different medical imaging fields, the use of stereoscopic images has increased. Furthermore, with the rapid growth in telemedicine for remote diagnosis, treatment, and surgery, there is a need for watermarking. This is for copyright protection and tracking of digital media. Also, the efficient use of bandwidth for transmission of such data is another concern. In this paper an adaptive watermarking scheme is proposed that considers human visual system in depth perception. Our proposed scheme modifies maximum singular values of wavelet coefficients of stereo pair for embedding watermark bits. Experimental results show high 3D visual quality of watermarked video frames. Moreover, comparison with a compatible state of the art method shows that the proposed method is highly robust against attacks such as AWGN, salt and pepper noise, and JPEG compression.

Robust catheter identification and tracking in X-ray angiographic sequences.

Coronary artery disease (CAD) is one of the major causes of death worldwide. Today X-ray angiography is a standard method for CAD diagnosis. Usually, the quality of these images is not good enough. Noise, camera and heart motions, non-uniform illumination and even the presence of catheter are sources of quality degradation. The existence of catheter can produce difficulties in vessel extraction methods because catheter is structurally similar to arteries. In this paper we propose a fully automatic method for catheter detection and tracking during the whole angiography sequence. In this method with a vesselness map, we smooth each frame using guided filter. The catheter is detected in the first frame using Hough transform. We then fit a second order polynomial on the catheter and accurately track it throughout the sequence. Our method is tested on 25 X-ray angiography sequences where a precision of 0.9597 is achieved.

Five-year follow-up of prophylactic vena cava filters in high-risk trauma patients.

OBJECTIVE: To assess the short- and long-term outcomes of vena cava filter (VCF) placement for prophylaxis against pulmonary embolism in patients at high risk due to trauma. DESIGN AND SETTING: Case series at a level I trauma center. PATIENTS: Patients were considered for prophylactic VCF placement if they met 1 of the injury criteria--spinal cord injuries with neurologic deficit, severe fractures of the pelvis or long bone (or both), and severe head injury--and had a contraindication to anticoagulation. INTERVENTION: Vena cava filters were placed percutaneously by the interventional radiologists when the acute trauma condition was stabilized following admission. MAIN OUTCOME MEASURES: Filter tilt of 14 degrees or more, strut malposition, insertion-related deep vein thrombosis, pulmonary embolism, or inferior vena cava patency. RESULTS: There were 132 prophylactic VCFs placed. A 3.1% rate of insertion-related deep vein thrombosis occurred, all of which were asymptomatic. Filter tilt occurred in 5.5% of patients and strut malposition in 38%. Three cases of pulmonary embolism (1 fatal) occurred in a prophylactic VCF, and all patients had either filter tilt or strut malposition. The risk of pulmonary embolism developing was higher in those patients with filter tilt or strut malposition than in those who did not have these complications (6.3% vs 0%; P=.05; Fisher exact test). The 1-, 2-, and 3-year inferior vena cava patency rates (+/-SD) were 97%+/-3%. CONCLUSIONS: Prophylactic VCF can be placed safely with an acceptable rate of insertion-related deep vein thrombosis and long-term inferior vena cava patency. Patients with prophylactic VCF remain at risk for pulmonary embolism if the filter is tilted 14 degrees or more or has strut malposition. In such patients, consideration should be given to placing a second filter.

Arterial embolization in the chest.

Transcatheter therapy is an effective technique in the treatment of vascular abnormalities in the chest. Pulmonary arteriovenous malformations, associated with the clinical sequellae of dyspnea, stroke, brain abscess, and hemoptysis, can be treated by transcatheter embolization with metallic coils or occlusion balloons. The results of treatment are excellent, with improvement seen in symptoms and physiologic parameters. The bronchial arteries and systemic nonbronchial collateral arteries supplying the tracheobronchial tree are most frequently involved in massive hemoptysis, a serious disorder with associated high rates of morbidity and mortality. Transcatheter embolization of these arteries is both safe and effective, requiring a thorough understanding of the arterial anatomy and technique of embolization.

Signal processing for functional analysis of protein mutants.

