The Atlantic salmon maturation process has been studied for decades to increase the quantity and quality of the production in farming facilities. An important topic in this context is the salmon egg maturation process. Ultrasound imaging is considered an effective tool for monitoring the egg development stage of salmon, but manual inspection is time-consuming and dependent on operator experience. We propose a method for automated monitoring of the egg maturation stage in salmon using deep learning, providing complimentary decisions on egg morphology. A segmentation network was developed to solve the challenge of separating and measuring individual eggs in the ovary. The segmentation part was combined with a classification network to determine the maturation stage of the eggs. Our model was able to segment eggs and classify their development stage with over 88% accuracy, outperforming established methods designed for similar tasks. A real-time application was developed which provided an estimation of size and maturity stage while scanning. The egg state estimation showed potential for replacing manual evaluations and can enable fully automatic evaluation of maturation in Atlantic salmon.
Non-pulsatile cardiopulmonary bypass (CPB) may induce microvascular dysregulation. In piglets, we compared ocular surface microcirculation during pulsatile versus continuous flow (CF) bypass.Ocular surface microcirculation in small tissue volumes (~0.1 mm3 ) at limbus (high metabolic rate) and bulbar conjunctiva (low metabolic rate) was examined in a porcine model using computer assisted video microscopy and diffuse reflectance spectroscopy, before and after 3 and 6 h of pulsatile (n = 5 piglets) or CF (n = 3 piglets) CPB. Functional capillary density, capillary flow velocity and microvascular oxygen saturation were quantified.At limbus, velocities improved with pulsatility (p < 0.01) and deteriorated with CF (p < 0.01). In bulbar conjunctiva, velocities were severely reduced with CF (p < 0.01), accompanied by an increase in capillary density (p < 0.01). Microvascular oxygen saturation decreased in both groups.Ocular surface capillary densities and flow patterns are better preserved with pulsatile versus CF during 6 h of CPB in sleeping piglets.
The extraction of oil and gas typically involves drilling boreholes through various subsurface strata. Abrupt variations in the formation properties can severely compromise the borehole stability if the drilling process is not adequately controlled. Certain events can result in influx/efflux of fluids through fissures or fractures in the borehole wall. This causes a destabilization in the drilling operation known as a "kick" which, if not controlled, may lead to serious well incidents. Modern drilling operations are continuously monitored using various logging-while-drilling (LWD) tools including ultrasonic imaging to assure a high degree of control over the drilling process. However, none of these tools directly monitor fluid flow from influx/efflux events in the borehole while drilling. This work explores the use of pulsed-wave (PW) Doppler ultrasound for mapping fluid influx through fractures in the borehole wall using ultrasonic transducers similar to those used in conventional LWD tools. This paper aims to define optimal parameters for PW Doppler specific to its application in LWD. The use of short pulse-lengths (2-4 wavelengths) for velocity estimation is explored. Mean velocity estimates with a standard deviation of 0.05 m/s is demonstrated using pulse-length of 4 wavelengths and the lower limit on radial flow velocities estimates was found to be around 0.2 m/s (15 ml/s from ⌀10 mm orifice) for LWD conditions. Furthermore, imaging the fracture geometry with a high SNR of about 30 dB is demonstrated using the power of PW Doppler spectra and compared with conventional pulse-echo amplitude imaging used in LWD tools. Estimation of lateral dimensions of the fracture using the power of PW Doppler spectra with a resolution approaching the size of the point-spread function of the probe has also been demonstrated.
A simulation study of transmit ultrasound beams from several transducer configurations is conducted to compare second-harmonic imaging at 3.5 MHz and 11 MHz. Second- harmonic generation and the ability to suppress near field echoes are compared. Each transducer configuration is defined by a chosen f-number and focal depth, and the transmit pressure is estimated to not exceed a mechanical index of 1.2. The medium resembles homogeneous muscle tissue with nonlinear elasticity and power-law attenuation. To improve computational efficiency, the KZK equation is utilized, and all transducers are circular-symmetric. Previous literature shows that secondharmonic generation is proportional to the square of the transmit pressure, and that transducer configurations with different transmit frequencies, but equal aperture and focal depth in terms of wavelengths, generate identical second-harmonic fields in terms of shape. Results verify this for a medium with attenuation f 1 . For attenuation f 1.1 , deviations are found, and the high frequency subsequently performs worse than the low frequency. The results suggest that high frequencies are less able to suppress near-field echoes in the presence of a heterogeneous body wall than low frequencies.
