By Marlene Iversen Halvorsrød, PhD Candidate at CIUS

Treatment of ischemic heart disease has improved considerably the last decades. Not only the treatment itself, but also timing of treatment has been of great importance for this success. Nevertheless, ischemic heart disease is still the leading cause of death in the world.

The heart muscle, as every other muscle in the body, needs oxygen to maintain proper function. The coronary arteries supply the heart muscle with oxygenated blood. Ischemic heart disease is caused by stenosis or totally occlusion of this arteries and thereby reduced amount of available oxygen. This lack of oxygen can lead to permanent heart muscle damage. The longer the muscle have no available oxygen, the worse the damage will be. Treatment of ischemic heart disease is based on opening stenotic or occluded arteries to improve the accessibility of oxygen. Either by invasive blocking or medical (fibrinolytic) treatment. Time from debut of symptoms to treatment is a crucial determinant for success. The shorter ischemic time, the less heart muscle damage. After 6 hours without oxygen, the majority of heart muscle cells are likely to suffer from irreversible damage.

Classical symptoms of ischemic heart disease as heart attack are chest pain with radiation to neck and shoulder, shortness of breath and sweating. The symptoms are often related to physical activity. But this stenotic or occluded artery manifest with a great variability of symptoms that makes the diagnostic process hard and sometimes time consuming.

In addition to a patient’s symptoms, the most important quick diagnostic tool is the electrocardiogram (ECG). ECG was first used in late 1800 and is still a cornerstone in diagnosing ischemic heart diseases. ECG records the electrical signals through the heart. We classify heart attacks by the look of the ECG. A patient with chest pain and a certain pattern in ECG called ST-elevation, usually has an occluded artery and will be treated immediately. In some cases, patients lack this typical pattern in ECG, but have the same occluded artery. This group of patients does not have the same algorithm for immediate treatment and the treatment can be delayed. The consequence is more damaged heart muscles than necessary.

A lot of effort is put down to solve this diagnostic puzzle. In our research group at CIUS (Centre of Innovative Ultrasound Solutions), NTNU (Norwegian University of Science and Technology) we try to take advantage of new advanced ultrasound technology to come closer to a solution. Ultrasound is routinely used in diagnosing ischemic heart disease today by looking at regional differences in how the heart muscle contract. To see this regional impairment, you need a trained eye and good image quality. This method is also highly subjective. We want to improve this diagnostic step by using new technology developed at NTNU with high frame rate imaging that gives better time resolution, tissue doppler that can measure the speed of the heart muscle in different regions, and 3D ultrasound.

We know that it is crucial that patients with blocked coronary artery get treatment immediately. We want to improve the way we select these patients. Medical technical solutions can take time to develop, but in time we hope to brake some barriers in this important issue concerning the world’s deadliest disease.

Yucel Karabiyik

Physical examination at the doctor’s office often involves palpation where the physician tries to feel the location, size, shape and stiffness of masses or organs in the body by touching. Sometimes, even the patients themselves can notice lumps or abnormalities in their body and refer to a physician. Examples include palpation for masses or abnormalities in breast, spleen, liver and thyroid. However, this process has its limitations. It is subjective, meaning that the diagnosis is highly dependent on the skills of the physician and it is not sensitive to small and deep tumors. Therefore, nowadays it is usually used for preliminary screening before referring the patient for further investigation.

Elastography is an imaging modality that can overcome these limitations. It is a quantitative method (as opposed to palpation) and the estimated tissue stiffness can sometimes be directly related to certain diseases or even stages of a disease. For instance, it has been found that the stiffness of liver estimated with elastography can be used for staging of liver fibrosis.

