Andrzej Nowicki

Bio: Andrzej Nowicki is an academic researcher from Polish Academy of Sciences. The author has contributed to research in topics: Ultrasonic sensor & Binary Golay code. The author has an hindex of 27, co-authored 233 publications receiving 2070 citations. Previous affiliations of Andrzej Nowicki include Drexel University & Military Medical Academy.

Transfer learning with deep convolutional neural network for liver steatosis assessment in ultrasound images.

Abstract: The nonalcoholic fatty liver disease is the most common liver abnormality. Up to date, liver biopsy is the reference standard for direct liver steatosis quantification in hepatic tissue samples. In this paper we propose a neural network-based approach for nonalcoholic fatty liver disease assessment in ultrasound. We used the Inception-ResNet-v2 deep convolutional neural network pre-trained on the ImageNet dataset to extract high-level features in liver B-mode ultrasound image sequences. The steatosis level of each liver was graded by wedge biopsy. The proposed approach was compared with the hepatorenal index technique and the gray-level co-occurrence matrix algorithm. After the feature extraction, we applied the support vector machine algorithm to classify images containing fatty liver. Based on liver biopsy, the fatty liver was defined to have more than 5% of hepatocytes with steatosis. Next, we used the features and the Lasso regression method to assess the steatosis level. The area under the receiver operating characteristics curve obtained using the proposed approach was equal to 0.977, being higher than the one obtained with the hepatorenal index method, 0.959, and much higher than in the case of the gray-level co-occurrence matrix algorithm, 0.893. For regression the Spearman correlation coefficients between the steatosis level and the proposed approach, the hepatorenal index and the gray-level co-occurrence matrix algorithm were equal to 0.78, 0.80 and 0.39, respectively. The proposed approach may help the sonographers automatically diagnose the amount of fat in the liver. The presented approach is efficient and in comparison with other methods does not require the sonographers to select the region of interest.

In vitro ultrasound-mediated leakage from phospholipid vesicles.

Abstract: Interest in using ultrasound energy in wound management and intracellular drug delivery has been growing rapidly. Development and treatment optimization of such non-diagnostic applications requires a fundamental understanding of interactions between the acoustic wave and phospholipid membranes, be they cell membranes or liposome bilayers. This work investigates the changes in membrane permeation (leakage mimicking drug release) in vitro during exposure to ultrasound applied in two frequency ranges: “conventional” (1 MHz and 1.6 MHz) therapeutic ultrasound range and low (20 kHz) frequency range. Phospholipids vesicles were used as controllable biological membrane models. The membrane properties were modified by changes in vesicle dimensions and incorporation of poly(ethylene glycol) i.e. PEGylated lipids. Egg phosphatidylcholine vesicles with 5 mol% PEG were prepared with sizes ranging from 100 nm to 1 μm. Leakage was quantified in terms of temporal fluorescence intensity changes observed during carefully controlled ultrasound ON/OFF time intervals. Custom-built transducers operating at frequencies of 1.6 MHz (focused) and 1.0 MHz (unfocused) were used, the I spta of which were 46.9 W/cm 2 and 3.0 W/cm 2 , respectively. A commercial 20 kHz, point-source, continuous wave transducer with an I spta of 0.13 W/cm 2 was also used for comparative purposes. Whereas complete leakage was obtained for all vesicle sizes at 20 kHz, no leakage was observed for vesicles smaller than 100 nm in diameter at 1.6 or 1.0 MHz. However, introducing leakage at the higher frequencies became feasible when larger (greater than 300 nm) vesicles were used, and the extent of leakage correlated well with vesicle sizes between 100 nm and 1 μm. This observation suggests that physico-chemical membrane properties play a crucial role in ultrasound mediated membrane permeation and that low frequency (tens of kilohertz) ultrasound exposure is more effective in introducing permeability change than the “conventional” (1 MHz) therapeutic one. The experimental data also indicate that the leakage level is controlled by the exposure time. The results of this work might be helpful to optimize acoustic field and membrane parameters for gene or drug delivery. The outcome of this work might also be useful in wound management.

