Single-sided magnetic resonance-based sensor for point-of-care evaluation of muscle

Magnetic resonance imaging is a widespread clinical tool for the detection of soft tissue morphology and pathology. However, the clinical deployment of magnetic resonance imaging scanners is ultimately limited by size, cost, and space constraints. Here, we discuss the design and performance of a low-field single-sided magnetic resonance sensor intended for point-of-care evaluation of skeletal muscle in vivo. The 11 kg sensor has a penetration depth of >8 mm, which allows for an accurate analysis of muscle tissue and can avoid signal from more proximal layers, including subcutaneous adipose tissue. Low operational power and shielding requirements are achieved through the design of a permanent magnet array and surface transceiver coil. The sensor can acquire high signal-to-noise measurements in minutes, making it practical as a point-of-care tool for many quantitative diagnostic measurements, including T2 relaxometry. In this work, we present the in vitro and human in vivo performance of the device for muscle tissue evaluation.


REVIEWER COMMENTS
Reviewer #1 (Remarks to the Author): -What are the noteworthy results?
A low-field single-sided MR sensor was proposed to evaluate the skeletal muscle in vivo.
-Will the work be of significance to the field and related fields?How does it compare to the established literature?If the work is not original, please provide relevant references.
The work has a certain significance to the applicafion field of single sided NMR.The proposed sensor is similar to other researchers', but uses more discrefized magnets, which is somewhat innovafive in the detecfion applicafion of skeletal muscles.For the sensor itself, there was no special opfimizafion, only ordinary parameter scanning methods were used (Line 296 The magnet size, orientafion, locafion, spacing, etc were input parameters.These were swept to determine opfimal magnet array design).For the analyfical methods, a biexponenfial fit of the T2 decays was used.There are sfill some doubts about the analyfical method of using the double exponenfial fifting of T2 decay, and further explanafion is needed.
-Does the work support the conclusions and claims, or is addifional evidence needed?
The addifional evidence is needed to explain why the biexponenfial fit of the T2 decays is suitable for the detecfion of the skeletal muscle.The relafionship bewteen caculated T2 by biexponenfial fit of the CPMG decay and skeletal muscle evaluafion is needed to provide.There is no theorefical descripfion for this relafionship, which is vital for the applicability of the designed proble.
-Are there any flaws in the data analysis, interpretafion and conclusions?-Do these prohibit publicafion or require revision?
There are some errors or doubts that will be detailed in the following quesfions.These require revision.
-Is the methodology sound?Does the work meet the expected standards in your field?
The method sfill needs further explanafion.This work can be further improved to meet the expected standards in our field.
-Is there enough detail provided in the methods for the work to be reproduced?
In the design method of sensor, as the parameter scanning method used is very common, these details are sufficient.From the method of biexponenfial fit of the T2 decays, the method is also commonly used and the details are sufficient.The processing of images and data stafisfics can provide some more details.
major points: 1.In this manuscript, the authors propose a low-field single-sided MR sensor, aiming for skeletal muscle in vivo.Although it was reported that similar MRI based systems and methods have been used for skeletal muscle evaluafion, the low-field single-sided MR sensor designed in this manuscript presents a different way for skeletal muscle measurements.The manuscript is well wriften and the concept of sensor design is reasonable, but the method sfill needs further explanafion.Some major concerns are needed to be addressed before it is acceptable.

2.
In line 172, it states that " Data was collected from the phantoms at 8. 48, 8.43, 8.38, and 8.29 MHz, corresponding to depths of approximately 2-, 5-, 8-, and 10-mm from the surface of the sensor…".In line 113, it states that " The mapped homogeneous region following fabricafion has a field strength of 0.2 T and sits 8mm above the surface of the array with a natural descending gradient in the Z direcfion of 1 T/m…".These two paragraphs seem to contradict each other.Because the spin magnefic rafio of hydrogen protons is approximately 42.58 MHz/T, so the frequency changed 0.04258 Mhz/mm.The phantoms at 8. 48, 8.43, 8.38, and 8.29 MHz seem not corresponding to depths of approximately 2-, 5-, 8-, and 10-mm.Please check it.
In addifion, the field strength of 0.2 T is corresponding to 8.516 MHz, then there will be an undeniable millimeter level error.
3. In Line 179 to 184, it states that "The signal from the layered phantom acquired at and below 8.38MHz is stafisfically the same as the muscle fissue phantom.This demonstrates we are only capturing the signal from the muscle porfion of the layered phantom.The signal reflects amplitudes between the muscle and adipose phantoms at 8.43 MHz.The signal at this frequency contains contribufions from both phantom types near the layer juncfion.The signal acquired at 8.48 MHz, however, stafisfically reflects the adipose phantom; verifying that we are fully below the phantom layer juncfion, we can achieve an accurate signal from muscle phantom above a 6-mm-thick layer of adipose phantom at 8.38 MHz using a biexponenfial fit of the decay…" In fact, for unilateral magnets, due to the presence of natural gradients, it is relafively easy to choose a specific RF excitafion frequency to determine which height the sample is excited to above the surface of the magnet.This has nothing to do with the biexponenfial fit of the decay method.
In addifion, from the spectral analysis of Transverse relaxafion fime(T2), the composifion of skeletal muscle or adipose layer containing hydrogen atom groups is also relafively complex, and they do not correspond to a specific Transverse relaxafion fime(T2).So the method of biexponenfial fit of the decay is also an approximate method.The T2 spectra by ILT from the acquired CPMG echo curves could be used to as a further proof.
4. How to define the relafionship bewteen caculated T2 by biexponenfial fit of the CPMG decay and skeletal muscle evaluafion?There is no theorefical descripfion for this relafionship, which is vital for the applicability of the designed proble.Taking MRI for example, it directly gives a visual image showing the skeletal muscle status.
5. In Line 69, it states that "Previous clinical studies with SSMR sensors were limited by the penetrafion depth (< 6 mm) and signal sensifivity…".In fact, there are many SSMR sensors with penetrafion depth more than 6 mm and relafively high signal sensifivity.Previous clinical studies with SSMR sensors maybe limited by other reasons such as engineering issues.
6.In Line 154, it states that "Slice selecfion characterizafion was performed by adjusfing the pulsed (B1) frequency from 8.32 -8.42 MHz in 0.1 MHz increments… " The 0.1 MHz increments should be 0.01 MHz increments from Figure 3 B.