The work presented in this thesis is focused on developing a method for imaging of nonlinear scattering from stiff particles using dual-frequency band pulses. The pulse complexes are comprised of a low-frequency manipulation pulse and a high-frequency imaging pulse where the the two pulses overlap in time and there is a frequency relationship of 1:8-10. It may be shown that the polarity of the nonlinear scattering follows the polarity of the low-frequency pulse, while linear scattering does not. By transmitting two such dual-frequency band pulses in each beam direction where the polarity of the low-frequency pulse is inverted from the first to the second, nonlinear scattering may be detected. The low-frequency pulse not only manipulates the scattering but also the propagation of the high-frequency imaging pulse. These nonlinear propagation effects will mask the nonlinear scattering and must be corrected for in order to suppress the linear scattering and detect the nonlinear scattering.In the first paper of this thesis, the nonlinear propagation effects using confocal low-frequency and high-frequency beams are investigated in a water tank setup. A dual-frequency band annular array, where the low-frequency element is place behind the high-frequency element, to form a stack, was used. When the high-frequency pulse is short compared to the low-frequency pulse period, the nonlinear propagation effects can be approximated by a nonlinear propagation delay and frequency shift. It is shown how the delay and frequency shift increases close to linearly with increasing manipulation pressure and how axis the profiles of the high-frequency beam are affected. On transmit, the size relationship between the low and high-frequency apertures can be varied, and it is shown how the nonlinear propagation effects is dependent on the array setup.By transmitting an unfocused low-frequency beam together with a focused high-frequency beam, the position of the high-frequency pulse relative to the low-frequency pulse can be kept close to constant over the whole imaging region. By placing the imaging pulse at the peak of the manipulation pulse, the frequency shift due to nonlinear propagation can be minimized. In the second paper, the suppression of linear scattering using such a beam setup and only correcting for the propagation delay is investigated. Applying a low-frequency pressure of 85 to 500 kPa, the linear scattering could be suppressed 35 to 17 dB. It is shown that there is an amplitude difference between the first and second received pulse which is due to diffraction differences of the first and second beam. Since the low-frequency beam is unfocused, the manipulation pressure will vary over the focused high-frequency beam and distort the spherical focusing. This distortion will be different for the first and second beam and produce different diffraction of the two beams, which will yield an amplitude difference. Frequency shift due to nonlinear propagation will also affect the diffraction but it is indicated that the nonlinear aberration is the dominating factor.In the third paper three different beamforming strategies for dual-frequency band imaging is investigated; 1. Focused low freq. + Focused high freq., 2. Unfocused low freq. + Focused high freq. and 1. Unfocused low freq. + Unfocused high freq. The nonlinear propagation delay and frequency shift are estimated and predicted based on the estimated low-frequency manipulation pressure experienced by the high-frequency pulse. There is good accordance between the estimated and predicted values until diffraction becomes significant. When diffraction becomes significant, differences in diffraction between the first and second pulse will also introduce a frequency shift and delay other than that generated by the nonlinear manipulation pressure. Differences in the pulse form of the first and second pulse is thus not only due to manipulation of the propagation of the high-frequency pulse by the low-frequency, but also by differences in diffraction.The nonlinear propagation and scattering are generated by equal processes but are different in the way that nonlinear propagation is an accumulative effect while scattering is a local effect. In the last part of the thesis the difference between nonlinear propagation and scattering is investigated using simulations, where the bandwidth of the high-frequency pulse relative to the center frequency of the manipulation pulse is varied. It is shown that when the high-frequency pulse is shorter in time than one period of the low-frequency pulse, the nonlinear propagation and scattering becomes different and the nonlinear scattering can be detected if the nonlinear propagation is corrected for.The correction of nonlinear propagation can be in the form of a filter, and a method for estimating this filter is also presented in the last part. Based on statistical analysis of the filter, it is shown that the average suppression of linear scattering using the proposed correction filter, is dependent on the homogeneity of the relation between the first and second pulse over the receive beam. Said in another way; if this relation is not constant over the receive beam, the optimal correction for a given signal segment is dependent on the unknown distribution of scatterers within the beam.The level of suppression of linear scattering using the proposed filter method will be dependent on the transmit beam setup. A simulation study where the effect of aperture size relationship between the low- and high-frequency beams and f-number of the high-frequency beam on the level of suppression of linear scattering is presented. In order to achieve a high degree of homogeneity, the diffraction of the HF and LF beams should be equal, which is not trivial to achieve in a medium with attenuation. Choosing the aperture sizes in order for the fresnel numbers to be equal for the two beams was thought to yield the optimal setup, but as attenuation affects the low and high-frequency pulses differently, this is not necessarily true. The level of suppression of linear scattering increases when the the high-frequency aperture is increased, making the beam narrower, but the low-frequency aperture must also be increased accordingly.
