Unsupervised deep learning based approach to temperature monitoring in focused ultrasound treatment.

Temperature monitoring in ultrasound (US) imaging is important for various medical treatments, such as high-intensity focused US (HIFU) therapy or hyperthermia. In this work, we present a deep learning based approach to temperature monitoring based on radio-frequency (RF) US data. We used Siamese neural networks in an unsupervised way to spatially compare RF data collected at different time points of the heating process. The Siamese model consisted of two identical networks initially trained on a large set of simulated RF data to assess tissue backscattering properties. To illustrate our approach, we experimented with a tissue-mimicking phantom and an ex-vivo tissue sample, which were both heated with a HIFU transducer. During the experiments, we collected RF data with a regular US scanner. To determine spatiotemporal variations in temperature distribution within the samples, we extracted small 2D patches of RF data and compared them with the Siamese network. Our method achieved good performance in determining the spatiotemporal distribution of temperature during heating. Compared with the temperature monitoring based on the change in radio-frequency signal backscattered energy parameter, our method provided more smooth spatial parametric maps and did not generate ripple artifacts. The proposed approach, when fully developed, might be used for US based temperature monitoring of tissues.

Comparison of the radial and brachial artery flow-mediated dilation in patients with hypertension.

BACKGROUND: Blood flow-mediated dilation (FMD) is a noninvasive assessment of vascular endothelial function in humans. The study of the FMD in hypertensive (HT) patients is an important factor supporting the recognition of the early mechanisms of cardiovascular pathologies, and also of the pathogenesis related to hypertension. OBJECTIVES: To investigate whether FMD measured on the radial artery (FMD-RA) using high-frequency ultrasounds can be used as an alternative to FMD assessed with the lower frequency system on the brachial artery in patients with HT. MATERIAL AND METHODS: The simultaneous measurements of FMD-RA and FMD measurements in the brachial artery (FMD-BA) were performed on 76 HT patients using 20 MHz and 7-12 MHz linear array probes, and were compared to the FMD measured in healthy groups. All quantitative data are presented as mean ± standard deviation (SD); the p-values of the normality and tests for variables comparisons are listed. The agreement of the FMD-RA and FMD-BA in HT patients was assessed with the Bland-Altman method, and using the intraclass correlation coefficient (ICC). In some statistical calculations, the FMD-RA values were rescaled by dividing them by a factor of 2. RESULTS: The mean FMD-RA and FMD-BA in HT patients were 5.16 ±2.18% (95% confidence interval (95% CI): [4.50%, 5.82%]) and 2.13 ±1.12% (95% CI: [1.76%, 2.49%]), respectively. The FMD-RA and FMD-BA values of HT patients were significantly different than those in respective control groups. The p-values of Mann-Whitney-Wilcoxon tests were less than 0.05. The Bland-Altman coefficient for both measurement methods, FMD-RA and FMD-BA, was 3%, and the ICC was 0.69. CONCLUSIONS: Our findings show that FMD-RA, supplementary to FMD-BA measurements, can be used to assess endothelial dysfunction in the group of HT patients. In addition, the FMD-RA measurements met the criteria of high concordance with the FMD-BA measurements.

DOES FLOW-MEDIATED DILATION NORMALIZATION FOR BASE-SCALED SHEAR RATE IMPROVE ITS VALUE IN CORONARY ARTERY DISEASE DIAGNOSTICS?

The article presents a new normalization of flow-mediated dilation (FMD) in the radial artery, taking into account the parameter BSSR being equal to the ratio of the basal shear rate (BS) measured before the cuff inflation and post occlusive shear rate (SR). The in vivo usefulness of the new normalization algorithm was evaluated in two groups of patients. In group I, comprising 15 healthy volunteers, the normalized FMD/SR was (3.19 ± 1.4)•10-4, while in group II, comprising 13 patients with stable coronary artery disease (CAD), it was (1.02 ± 0.76)•10-4. We calculated almost 50% larger difference between the average values after normalizing FMD/BSSR. Specifically, the FMD/BSSR was equal to 28 ± 9.40 in group I and 6.01 ± 3.74 in group II. The prediction of CAD patients based on FMD/SR values had a sensitivity of 83.3% and a specificity of 84.6%, whereas the prediction of CAD patients based on the FMD/BSSR values revealed 100% sensitivity and specificity. These results confirm the usefulness of the novel normalization algorithm of the FMD in differentiation of normal patients from those with stable CAD.

