Bartosz Borucki

Speckle noise reduction of ultrasound images using Extra-Energy Reduction function

A new effective method for speckle noise reduction in ultrasound imaging is introduced in this paper. The proposed method is based on Extra-Energy Reduction (EER) function in the vector field, which is addressed by vector triangular formula. It is reasonable to decide that the calculated the energy of a pixel by vector triangular formula is less than the usual absolute distance energy of the same pixel then the pixel must be noisy. The noisy pixel with higher energy is moved to equilibrium situation by subtracting the extra energy for any type of functional operation (e.g. image segmentation, pattern recognition, objects classification). The implemented method reduces extra-energy from an image and provides a proper gray level distribution into the entire image. The technique demonstrates very impressive noise suppression without loss of important information compared to some existing state-of-the-art speckle denoising methods.

Error assessment and minimization in 4D motion tracking for functional orthopaedics diagnostics

Purpose Against traditional medical teaching, web-based medical learning tools offer a number of advantages, like timeand distance-independent learning [1]. There exist a few online medical learning platforms, e. g., SurgyTech, BioDigital Systems or Zygote Body. But the majority of these online platforms only provide movies of surgical interventions or they are often limited to a small number of cases. The anatomical situs is often reconstructed using 3D-modeling tools [2]. Real patient-specific anatomical and pathological conditions and selfassessment tools that provide immediate feedback to the learner are largely missing. In this article, we describe the development of a webbased medical learning tool: the LiverAnatomyExplorer. The system provides clinical 2D image data as well as interactive polygonal 3D models. In close collaboration with medical experts at the Asklepios Clinic Barmbek, Hamburg, Germany, we selected appropriate patient cases with liver tumors, metastases and vessel anomalies. We provide an integrated training package that includes diagnosis and surgical reports, clinical image data with colored overlays, easy-to-use interactive 3D models, surgical video clips and self-assessment exercises. Methods We started with a detailed requirements analysis with potential end users: 176 medical students and 19 clinicians filled an online questionnaire. It turned out that many students would benefit from clinical case collections with medical imagery and 3D representations of anatomical basics. Most of the clinical subjects (76 %, 10/13) are interested in providing interactive 3D graphics, clinical imagery and movies as complementary learning materials for trainees. The analysis also showed that 97 % (171/176) of the students and 54 % (7/13) of the medical experts never or rarely used 3D visualizations so far. We used MeVisLab to process and convert 13 real clinical cases. After identifying and segmenting the important structures, they need to be exported in a web-compatible format. In our case, the segmentations of liver, tumors and vascular structures have been performed by medical experts with deep anatomical knowledge at MeVis Distant Services, Bremen, Germany. In order to achieve a fast online access of anonymized patient data, we chose JPG as export image format. The resulting segmented areas are automatically exported as SVG files. For web-based rendering of 3D models, the size of the meshes is more crucial than for local use. The simplified surface mesh has an overall size of about 5–10 MB (ca. 100,000 polygons), which enables fast web-based rendering. The surface mesh is exported as a single X3D file. We use X3D, an ISO Web3D standard, since X3D files can be easily integrated and rendered in real-time, without any plugin, using WebGL and X3DOM [3]. The exported clinical 2D data and segmented areas are presented in our combined 2D/3D viewer based on HTML, SVG, JavaScript and WebGL (Fig. 1left/right). The WebGLrendered 3D scene can be rotated, panned and zoomed freely in the 3D viewer. Immediate customized feedback mechanisms are essential for students to gain knowledge and to monitor individual learning curves [4]. Therefore, we have integrated a multifunctional selfassessment tool. Beside typical textual multiple choice questions, we enhance the quiz to interactive 2D and 3D click answer options. Highquality surgical images or movies composed and annotated by an expert surgeon, can be used by the learner as additional learning material. Results We conducted an evaluation with 54 medical students (average age: 24, gender: 40 f/14 m, 7th to 9th semester) to investigate the user experience and learning aspects of the LiverAnatomyExplorer. After testing the web application, the subjects were asked to fill an online questionnaire (statements are scaled using a five-point likert scale (1 = ‘‘Strongly disagree’’ to 5 = ‘‘Strongly agree’’). The analysis of the user study showed that the website has a modern and attractive design (average of 3.94) and the navigation is self-explanatory (average of 4.22). The orientation guides in the 3D viewer and the tutorial were highlighted by many subjects to be very helpful. The learning aspects of the LiverAnatomyExplorer were also rated ‘‘good’’, with some exceptions. On the one hand, the subjects highly rated the reality of the learning contents (average of 3.98) and the knowledge gain due to the multi-modal, individual liver anatomy data (average of 3.96). On the other hand, it turned out that the questions concerning liver segments and vessel anomalies are too specialized for medical students at this education level. Conclusion Our architecture could be easily adapted to other organs, since we use X3D as free ISO exchange file format and WebGL as heterogeneous free rendering engine. Further interviews with medical experts are

Estimating Achilles Tendon Healing Progress with Convolutional Neural Networks.