Skin imaging with high frequency ultrasound — preliminary results

Abstract: This study presents the detailed construction and the principle of performance of high frequency (HF) ultrasound scanner for skin examination. The aim of this study was to show a difference between diseased and healthy skin and to evaluate the usefulness of the scanner in monitoring of therapeutic efficacy of morphea and lichen sclerosus et atrophicus (LSA). We examined 48 patients aged between 15 and 64 years; 25 patients with plaque-type morphea, nine patients with linear morphea and 14 patients with LSA. In the course of 18 months all patients were examined before, during and after treatment. In 29 patients ultrasonographic evidence of regression (decreasing of the skin thickness) was observed, in eight patients ultrasound examination showed progression and in 12 patients no difference in the ultrasound scan before and after treatment could be shown. Our study shows that HF ultrasound scanner is suitable to differentiate between the healthy and diseased skin in morphea and LSA as well as to evaluate the treatment efficacy of these diseases.

Hydrophone spatial averaging corrections from 1 to 40 MHz

Abstract: The purpose of this study was to develop and experimentally verify a practical spatial averaging model for frequencies up to 40 MHz. The model is applicable to focused sources of circular geometry, accounts for the effects of hydrophone probe finite aperture, and allows calibration by substitution to be performed when the active elements of reference and tested hydrophone probes differ significantly. Several broadband sources with focal numbers between 3 and 20 were used to produce ultrasound fields with frequencies up to 40 MHz. The effective diameters of the ultrasonic hydrophone probes calibrated in the focal plane of the sources ranged from 150 to 500 /spl mu/m. Prior to application of the spatial averaging corrections, the hydrophones with diameters smaller than that of the reference hydrophone exhibited experimentally determined absolute sensitivities higher than the true ones. This discrepancy increased with decreasing focal numbers and increasing frequency. It was determined that the error was governed by the cross-section of the beam in the focal plane and the ratio of the effective diameters of the reference and tested hydrophone probes. In addition, the error was found to be reliant on the frequency-dependent effective hydrophone radius. After applying the spatial averaging correction, the overall uncertainty in the hydrophone calibration was on the order of /spl plusmn/1 dB. The model developed is being extended to be applicable to frequencies beyond 40 MHz, which are becoming increasingly important in diagnostic ultrasound imaging applications.

Introduction to ultrasound elastography.

Abstract: For centuries tissue palpation has been an important diagnostic tool. During palpation, tumors are felt as tissues harder than the surrounding tissues. The significance of palpation is related to the relationship between mechanical properties of different tissue lesions. The assessment of tissue stiffness through palpation is based on the fact that mechanical properties of tissues are changing as a result of various diseases. A higher tissue stiffness translates into a higher elasticity modulus. In the 90's, ultrasonography was extended by the option of examining the stiffness of tissue by estimating the difference in backscattering of ultrasound in compressed and non-compressed tissue. This modality is referred to as the static, compression elastography and is based on tracking the deformation of tissue subjected to the slowly varying compression through the recording of the backscattered echoes. The displacement is estimated using the methods of cross-correlation between consecutive ultrasonic lines of examined tissue, so calculating the degree of similarity of ultrasonic echoes acquired from tissue before and after the compression was applied. The next step in the development of ultrasound palpation was to apply the local remote tissue compression by using the acoustic radiation force generated through the special beam forming of the ultrasonic beam probing the tissue. The acoustic radiation force causes a slight deformation the tissue thereby forming a shear wave propagating in the tissue at different speeds dependent on the stiffness of the tissue. Shear wave elastography, carries great hopes in the field of quantitative imaging of tissue lesions. This article describes the physical basis of both elastographic methods: compression elastography and shear wave elastography.

Statistical optics

Abstract: Course Description This is an advanced course in which we explore the field of Statistical Optics. Topics covered include such subjects as the statistical properties of natural (thermal) and laser light, spatial and temporal coherence, effects of partial coherence on optical imaging instruments, effects on imaging due to randomly inhomogeneous media, and a statistical treatment of the detection of light. Development of this more comprehensive model of the behavior of light draws upon the use of tools traditionally available to the electrical engineer, such as linear system theory and the theory of stochastic processes.