Ultrasound assessment of the conversion of sound energy into heat in tissue phantoms enriched with magnetic micro- and nanoparticles.

PURPOSE: Nowadays, the improvement of ultrasonic hyperthermia therapy is often achieved by adding hard particles to the sonicated medium in order to increase the heating efficiency. The explanation of the phenomenon of ultrasonic heating still requires testing on tissue mimicking materials (TMMs), enriched with particles of different sizes and physical properties. Our goal was to determine, by comparing their quantitative acoustic properties, which TMMs, with magnetic micro- or nanoparticles, convert more ultrasonic energy into heat or which of the particles embedded in the agar gel act as more effective thermal sonosensitizers. METHODS: We manufactured a pure agar gel and an agar gel with the addition of magnetic micro- or nanoparticles in two proportions of 8 and 16 mg/ml. Ultrasound quantitative techniques, the broadband reflection substitution technique and backscattered spectrum analysis were used to characterize the samples by speed of sound (SOS), frequency-dependent attenuation, and backscattering coefficients. The integrated backscattering coefficients were also calculated. The quantitative parameters, scattering, and attenuation coefficients of ultrasound in phantoms with micro- and nanoparticles were estimated. Based on the attenuation and scattering of ultrasound in the samples, the ultrasonic energy absorption, which determines the heating efficiency, was evaluated. Additionally, the temperature increase during sonication of the phantoms by an ultrasonic beam was directly measured using thermocouples. RESULTS: The density of the materials with nanoparticles was higher than for the materials with microparticles with the same fractions of particles. The SOS for all materials ranged from 1489 to 1499 m/s. The attenuation in the whole frequency range (3-8 MHz) was higher for the materials with nanoparticles than for the materials with microparticles. For the materials with the lower content (8 mg/ml) of particles, the attenuation coefficient was 0.2 dB/(MHz cm). For the 16 mg/ml concentration of nanoparticles and microparticles, the attenuation coefficients were 0.66 and 0.45 dB/(MHz cm), respectively. The value of backscattering coefficient in the whole frequency range was greater for the materials with microparticles than for the materials with nanoparticles. The values of the integrated backscattering coefficient were 0.05 and 0.08 1/m for the materials with nanoparticles and 0.46 and 0.82 1/m for the materials with microparticles and concentrations of 8 and 16 mg/ml, respectively. The rates of temperature increase in the first 3 s due to ultrasonic heating were higher for the materials with nanoparticles than for the materials with microparticles. CONCLUSIONS: Based on acoustical measurements, we confirmed that all materials can be used as tissue phantoms in the study of ultrasonic hyperthermia, as their properties were in the range of soft tissue properties. We found that the nanoparticle-doped materials had greater attenuation and smaller scattering of ultrasound than the materials with microparticles, so absorption in these materials is greater. Thus, the TMMs with nanoparticles convert more acoustic energy into heat and we conclude that magnetic nanoparticles are more effective thermal sonosensitizers than microparticles. This conclusion is confirmed by direct measurement of the temperature increase in the samples subjected to sonification.

20-MHz Ultrasound for Measurements of Flow-Mediated Dilation and Shear Rate in the Radial Artery.

