Image processing plays an important role in medical image analysis. The most popular methods for image processing and analysis are very resource hungry, which leads to some disadvantages in their applications even on a powerful desktop computers. On the other side, modern mobile devices are equipped with powerful processors with an efficient instruction architecture. This lead to better performance per watt than a desktop CPUs. This work investigates the performance of a widely used medical analysis algorithm implemented on a modern mobile devices and desktop CPU. The results obtained with ARM NEON instructions show speed improvements up to 2 times. As this research shows mobile devices cannot yet compete with powerful desktop CPUs, even with using highly optimizations or multiple threads. In the last part of the paper conclusions are drawn for acceptable image input and parameter sizes.

An accurate and efficient computer-aided mammography diagnosis system plays an important role as a second opinion to assist radiologists. Finding an accurate and robust computer-aided diagnosis system for classification of the abnormalities in the mammograms as malignant or benign still remains a challenge in the digital mammography. In this paper, a fully autonomous classification system is presented and it consists of the three stages. The input Regions of Interest (ROIs) are obtained using an efficient Otsu's N thresholding and further subjected to a number of preprocessing stages. After preprocessing stage, from the ROIs, a group of 32 Zernike moments with different orders and iterations have been extracted. These moments have been applied to the neural network classifier. The experimental results show that the proposed algorithm is efficient comparing to the ground truth table given in the Mammography Image Analysis Society (MIAS) database.

This paper deals with the design of a novel spectrometer of fast neutrons in nuclear safeguards applications based on the liquid organic scintillator EJ-309 with materials of different thickness surrounding the detector. The investigation was performed on the simulated data obtained by the MCNPX-PoliMi numerical code based on the Monte Carlo method. Among the various materials (polyethylene, iron, aluminum, and graphite) investigated as layers around the scintillator, polyethylene and iron have shown the most promising characteristics for evaluation of fast neutron energy spectra. The simulated pulse height distributions were summed up for each energy bin in the neutron energy range between 1 MeV and 15 MeV in order to obtain better counting statistics. The unfolded results for monoenergetic neutron sources obtained by a first order of Tikhonov regularization and non-linear neural network show very good agreement with the reference data while the evaluated spectra of neutron sources continuous in energy follow the trend of the reference spectra. The possible advantages of a novel spectrometer include a less number of input data for processing and a less sensitivity to the noise compared to the scintillation detector without surrounding materials.

We present a new approach to regularize the displacement field of the accelerated Demons registration algorithm. The accelerated Demons algorithm uses Gaussian smoothing to penalize oscillatory motion in the displacement fields during registration. This regularization approach is often applied and ensures a smooth deformation field. However, when registering images with discontinuities in their motion field such as from organs sliding along the chest wall, the assumption of a smooth deformation field is invalid. In this work, we propose using total variation based smoothing that is known to better retain the discontinuities in the deformation field. The proposed approach is a first step towards automatically recovering breathing induced organ motion with good accuracy.

Image enhancement plays an important role in different research fields such as medical image analysis. Since the same computations are usually performed on many image elements, those computations can be easily parallelized. Modern Graphics Processing Units (GPUs) are capable for doing many tasks in parallel. However, improving running times on GPUs usually leads to a loss of floating point precision. In this paper we evaluate the impact of GPU hardware implemented native functions on three GPUs, and one Central Processing Unit (CPU). As an example, the bilateral filter with built-in and native math functions was implemented and used for smoothing noisy brain Magnetic Resonance Images (MRI). For all experiments widely used error metrics were calculated. Experiments shows that native versions improve running times significantly (up to 155 times). As expected precision is lower for the measures which include a lot additions without normalization.

An excessive power consumption characterizes all industries, even IT. Today, different methods of energy savings are in use, due to the pollution of the environment and the greenhouse effect. One of the newest methods in the IT is green IT that is directly connected with the virtualization of hardware resources. This study illustrates the procedure of hardware virtualization using real-world example and the analyses of the results in power consumption before and after the virtualization.

In this paper we present a new approach for the smoothing of displacement field in the nonrigid accelerated demons algorithm for a discontinuities MRI image registration. accelerated demons algorithm use Gaussian smoothing filter to penalize oscillatory motion in the displacement fields during the registration, which has significant impact on registration quality. Simulated images show that Gaussian smoothing cannot properly handle these discontinuities. For this reason we introduce diffusion-like smoothing which shows promising results, and start as base for our further research on the topic of discontinuous image registration. This significant problem impedes recovering of organ motion in the case for example persons breathing cycle.

We present a hierarchical approach to elastic registration of medical images which use thin-plate splines approximation method. The thin-plate splines approximation method takes into account the landmark localization errors as anisotropic landmark errors. This inclusion is very important in clinical applications, because the landmark localization is always prone to error. Experimental results show that application of this TPS method can improve the registration result of the hierarchical elastic registration approach.