We present a new technique to predict the functional behaviour of protein mutants using their primary sequence. The amino acid sequences of proteins are used to form biochemical signals. Signal processing techniques are used to extract functionally meaningful measures from these signals. These computational measures are then used to distinguish functional classes of mutant proteins.

New observations in the sonographic evaluation of intrauterine contraceptive devices.

The in vitro sonographic appearance of the four currently available intrauterine contraceptive devices (IUDs) now used in the United States and Canada were analyzed and compared with in vivo scans. While plastic IUD limbs demonstrated entrance-exit echoes in all scan planes, copper IUDs showed posterior reverberation when optimal gain settings were used. Regardless of the transducer used, posterior acoustic shadowing was only seen when IUD limbs were scanned perpendicular to their long axis and at the junction of long and short limbs. The authors feel that these refined ultrasound characteristics will be helpful in accurately identifying IUDs within the uterus.

PIP: This study analyzed the in vitro sonographic appearance of 4 IUDs Lippes Loop, Progestasert, Tatum T, and Copper 7 and compared this with in vivo scans. The following ultrasound characteristics were identified as useful in confirming the presence and type of an in utero IUD: 1) entrance-exit echoes in all plastic limbs; 2) reverberation artifact from the metal component of all IUDs at proper gain settings; 3) short axis limb shadowing in all IUDs, especially the plastic limbs; and 4) shadowing at the junction between long and short limbs. Regardless of composition, all IUDs shadow when the limbs are imaged in short axis but not in long axis. When scanned in either axis, there is shadowing only at the point where the arms join. Copper-wound IUDs show repetitive posterior echoes or reverberation. When scanned in vivo, the reverberation artifact can be more easily demonstrated with optimum and higher gain settings. Low-level internal echoes, presumably due to the barium-progesterone mixture, were observed between the entrance-exit echo of the Progestasert IUD. In addition, when imaged in long axis, there was more posterior attenuation of sound with Progestasert than other IUDs without frank shadowing.

Transjugular intrahepatic portosystemic shunt for bleeding stomal varices associated with chronic portal vein occlusion: long-term angiographic, hemodynamic, and clinical follow-up.

The efficacy of the transjugular intrahepatic portosystemic shunt procedure (TIPS) in controlling refractory hemorrhage from stomal varices at the mucocutaneous junction has been previously described, but the durability of this procedure for this indication is unknown. Conservative therapy may be unsuccessful in preventing recurrent hemorrhage. Some authors believe that portosystemic shunting is the treatment of choice for patients with bleeding stomal varices who are good surgical candidates, because of the low incidence of recurrent bleeding and the longest survival. We report the 39-month angiographic and hemodynamic follow-up, and the 48-month clinical follow-up of a patient with refractory hemorrhage from stomal varices and coexistent chronic portal vein occlusion successfully treated with a TIPS procedure.

Nonsurgical treatment of acute iatrogenic renal artery injuries occurring after renal artery angioplasty and stenting.

OBJECTIVE: The objective of this study was to determine the success of the nonsurgical treatment of acute iatrogenic renal artery injuries that occur after renal artery angioplasty and stenting at a tertiary referral center. MATERIALS AND METHODS: During a 5-year period, 212 patients (308 renal arteries) underwent percutaneous transluminal angioplasty or stent dilatation of the renal artery. Through a retrospective review of medical and radiology records, we determined that 13 of these patients suffered iatrogenic renal artery injuries. RESULTS: The renal arterial complication rates were 4.2% per artery treated and 6.1% per patient treated. All 13 patients were successfully treated nonsurgically. Five patients with acute rupture of the renal artery were treated immediately with balloon tamponade or with placement of an additional stent or stent-graft. Six patients suffered acute thrombotic occlusion; five were successfully treated with thrombolysis, and one was successfully treated without thrombolysis by the placement of an additional stent. Presumed distal guidewire perforation caused subcapsular hematoma in one patient and a perirenal and pararenal hematoma in another; both were successfully treated with conservative management. During the clinical follow-up period (mean period, 19 months), one patient required long-term hemodialysis. No other patients required additional treatment. CONCLUSION: The nonsurgical treatment of acute iatrogenic renal artery injuries occurring after renal artery angioplasty and stenting can be successful and may obviate additional surgery.