Automatic Detection of Coronavirus Disease (COVID-19) Using X-ray Images and Deep Convolutional Neural Networks

Abstract: The 2019 novel coronavirus disease (COVID-19), with a starting point in China, has spread rapidly among people living in other countries, and is approaching approximately 34,986,502 cases worldwide according to the statistics of European Centre for Disease Prevention and Control. There are a limited number of COVID-19 test kits available in hospitals due to the increasing cases daily. Therefore, it is necessary to implement an automatic detection system as a quick alternative diagnosis option to prevent COVID-19 spreading among people. In this study, five pre-trained convolutional neural network based models (ResNet50, ResNet101, ResNet152, InceptionV3 and Inception-ResNetV2) have been proposed for the detection of coronavirus pneumonia infected patient using chest X-ray radiographs. We have implemented three different binary classifications with four classes (COVID-19, normal (healthy), viral pneumonia and bacterial pneumonia) by using 5-fold cross validation. Considering the performance results obtained, it has seen that the pre-trained ResNet50 model provides the highest classification performance (96.1% accuracy for Dataset-1, 99.5% accuracy for Dataset-2 and 99.7% accuracy for Dataset-3) among other four used models.

Ultrasonic arrays for non-destructive evaluation: A review

Abstract: An ultrasonic array is a single transducer that contains a number of individually connected elements. Recent years have seen a dramatic increase in the use of ultrasonic arrays for non-destructive evaluation. Arrays offer great potential to increase inspection quality and reduce inspection time. Their main advantages are their increased flexibility over traditional single element transducer methods, meaning that one array can be used to perform a number of different inspections, and their ability to produce immediate images of the test structure. These advantages have led to the rapid uptake of arrays by the engineering industry. These industrial applications are underpinned by a wide range of published research which describes new piezoelectric materials, array geometries, modelling methods and inspection modalities. The aim of this paper is to bring together the most relevant published work on arrays for non-destructive evaluation applications, comment on the state-of the art and discuss future directions. There is also a significant body of published literature referring to use of arrays in the medical and sonar fields and the most relevant papers from these related areas are also reviewed. However, although there is much common ground, the use of arrays in non-destructive evaluation offers some distinctly different challenges to these other disciplines.

Automatic detection of coronavirus disease (COVID-19) using X-ray images and deep convolutional neural networks.

Abstract: The 2019 novel coronavirus disease (COVID-19), with a starting point in China, has spread rapidly among people living in other countries and is approaching approximately 101,917,147 cases worldwide according to the statistics of World Health Organization. There are a limited number of COVID-19 test kits available in hospitals due to the increasing cases daily. Therefore, it is necessary to implement an automatic detection system as a quick alternative diagnosis option to prevent COVID-19 spreading among people. In this study, five pre-trained convolutional neural network-based models (ResNet50, ResNet101, ResNet152, InceptionV3 and Inception-ResNetV2) have been proposed for the detection of coronavirus pneumonia-infected patient using chest X-ray radiographs. We have implemented three different binary classifications with four classes (COVID-19, normal (healthy), viral pneumonia and bacterial pneumonia) by using five-fold cross-validation. Considering the performance results obtained, it has been seen that the pre-trained ResNet50 model provides the highest classification performance (96.1% accuracy for Dataset-1, 99.5% accuracy for Dataset-2 and 99.7% accuracy for Dataset-3) among other four used models.

Measurement of blood flow by ultrasound: accuracy and sources of error.

Abstract: Doppler ultrasound has now developed to the point where the rate of flow of blood in a given vessel can be measured with appropriate instrumentation. The theoretical basis of Doppler flow measurement is reviewed in this paper, with particular emphasis on the potential and actual sources of error. Three distinct approaches are identified, and the strengths and weaknesses of each discussed. The separate errors involved in estimating the vessel cross-sectional area, the angle of approach, and the Doppler shift are analyzed, together with the question of the uniformity of scattering from the blood. In vivo and in vitro tests of the accuracy obtained using a number of Doppler flow measuring instruments are then reviewed. It is concluded that the Doppler methods are capable of good absolute accuracy when suitably designed equipment is used in appropriate situations, with systematic errors of 6% of less. There are, however, considerable random errors, attributable primarily to errors in measuring the cross-sectional area and the angle of approach. Repeating the measurement of flow several times and averaging the results can reduce these random errors to an acceptable level.