A high-frequency scanning system consisting of a 20-MHz linear array transducer combined with a 20-MHz pulsed Doppler probe was introduced to evaluate the degree of radial artery flow-mediated dilation (FMD [%]) in two groups of patients after 5 min of controlled forearm ischemia followed by reactive hyperemia. In group I, comprising 27 healthy volunteers, FMD (mean ± standard deviation) was 15.26 ± 4.90% (95% confidence interval [CI]: 13.32%-17.20%); in group II, comprising 17 patients with chronic coronary artery disease, FMD was significantly less at 4.53 ± 4.11% (95% CI: 2.42%-6.64%). Specifically, the ratio FMD/SR (mean ± standard deviation), was equal to 5.36 × 10-4 ± 4.64 × 10-4 (95% CI: 3.54 × 10-4 to 7.18 × 10-4) in group I and 1.38 × 10-4 ± 0.89 × 10-4 (95% CI: 0.70 × 10-4 to 2.06 × 10-4) in group II. Statistically significant differences between the two groups were confirmed by a Wilcoxon-Mann-Whitney test for both FMD and FMD/SR (p <0.01). Areas under receiver operating characteristic curves for FMD and FMD/SR were greater than 0.9. The results confirm the usefulness of the proposed measurements of radial artery FMD and SR in differentiation of normal patients from those with chronic coronary artery disease.

Determining temperature distribution in tissue in the focal plane of the high (>100 W/cm(2)) intensity focused ultrasound beam using phase shift of ultrasound echoes.

In therapeutic applications of High Intensity Focused Ultrasound (HIFU) the guidance of the HIFU beam and especially its focal plane is of crucial importance. This guidance is needed to appropriately target the focal plane and hence the whole focal volume inside the tumor tissue prior to thermo-ablative treatment and beginning of tissue necrosis. This is currently done using Magnetic Resonance Imaging that is relatively expensive. In this study an ultrasound method, which calculates the variations of speed of sound in the locally heated tissue volume by analyzing the phase shifts of echo-signals received by an ultrasound scanner from this very volume is presented. To improve spatial resolution of B-mode imaging and minimize the uncertainty of temperature estimation the acoustic signals were transmitted and received by 8 MHz linear phased array employing Synthetic Transmit Aperture (STA) technique. Initially, the validity of the algorithm developed was verified experimentally in a tissue-mimicking phantom heated from 20.6 to 48.6 °C. Subsequently, the method was tested using a pork loin sample heated locally by a 2 MHz pulsed HIFU beam with focal intensity ISATA of 129 W/cm(2). The temperature calibration of 2D maps of changes in the sound velocity induced by heating was performed by comparison of the algorithm-determined changes in the sound velocity with the temperatures measured by thermocouples located in the heated tissue volume. The method developed enabled ultrasound temperature imaging of the heated tissue volume from the very inception of heating with the contrast-to-noise ratio of 3.5-12 dB in the temperature range 21-56 °C. Concurrently performed, conventional B-mode imaging revealed CNR close to zero dB until the temperature reached 50 °C causing necrosis. The data presented suggest that the proposed method could offer an alternative to MRI-guided temperature imaging for prediction of the location and extent of the thermal lesion prior to applying the final HIFU treatment.

Stochastic modelling of the eukaryotic heat shock response.

The heat shock response (HSR) is a highly evolutionarily conserved defence mechanism allowing the cell to promptly react to elevated temperature conditions and other forms of stress. It has been subject to intense research for at least two main reasons. First, it is considered a promising candidate for deciphering the engineering principles underlying regulatory networks. Second, heat shock proteins (main actors of the HSR) play crucial role in many fundamental cellular processes. Therefore, profound understanding of the heat shock response would have far-reaching ramifications for the cell biology. Recently, a new deterministic model of the eukaryotic heat shock response has been proposed in the literature. It is very attractive since it consists of only the minimum number of components required by any functional regulatory network, while yet being capable of biological validation. However, it admits small molecule populations of some of the considered metabolites. In this paper a stochastic model corresponding to the deterministic one is constructed and the outcomes of these two models are confronted. The aim with this comparison is to show that, in the case of the heat shock response, the approximation of a discrete system with a continuous model is a reasonable approach. This is not always the truth, especially when the numbers of molecules of the considered species are small. By making the effort of performing and analysing 1000 stochastic simulations, we investigate the range of behaviour the stochastic model is likely to exhibit. We demonstrate that the obtained results agree well with the dynamics displayed by the continuous model, which strengthens the trust in the deterministic description. A proof of the existence and uniqueness of the stationary distribution of the Markov chain underlying the stochastic model is given. Moreover, the obtained view of the stochastic dynamics and the performed comparison to the outcome of the continuous formulation provide more insight into the dynamics of the heat shock response mechanism.