<p>Ultrasound image quality is of utmost importance for a clinician to reach a correct diagnosis. Conventionally, image quality is evaluated using metrics to determine the contrast and resolution. These metrics requires localization of specific regions and targets in the image such as a region of interest (ROI), a background region, and or, a point scatterer. Such objects can all be difficult to identify in in-vivo images, especially for automatic evaluation of image quality in large amounts of data.</p> <p>Using a matrix array probe, we have recorded a Very Large cardiac Channel data Database (VLCD) to evaluate coherence as an in-vivo image quality metric. The VLCD consists of 33 280 individual image frames from 538 recordings of 106 patients. We also introduce a Global Image Coherence (GIC), an in-vivo image quality metric that does not require any identified ROI since it is defined as an average coherence value calculated from all the data pixels used to form the image, below a pre-selected range. The GIC is shown to be a quantitative metric for in-vivo image quality when applied to the VLCD. We demonstrate, on a subset of the dataset, that the GIC correlates well with the conventional metrics contrast ratio (CR) and the generalized contrast-to-noise ratio (gCNR) with R=0.74 (p<0.005) and R=0.62 (p<0.005) respectively. </p> <p>There exists multiple methods to estimate the coherence of the received signal across the ultrasound array. We further show that all coherence measures investigated in this study are highly correlated (R>0.9, p<0.001) when applied to the VLCD. Thus, even though there are differences in the implementation of coherence measures, all quantify the similarity of the signal across the array and can be averaged into a GIC to evaluate image quality automatically and quantitatively.</p>
In piglets we tested the applicability of digital video microscopy and diffuse reflectance spectroscopy for non-invasive assessments of limbal and bulbar conjunctival microcirculation. A priori we postulated that the metabolic rate is higher in limbal as compared to bulbar conjunctiva, and that this difference is reflected in microvascular structure or function between the two locations. Two study sites, Oslo University Hospital (OUH), Norway and Cleveland Clinic (CC), USA, used the same video microscopy and spectroscopy techniques to record limbal and bulbar microcirculation in sleeping piglets. Recordings were analyzed with custom-made software to quantify functional capillary density, capillary flow velocity and microvascular oxygen saturation in measuring volumes of approximately 0.1 mm3. The functional capillary density was higher in limbus than in bulbar conjunctiva at both study sites (OUH: 18.1 ± 2.9 versus 12.2 ± 2.9 crossings per mm line, p < 0.01; CC: 11.3 ± 3.0 versus 7.1 ± 2.8 crossings per mm line, p < 0.01). Median categorial capillary blood flow velocity was higher in bulbar as compared with limbal recordings (CC: 3 (1–3) versus 1 (0–3), p < 0.01). Conjunctival microvascular oxygen saturation was 88 ± 5.9% in OUH versus 94 ± 7.5% in CC piglets. Non-invasive digital video microscopy and diffuse reflectance spectroscopy can be used to obtain data from conjunctival microcirculation in piglets. Limbal conjunctival microcirculation has a larger capacity for oxygen delivery as compared with bulbar conjunctiva.
Simulations of iterative transmit-beam aberration correction using a time-delay and amplitude filter have been performed to study the convergence of such a process. Aberration in medical ultrasonic imaging is usually modeled by arrival-time and amplitude fluctuations concentrated on the transducer array. This is an approximation of the physical aberration process, and may be applied to correct the transmitted signal using a time-delay and amplitude filter. Estimation of such a filter has proven difficult in the presence of severe aberration. Presented here is an iterative approach, whereby a filter estimate is applied to correct the transmit-beam. This beam induces acoustic backscatter better suited for arrival-time and amplitude estimation, thus facilitating an improved filter estimate. Two correlation-based methods for estimating arrival-time and amplitude fluctuations in received echoes from random scatterers were employed. Aberration was introduced using eight models emulating aberration produced by the human abdominal wall. Results show that only a few iterations are needed to obtain corrected transmit-beam profiles comparable to those of an ideal aberration correction filter. Furthermore, a previously developed focusing criterion is found to quantify the convergence accurately.
The integrity of cement in cased boreholes is typically evaluated using well logging. However, well logging results are complex and can be ambiguous, and decisions associated with significant risks may be taken based on their interpretation. Cement evaluation logs must therefore be interpreted by trained professionals. To aid these interpreters, we propose a system for automatically interpreting cement evaluation logs, which they can use as a basis for their own interpretation. This system is based on deep convolutional neural networks, which we train in a supervised manner using a dataset of around 60 km of interpreted well log data. Thus, the networks learn the connections between data and interpretations during training. More specifically, the task of the networks is to classify the bond quality (among 6 ordinal classes) and the hydraulic isolation (2 classes) in each 1m depth segment of each well based on the surrounding 13 m of well log data. We quantify the networks' performance by comparing over all segments how well the networks' interpretations of unseen data match the reference interpretations. For bond quality, the networks' interpretation exactly matches the reference 51.6% of the time and is off by no more than one class 88.5% of the time. For hydraulic isolation, the interpretations match the reference 86.7% of the time. For comparison, a random-guess baseline gives matches of 16.7%, 44.4%, and 50%, respectively. We also compare with how well human reinterpretations of the log data match the reference interpretations, finding that the networks match the reference somewhat better. This may be linked to the networks learning and sharing the biases of the team behind the reference interpretations. An analysis of the results indicates that the subjectivity inherent in the interpretation process (and thereby in the reference interpretations we used for training and testing) is the main reason why we were not able to achieve an even better match between the networks and the reference.