Quantitative assessment of a treatment progress in the Achilles tendon healing process - one of the most common musculoskeletal disorders in modern medical practice - is typically a long and complex process: multiple MRI protocols need to be acquired and analysed by radiology experts for proper assessment. In this paper, we propose to significantly reduce the complexity of this process by using a novel method based on a pre-trained convolutional neural network. We first train our neural network on over 500 000 2D axial cross-sections from over 3 000 3D MRI studies to classify MRI images as belonging to a healthy or injured class, depending on the patient’s condition. We then take the outputs of a modified pre-trained network and apply linear regression on the PCA-reduced space of the features to assess treatment progress. Our method allows to reduce up to 5-fold the amount of data needed to be registered during the MRI scan without any information loss. Furthermore, we are able to predict the healing process phase with equal accuracy to human experts in 3 out of 6 main criteria. Finally, contrary to the current approaches to healing assessment that rely on radiologist subjective opinion, our method allows to objectively compare different treatments methods which can lead to faster patient’s recovery.

Augmented-reality computed tomography-guided transcatheter pacemaker implantation in dextrocardia and congenitally corrected transposition of great arteries

A 30-year-old man with congenitally corrected transposition of great arteries and dextrocardia was admitted for leadless transcatheter pacemaker implantation (MicraTM TPS, Medtronic, MN, USA). At the age of 4, he underwent patch repair of the ventricular septal defect complicated by complete atrioventricular block. Computed tomography angiography (CTA) using a Somatom Force scanner (Siemens, Erlangen, Germany) was performed for better visualization of the position, orientation and morphology of the subpulmonic ventricle (left ventricular morphology). Three-dimensional CTA reconstructions of the subpulmonic ventricle simulating the angulations of the C-arm in the catheterization laboratory were transmitted to a head-mounted computer (Google Glass, Google Inc, Mountain View, CA) for display in a mobile application equipped with a hands-free voice recognition system developed by the authors (Fig. 1A). The CTA-defined right anterior oblique 60 view served to guide Micra advancement through the right atrium toward the subpulmonic ventricular apex (Fig. 1B, C), whereas the anterior-posterior projection facilitated lead positioning relative to the lateral orientation of the septum (Fig. 1D, E). Several technical challenges can be encountered when implanting pacemakers in patients with congenital heart disease and dextrocardia. In the present case, both ventricles were inversed and rotated to the right, and identification of the most optimal stimulation site was restricted due to trabeculation of the subpulmonic ventricle and interventricular patch. Herein, augmented-reality glass was used for display of CTA datasets to define the most optimal fluoroscopic angles and guide pacemaker positioning. This case confirms the utility of augmented-reality technology to optimize complex percutaneous interventions in patients with altered fluoroscopic orientation.

Large Java arrays and their applications

All current implementations of Java Virtual Machines allow the creation of one-dimensional arrays of length smaller than 231 elements. In addition, since Java lacks true multidimensional arrays, most of numerical libraries use one-dimensional arrays to store multidimensional data. With the current limitation, it is not possible to store volumes of size larger than 12903. On the other hand, the data from scientific simulations or medical scanners continuously grow in size and it is not uncommon to go beyond that limit. This work addresses the problem of maximal size of one-dimensional Java arrays. JLargeArrays is a Java library of one-dimensional arrays that can store up to 263 elements. Performance comparison with native Java arrays and Fastutil library shows that JLargeArrays is the fastest solution overall. Possible applications in Java collections as well as numerical and visualization frameworks are also discussed.

Assessment of anisotropic denoiser enhanced cone beam CT for patient dose reduction

Biopsy is commonly used to confirm cancer diagnosis when radiologically indicated. Given the ability of PET to localize malignancies in heterogeneous tumors and tumors that do not have a CT correlate, PET/CT guided biopsy may improve the diagnostic yield of biopsies. To facilitate PET/CT guided needle biopsy, we developed a workflow that allows us to bring PET image guidance into the interventional CT suite. In this abstract, we present SlicerPET, a user-friendly workflow based module developed using open source software libraries to guide needle biopsy in the interventional suite.