Acoustic Noise Reduction by Avoiding Resonance Peaks

Abstract

Purpose The use switching gradient fields and radio frequency pulses in Magnetic Resonance Imaging affects safety and can reduce patient comfort. Three of those effects are Specific Absorption Rate (SAR), Peripheral Nerve Stimulation (PNS) and acoustic noise. The first two can be mitigated by simply slowing down the sequence by increasing the repetition time (TR). For acoustic noise it is often assumed that by reducing the gradient strength or slew rate, the noise will be lower. Overall this is true, however, it may occur that by reducing the gradient performance, the timing of the sequence will be affected such that its frequency content coincides with one or more of the mechanical or acoustical resonance frequencies of the scanner and the acoustic noise actually increases [1],[2]. Methods The goal of our method is to achieve acoustic noise reduction by changing the timing of the sequence such that the acoustic resonance peaks of the scanner are avoided. In TSE and EPI scans, the echo spacing is increased; in other scans the TR is increased. First we predict the sound pressure level (SPL) of a sequence by calculating the product of the gradient waveform spectrum and the frequency transfer function of the scanner [3]. By using a longer echo spacing or increased TR, the spectrum condenses towards lower frequencies. In a first order estimate, the number of peaks and their amplitude will remain the same; they only move to lower frequencies. With this assumption it is straightforward to predict the gradient spectrum for a range of different timings, for which now also the SPL can be calculated. Changing the timing is generally undesirable. To avoid too drastic changes the predicted SPL is multiplied with a weighting function, favoring small noise improvements at limited cost over bigger improvements at high cost. Experiments are performed on several types of scanners on field strengths 1.0, 1.5 and 3.0 T. Results A gradient echo scan with a TR of 2.17 ms has a gradient spectrum with a first peak around 460 Hz, see figure 1. By increasing the repetition time, predicting the gradient spectrum and calculating the SPL, the graph in figure 2 is obtained. The algorithm searches for a minimum value in the weighted SPL curve. At a TR of 2.44 ms, the reduction in SPL is predicted as 6.1 dB. It is interesting to note that for shorter as well as longer repetition times, the acoustic noise will become higher. Even lower predicted values of the SPL, around a TR of 2.7 ms and 2.8 ms are considered as less desirable due to the weighting factor. With the proposed increase in TR by just 12% applied, the actual SPL reduction is 7.1 dB. The estimated and actual gradient spectrums for the TR of 2.44 ms are shown in figure 3. Our algorithm has been applied to the standard database of protocols as used on a standard 3.0T scanner. In about one third of the protocols, the protocol parameters allowed to change the timing. A histogram of the reduction in SPL after avoiding the resonance frequency is shown in figure 4. The average decrease in SPL was 3.9 dB.

Full text

Acoustic noise reduction by avoiding resonance peaks
Jouke Smink, Geert-Jan Plattèl, Paul R. Harvey and Jan. P. Groen
Philips Healthcare, Best, the Netherlands.
Purpose
The use switching gradient fields and radio frequency pulses in Magnetic Resonance Imaging affects safety and can reduce patient
comfort. Three of those effects are Specific Absorption Rate (SAR), Peripheral Nerve Stimulation (PNS) and acoustic noise. The first
two can be mitigated by simply slowing down the sequence by increasing the repetition time (TR). For acoustic noise it is often
assumed that by reducing the gradient strength or slew rate, the noise will be lower. Overall this is true, however, it may occur that by
reducing the gradient performance, the timing of the sequence will be affected such that its frequency content coincides with one or
more of the mechanical or acoustical resonance frequencies of the scanner and the acoustic noise actually increases [1],[2].
Methods
The goal of our method is to achieve acoustic noise reduction by changing the timing of the sequence such that the acoustic resonance
peaks of the scanner are avoided. In TSE and EPI scans, the echo spacing is increased; in other scans the TR is increased. First we
predict the sound pressure level (SPL) of a sequence by calculating the product of the gradient waveform spectrum and the frequency
transfer function of the scanner [3]. By using a longer echo spacing or increased TR, the spectrum condenses towards lower
frequencies. In a first order estimate, the number of peaks and their amplitude will remain the same; they only move to lower
frequencies. With this assumption it is straightforward to predict the gradient spectrum for a range of different timings, for which now
also the SPL can be calculated. Changing the timing is generally undesirable. To avoid too drastic changes the predicted SPL is
multiplied with a weighting function, favoring small noise improvements at limited cost over bigger improvements at high cost.
Experiments are performed on several types of scanners on field strengths 1.0, 1.5 and 3.0 T.
Results
A gradient echo scan with a TR of 2.17 ms has a gradient spectrum with a first peak around 460 Hz, see figure 1. By increasing the
repetition time, predicting the gradient spectrum and calculating the SPL, the graph in figure 2 is obtained. The algorithm searches for
a minimum value in the weighted SPL curve. At a TR of 2.44 ms, the reduction in SPL is predicted as 6.1 dB. It is interesting to note
that for shorter as well as longer repetition times, the acoustic noise will become higher. Even lower predicted values of the SPL,
around a TR of 2.7 ms and 2.8 ms are considered as less desirable due to the weighting factor. With the proposed increase in TR by
just 12% applied, the actual SPL reduction is 7.1 dB. The estimated and actual gradient spectrums for the TR of 2.44 ms are shown in
figure 3.
Our algorithm has been applied to the standard database of protocols as used on a standard 3.0T scanner. In about one third of the
protocols, the protocol parameters allowed to change the timing. A histogram of the reduction in SPL after avoiding the resonance
frequency is shown in figure 4. The average decrease in SPL was 3.9 dB.
0
5
10
15
20
25
0 500 1000 1500 2000
Frequency (Hz)
Amplitude (mT/m)
freq. response function TR = 2.17 ms TR = 2.25 ms TR = 2.44 ms
-10
-5
0
5
10
15
2 2.1 2.2 2.3 2.4 2.5 2.6 2.7 2.8 2.9 3
repetition time (ms)
difference in SPL (dB)
SPL weighting factor weighted SPL
0
5
10
15
20
25
0 500 1000 1500 2000
Frequency (Hz)
Amplitude (mT/m)
freq. response function estimated spectrum real spectrum
Conclusion
This algorithm is flexible and automatic. It can be applied to any scan as long as the
protocol definition allows some freedom in the timing. Protocols with a single dominant
frequency in the excited spectrum are efficiently optimized with respect to acoustic
noise, without significant penalties. Other scans with a rich spectrum like GraSE or
multi-shot EPI contain a large number of peaks and are thus difficult to optimize.
0
5
10
15
20
25
30
35
40
12345678910111213141516
SPL reduction (dB)
# protocols
References
[1] L.S. Petropoulos et al., ISMRM fourteenth scientific meeting, Seattle, 2006: 2048
[2] S. Schmitter et al., ISMRM fourteenth scientific meeting, Seattle, 2006: 2814
[3] R.A. Hedeen, W.A. Edelstein, Magnetic Resonance in Medicine, 37 (1997) 7-10