Figure 1

The method is based on generating very small amplitude displacements in the tissue and estimating how fast these displacements propagate in the form of waves. The stiffer the tissue, the faster the displacements propagate. The displacements are generated with actuators, which can be a loud speaker, an electrodynamic shaker or another kind depending on the application. The first step after creating the displacements is to image and estimate the displacement amplitudes. Diagnostic imaging techniques, magnetic resonance imaging (MRI) and medical ultrasound imaging are used for recording these displacements. Both techniques have their advantages and limitations. MRI can detect displacements in 3-D with good signal-to-noise ratio. However, it is costly and the scanning time is long compared to medical ultrasound. Ultrasound on the other hand, is portable, cheaper and faster but the displacement estimations have lower signal-to-noise ratio. Figure(1) shows an image from an ultrasound thyroid scanning. A 3-D printed neck holder is placed on the actuator and displacements are imaged using a linear ultrasound probe.

Figure 2

Figure (2) shows an example from a phantom recording. The shaker was driven with a 400 Hz sinusoidal signal. The phantom consists of inclusions with different stiffness values embedded in a tissue mimicking material. Figure (2).a shows an inclusion that is softer than the background material located between 25 – 45 mm depth. Wave images can be generated as shown in Figure (2).b after displacement estimation. One can see that in the upper part of the image, i.e., above 25 mm, the waves have constant wavelength. This wavelength is dependent on the stiffness of the material. Waves in the inclusion have shorter wavelengths which tells us that the material here is softer and the wave propagates with a lower velocity. Figure(2).c shows the final velocity estimations. As can be seen from the image, the part of the image where the inclusion is located has lower wave propagation velocity than the background. Using this velocity information, we can estimate the stiffness of the material.

Figure 3

Figure (3) shows the result of a recording from a stiff inclusion in the phantom. A 3-D ultrasound probe used for the recording and the stiff inclusion could be reconstructed in 3-D. Our main goal is to apply this technique in cardiac imaging and estimate the stiffness of the myocardium in 3-D using ultrasound imaging. Estimation of myocardial stiffness may potentially help in understanding and diagnosing of relaxation and contraction abnormalities in the heart.

By Marcus Wild

Marcus Wild

Within the maritime industry, ships use SONAR (Sound Navigation and Ranging) systems in order to make maps of the seabed. The problem with SONARs, is that they can have a tendency to overheat. Indeed, SONARs are composed of materials which exhibit energy loss, i.e., a percentage of the energy input into these materials is converted to heat. High temperatures can cause the SONARs to perform inefficiently and over time will damage the materials. Our aim is to predict the temperature rise in the materials used in SONARs. This will help us to design more efficient SONARs which in turn will allow us to push the performance of these systems.

High temperatures can cause the SONARs to perform inefficiently and over time will damage the materials.

SONARs emit sound waves, which then bounce off the seabed and return to the ship, in the same way that a bat or a dolphin uses sound to find its prey. SONARs are made of many components, but one of the most important components is the piezoelectric (piezo means squeeze) material. When pressure is applied to this type of material, an electric current is produced and reversely, when an electric current is applied to this material, it generates a sound wave (like a microphone and an amplifier). This behaviour is very useful and this material is the driving mechanism behind the sound wave generated by SONAR. However, some of this energy is also converted into heat, and too much heat will damage the SONAR.

In order to predict the heating in the piezoelectric material, the material properties such as the stiffness, electric characteristics (specifically, electric permittivity) and the electromechanical coupling factor (how easily the material converts mechanical energy like sound to electrical energy and vice versa) need to be characterised accurately.
Once these material characteristics have been determined, the power dissipation as a function of space and frequency can be calculated. The calculated power dissipation density, which can be thought as the amount of input energy converted to heat, for one particular frequency is shown in Fig. 1.

Figure 1: Power dissipation density as a function of position along the length of the rod.

