7 Tesla MRI artifacts and cardiac imaging

7 Tesla MRI

I discovered the concept of artifacts when I started learning MRI, back in another century. For most of us, an artifact is an antique object made by a human being from a forgotten culture. For the signal processing geeks among us, it has a completely different meaning: an MRI artifact, unlike noise, is usually from a deterministic phenomenon which creates an undesirable shift in the signal we are trying to measure. Eventually, you can hear a “what’s going on with the image? What is this artifact again?”.

Artifacts seem to multiply as the magnetic field strength of MRI increases. This is particularly true at 7 Tesla, the Formula 1 of MRIs. There is a bit of a myth that some technologies developed in Formula 1 racing can be transported to the everyday cars we are driving. I checked and could not pinpoint anything major¹, except if I drove a high Ferrari or a Porsche (I don't), but I digress.

Using 7T MRI, some clinical findings can be visualized, and traced back to lower, clinically available field strengths. How does it work? Sometimes, you do not know what you can see, until you actually see it. 7T MRI are hardly economically viable for healthcare, but can be justified through the clinical results they bring that can be translated to lower MRI field strengths. This, theoretically, also works with artifacts. Some MRI artifacts are a showstopper for 7T MRI. Developing technologies to correct them is a must. This technology could be translated to lower field strengths. A good example is Skope, who first brought precision in 7T MRI and later translated their work in clinical-ready MRI.

For what concerns Epsidy, cardiac MRI suffers from many potential sources of artifacts. For instance, motion artifacts, from breathing or cardiac beating, become more intense with cardiac MRI at 7T. Magnetic susceptibility is the source of another particularly annoying artifact. It presents as distortions or local signal change due to local magnetic field inhomogeneities, and quite obviously is enhanced at 7T, especially in the air-tissue interfaces found in the lungs. These limitations of 7T, plus some others, make imaging anatomies outside of the brain or joints, and particularly cardiac, particularly challenging.

One key feature of cardiac MRI is to detect ventricular contraction to synchronize the acquisition of data with heartbeats. Detecting QRS complexes has been the gold standard, as they precede ventricular contractions by a few milliseconds. QRS triggering presents a timing advantage over other methods. Oftentimes, when QRS detection fails, MRI technologists revert to pulse oximetry or real-time cardiac imaging, with quite a severe reduction in image quality, which defies the purpose of 7T cardiac MRI.

At 7T, detecting QRS complexes is particularly challenging because artifacts in the recorded electrical signal of the heart are greatly exacerbated. In reality, detecting QRS complexes at lower field strengths is also problematic. Performing 3T cardiac MRI presents challenges in a large percentage of patients. This will be the topic of a future blog post.

ECG artifacts when the patient is moved inside the MRI (1.5T)

What has the research community come up with to overcome this issue?

Various solutions have been studied and many published. Over a decade ago, an acoustic triggering was invented, and, despite the overwhelming noise of the 7T environment, proved to be efficient. Another approach used Holter data with multiple ECG leads (V1 to V6) to detect QRS complexes; the data was analyzed in multiple signal dimensions ², coined as "3DQRS". There is still work to be done in order to meet the challenges of using Holter devices, with their long cables, that are typically unsafe at 7T. Using the standard manufacturer 3 lead ECG acquisition hardware with a learning phase outside of the MRI, good results were achieved ^4,5. Another approach compared using a video camera aimed at the patient's forehead with pulse oximetry ⁶. Pulse oximetry can be used in some MRI pulse sequences, but not all, and sometimes compromises image quality. Another pulse-based approach used Doppler ultrasound (DUS) at the level of patients’ (or in this case volunteers’) hearts⁷. This approach presents the advantage of recording the heart wall motion; DUS delay to ventricular contraction is reduced compared to the pulse wave at the level of the forehead or, worst, at the tip of a finger. This approach was later applied to cranial MRI angiography⁸ and is also efficient for fetal heart triggering, which is performed at a lower, clinically acceptable, magnetic field strength. A creative approach used scatter signal from 8 transmit coil elements to create a sort of short wavelength radar recording ventricular contractions⁹. The principle of this approach resembles a promising new approach called pilot tone¹⁰, which is commercially available for respiration triggering and under development for cardiac triggering.

Considering the various approaches to the problem of 7T cardiac MRI triggering, and strikingly, a recent publication showing positive results on 84 consecutive examinations (72 volunteers)¹¹, one could think the problem is closed. However, there is still significant room to improve how the electrical signal of the heart is measured during MRI. Patients, technologists and radiologists demand a better experience with cardiac MRI. It is not only a matter of technology, but solving the problem of recording ECG without artifacts during MRI will contribute to this goal.

This will not only benefit cardiac MRI, but also other ultra-high resolution imaging of the nervous system (brain or spinal cord) currently limited by vascular pulsatility. There as well, precise synchronization with the cardiac cycle is required. Finally, MRI angiography is increasingly done without external contrast agents (for example, QISS¹²), and this as well would benefit from precise cardiac triggering.

Sources and references

1. https://www.racv.com.au/royalauto/transport/cars/f1-innovations-in-road-cars.html

2. J Cardiovasc Magn Reson. 2010 Nov 16;12(1):67. doi: 10.1186/1532-429X-12-67.

3. Magn Reson Med. 2014 Apr;71(4):1374-80.  doi: 10.1002/mrm.25078. Epub 2014 Jan 22

4. Tomography. 2016 Sep;2(3):167-174. doi: 10.18383/j.tom.2016.00193.

5. Tomography. 2021 Aug 4;7(3):323-332. doi: 10.3390/tomography7030029.

6. Biomed Eng Online. 2016 Nov 24;15(1):126.  doi: 10.1186/s12938-016-0245-3.

7. Magn Reson Med. 2018 Jul;80(1):239-247.  doi: 10.1002/mrm.27032. Epub 2017 Dec 1.

8. BMC Med Imaging. 2020 Dec 9;20(1):128.  doi: 10.1186/s12880-020-00523-x.

9. Magn Reson Med. 2018 Aug;80(2):633-640. doi: 10.1002/mrm.27038. Epub 2017 Dec 11.

10. Magn Reson Med. 2021 May;85(5):2403-2416. doi: 10.1002/mrm.28580. Epub 2020 Nov 23.

11. PLoS One. 2021 Jul 23;16(7):e0252797. doi: 10.1371/journal.pone.0252797. eCollection 2021.

12. PLoS One. 2014 Jan 16;9(1):e86274. doi: 10.1371/journal.pone.0086274. eCollection 2014.