Therefore, based on the calculated material parameters and the theory surrounding piezoelectric materials, the power dissipation at the centre of the rod should be more significant than at the edges of the rod. Using the curve in Fig. 1, we can now attempt to predict the temperature rise in a piezoelectric rod. An experimental setup was designed, shown in Fig. 2, in order to achieve this. The piezoelectric rod is placed in an insulating box in order to have a more stable convection rate (cooling due to air flow). A voltage is applied to the piezoelectric rod at a particular frequency through the copper wire in order to cause self-heating. Three thermocouples are attached at different positions along the side of the rod to measure the distribution of the temperature as the piezoelectric heats up.  This setup is duplicated in a Finite-Element-Method software called COMSOL in order to model the temperature rise at the thermocouple positions.

Fig. 3 shows the comparison between the modelled and measured temperature rise at the thermocouple positions. Although there are discrepancies between the measured and modelled temperature curves, the overall trend between the thermocouples does seem to have been captured well. This would suggest that the power dissipation density curve does indeed resemble the predicted curve shown in Fig. 1. Although this work does require a more accurate experimental setup and thermal material parameters in order to add confidence to this conclusion, we are confident that this a good first step to predicting the temperature in piezoelectric materials.

Making good digital ultrasound probes is extremely challenging because of the constrained amount of power that can be used. If significantly more than a couple of watts are consumed, the probe gets too hot to be allowed to touch your skin. When thousands of parallel ultrasound transmitters and receivers are preferred in a 3D ultrasound probe, it becomes clear that this power budget is tough.

In an ultrasound system, the handheld probe is the part that transmits ultrasound pulses into your body and captures the echoes. The probe tip has a number of transducer elements, which are tiny devices that act as ultrasound “speakers/microphones”. Normally, these elements are connected to the ultrasound machine, which converts all the echoes to digital signals used to make the image.

An alternative approach is to use a digital ultrasound probe, which converts echoes to digital signals in the probe itself. The probe can then send a digital data stream to the ultrasound machine, perhaps wirelessly. This will be a great advantage in 3D ultrasound, where the probe contains thousands of elements. It also enables the design of probes that are complete small ultrasound machines on their own, sending images to phones or tablets.

One of the most important parts of the ultrasound receivers are the analog-to-digital converters (ADCs), which convert ultrasound echoes into digital signals. In CIUS, we work on designing receiver ADCs that can be used in digital probes. These can only use a few hundred microwatts each to not exceed the power budget of the probe (one microwatt is 0.000001 watt). That is a tiny amount of power when we want to digitize the ultrasound echoes more than ten million times per second.

To make these low-power ADCs, we first and foremost need to make our own chips, using modern integrated circuit technology. Inside these we can have our custom designed ADCs, which we make as small and simple as possible, because smaller electronics use much less power. However, when electronic circuits are sized down, their accuracy eventually start to suffer, and we get too much noise and distortion in the ultrasound signals. The key to solving this is to figure out clever ways to reduce noise and distortion without making the circuits much larger.

A custom made analog-to-digital converter. Its physical dimensions are 110×160 micrometers.

We will have a closer look at distortion, which is unwanted, or false signals that appear in the digital output because of imperfections in the circuit components. For instance, the ADCs we make contain a group of capacitors, which are components that store electric charge, exactly like water tanks store water. In this group, the next capacitor always has to be exactly twice as large as the previous one, such that they get the sizes 1, 2, 4, 8, 16, up to 512. In practice, this does not happen, because of inevitable inaccuracy in the production of integrated circuits. The result is an ADC that generates distortion and works poorly.

One solution to this problem is to make the capacitors very large, because they are then easier to produce. However, this come at the cost of much higher power consumption. A better solution is to use calibration, and digitally correct for the errors of the capacitor bank. We are working on two different calibration approaches. One is to measure the errors of the capacitors directly, and use the error measurements to correct the ADC output. For this method, we have made an ADC where a part of it can be configured to measure the capacitors when you power up the equipment. Afterwards, the ADC starts normal operation and use the measurements to correct its output.

In the other method, we have developed an algorithm that can estimate the capacitor errors just by looking at the distortion present at the ADC output. Specifically, we send a test signal (a sinusoid) into the ADC, and look at the digitized signal and distortion at the output. The algorithm then figures out what the capacitor errors inside the ADC have to be to make up exactly the distortion that we observe. Then we can use the errors, which are now estimated rather than measured, to correct the output signal.

Output spectrum (output frequencies) from an analog-to-digital converter before (red) and after (blue) the calibration algorithm is used. The large spike is the test signal, and the other spikes are the distortion. They are suppressed after the calibration is done.

Transthoracic echocardiography is the most widely used imaging technique in cardiology. The examinations are typically performed following a protocol where an ultrasound probe is placed on the chest of a subject in order to acquire different standard views of the heart.

These images are used as reference when assessing cardiac function, and it is essential that the morphophysiological representations are correct. Image quality varies substantially between patients and is also operator dependent, which increases interobserver variability and lowers feasibility of detailed quantitative measurements in the clinic. Further, a new group of users is adopting echocardiography through the introduction of hand-held devices, making ultrasound available in the field and outpatient clinics. Increased availability is great news, but it also triggers a range of new challenges. Training opportunities are limited, and expert knowledge is often captivated by workload. How could these increasingly relevant problems be solved?

At the Centre of Innovative Ultrasound Solutions (CIUS) we believe modern machine learning methods, such as deep learning, can provide necessary solutions. In recent years, deep learning has received incredible attention by providing impressive results on image, speech and text recognition tasks. As opposed to traditional machine learning approaches with hand-crafted features, these methods learn both the feature extraction and classification directly from the training data. This is particularly interesting for ultrasound, where hand-crafting generic features can be extremely difficult. For details about the method, applications and challenges, read Postdoc Erik Smistad’s interesting blog on the topic.

Work package 4, led by Professor Lasse Løvstakken, covers new image processing and analysis methods for extracting relevant data from ultrasound images, and enhanced data visualization tools for displaying clinical information in an intuitive way. Our machine learning team is growing rapidly, and currently we are using deep learning to make intelligent systems that could standardize and quality assure the acquisition process, and ultimately help the operator achieve the best possible images for a given patient. Systems that could confirm that the images are good enough for quantitative measures, and also perform them fast and automatically on-site by extracting the relevant information directly from the images.

Two of the most essential building blocks of our systems are the view classification and segmentation algorithms. For both tasks, we employ convolutional neural networks trained on a vast amount of annotated ultrasound data. The annotations are performed by experts, which in turn incorporates their knowledge in the models. Our view classification tells us what we are looking at, for example, if it is one of the standard views shown in the video above. This can induce perception logic, and if images are considered applicable for segmentation, this is performed (see video below). The segmentation partitions the image into relevant objects by labeling the pixels to a known class, for instance the endocardial border of the left ventricle as can be seen in the video. Together, view classification and segmentation constitute a natural pipeline, that can facilitate automatic measurements and acquisition guidance. This can prove to be helpful for less experienced users, in point-of-care situations, and also reduce the workload in the outpatient clinic.

Undoubtedly, we see that deep learning has the potential of improving clinical workflow and diagnostic efficacy in echocardiography. At CIUS we will continue to pursue these methods, deliver novel and thoroughly evaluated solutions, and foremost, knowledge that will make it an essential component of tomorrow’s health care.

Bloggers: Researcher Debbie Hill og PhD Candidate Leslie Euceda Wood
MR Cancer group, Department for ciruclation and medical imaging

It’s that time of year again! And this year, the Movember Foundation is challenging YOU to move every day this month to tackle physical inactivity. In keeping with the Movember spirit, the MR Cancer group (and friends!) donned their best training gear, and ventured out into the crisp breeze of a lovely November morning. Not only to raise a few smiles, but also to raise awareness for men’s health and prostate cancer.

According to the Movember Foundation a lack of physical activity is the fourth leading risk factor for global mortality, causing 3.2 million deaths worldwide per year. There is a push to combat physical inactivity by:

1. Getting people moving (check out the MOVE campaign)
2. Raising awareness on the dangers of physical inactivity & benefits of activity for both physical and mental health
3. Finding new ways to encourage physical activity
4. Investing in projects that increase understanding of what motivates men to move.

Here at NTNU, the CERG group are focussed on investigating the effects of exercise in medicine (check out their website for information & fitness calculators!), and recent studies have uncovered that exercising just 1.5 hours per week can reduce mortality rates by 20%.

However, it’s not necessarily that simple – despite an increase in the amount we exercise, we are actually getting fatter! This could be explained by the prolonged periods of inactivity, which is commonplace in many lines of work. Making small changes to the daily routine, like cycling to work, using the stairs rather than lifts, or encouraging a ‘walking meeting’ rather than sitting at a desk, could all make a significant difference to your general health.

At the MR cancer group we are dedicated to improving our understanding of prostate cancer and to optimise imaging strategies to determine the best course of treatment. Exercise has been shown to decrease blood serum levels of metabolic and sex hormones thought to stimulate prostate cancer (1-3).

In addition, the incidence of prostate cancer increases with age. A build up of reactive oxygen species (free radicals) can alter the conditions inside cells, and has been linked with an increase in cancer incidence. Exercise combats these changes by decreasing the amount of reactive oxygen species, and could therefore contribute to reducing cancer risk (4). In prostate cancer patients, physical activity has been associated with reduced mortality, be it prostate-specific or not. While more vigorous physical training, like biking, jogging, or swimming around three hours per week may substantially improve prostate-specific survival (5,6).

We are currently carrying out a preliminary study into the effects of exercise on prostate cancer using a prostate cancer mouse model. (See our previous blog post: How moustaches, mice and magnets contribute to prostate cancer research.) We are interested in seeing whether mice that are allowed voluntary access to exercise wheels will develop cancer in a different way (or at all!) than their non-exercising counterparts. The mouse prostates are monitored periodically using magnetic resonance imaging (MRI), and we will validate our findings with metabolic and microscopic characterisations of the tissues, such as high resolution (HR) magic angle spinning (MAS) spectroscopy and immunohistochemistry.

Our research is ultimately directed towards improving clinical practice to help with our ongoing fight against cancer, and has been made possible by funding support from the Norwegian Cancer Society, Research Council of Norway, and the Liaison Committee between Central Norway RHA and NTNU.

So, join the Movember movement whether it’s by showing off that moustache, getting more active, or both. Raise funds and awareness for men’s health, and look after your own in the process!

References

1. Barnard RJ, Leung PS, Aronson WJ et al: A mechanism to explain how regular exercise might reduce the risk for clinical prostate cancer. Eur J Cancer Prev 2007; 16: 415.
2. Leung PS, Aronson WJ, Ngo TH et al: Exercise alters the IGF axis in vivo and increases p53 protein in prostate tumor cells in vitro. J Appl Physiol 2004; 96: 450.
3. Ornish D, Weidner G, Fair WR et al: Inten- sive lifestyle changes may affect the progres- sion of prostate cancer. J Urol 2005; 174: 1065.
4. Ji LL, Gomez-Cabrera MC and Vina J: Exercise and hormesis: activation of cellular antioxidant signaling pathway. Ann N Y Acad Sci 2006; 1067: 425.
5. Kenfeld SA, Stampfer MJ, Giovanucci E, Chan JM: Physical Activity and Survival After Prostate Cancer Diagnosis in the Health Professionals Follow-Up Study. J Clin Oncol 2011; 29:726-732
6. Bonn SE, Sjölander A, Lagerros YT, Wiklund F, Stattin P, Holmberg E, Grönberg H, Bälter K: Physical Activity and Survival among Men Diagnosed with Prostate Cancer. Cancer Epidemiol Biomarkers Prev 2014; 24(1): 57–64