For CPMG pluse, how about is the excitafion bandwidth of the RF excitafion pulse in kHz?
Reviewer #2 (Remarks to the Author): This manuscript described a single-sided MR sensor for measuring calf muscle, with a main focus on the tailored magnet design.While each component of the system did not show any significant breakthroughs in terms of novelty, it did enable a parficular yet limited in-vivo applicafion.The verificafion experiments were reasonably designed and the conclusions were generally sound.

Specific comments:
1. FEA design.If the goal was to increase SNR, via trade-offs between sweet spot volume/B0, and depth/weight etc., how exactly were the decisions made?Was there a quanfifiable metric (e.g.SNR itself as a funcfion of the outcome measures) that supported the statement and were any outcome measures play more important roles?These are important informafion currently missing.
2. The magnet design does not seem to deviate a lot from convenfional single-sided Halbach arrays.What were the key differences that made the gradient smaller in the ROI when compared with "other single-sided sensors"(Line 217-218)?This seems to be one highlight of the work and should be elaborated in greater details.
3. Line 389 -"A 16x16x 32 mm area in the center of the array containing the array sweet spot...-how large was the sweet spot volume relafive to the size of the area?Was the RF bandwidth (both hardware limits and pulse length), hence the thickness, matched to the determined volume (presumably +/-0.5% of the center frequency)?
4. The targeted signal source is jointly selected by B0 and B1 fields.It would be great to show the 3D simulated excited slice (determined by center freq/BW) laying on top of a secfion of the calf (may be taken from a simplified model or even actual MR images), and see how much signal comes from the muscle but not the subcutaneous fat/fissues.5. How was the size of the RF coil determined?It plays no lesser role in determining the ROI, parficularly undesired signal from subcutaneous regions.It is important to show the 3D slice profile in 4 as it would inevitably cross the skin, just a mafter of how far away from the coil sensifive areas and/or the flip angles.
6. How far was the coil placed above the magnet?Was ringdown a big issue?7. Figure 3. Was the signal magnitude measured by the peak value?I think a more precise way would be calculate the area below the peak, as a flat but lower-amplitude spectrum could lead to more signal than a narrow but higher peak.
8. Figure 4. Why were the error bars much longer for 8.29/8.43MHz?Does that have any implicafions?Another thing is what was the diameter of the phantom and did that properly mimic the human calf geometry?Would it be possible to construct the phantom as two concentric rings (outer ring 6mm thickness) with the longitudinal direcfion covering the crossings of the slice profile and the phantom?9. Figure 5B.Intuifively, the Human T2,1 should be closer to Rat Muscle T2,1 than Rat Fat T2,1.But that was not the case.Does that make sense?10.How was the SNR affect the fifting?Line 429 -since the authors had all the data, a plot of fifting accuracy as a funcfion of averaging numbers can be generated to provide some insights.
11.The diffusion effect was menfioned in the supplemental material but not in the main body.Did that provide poor differenfiafions compared with T2? Was the diffusion effect absorbed into T2 decay in the current simplified format, and if so, how large of the signal decay did it provide compared with T2?
The authors would like to express appreciation for the thorough review of our manuscript, "Single-sided magnetic resonance-based sensor for point-of-care evaluation of muscle".The feedback has been invaluable in supplementing and refining the content and quality of the manuscript.The authors have carefully considered all comments and are grateful for the expertise and time dedicated to this process.In the following response, we address each of the points, showing how they have contributed to the improvement of the article.

Reviewer #1:
-What are the noteworthy results?A low-field single-sided MR sensor was proposed to evaluate the skeletal muscle in vivo.
-Will the work be of significance to the field and related fields?How does it compare to the established literature?If the work is not original, please provide relevant references.The work has a certain significance to the application field of single sided NMR.The proposed sensor is similar to other researchers', but uses more discretized magnets, which is somewhat innovative in the detection application of skeletal muscles.For the sensor itself, there was no special optimization, only ordinary parameter scanning methods were used (Line 296 The magnet size, orientation, location, spacing, etc were input parameters.These were swept to determine optimal magnet array design).For the analytical methods, a biexponential fit of the T2 decays was used.There are still some doubts about the analytical method of using the double exponential fitting of T2 decay, and further explanation is needed.
Response: Further justification for the use of biexponential fitting has been included in the Background section.Additional context and references to past and recent literature context and references has been included to provide support for the use of multi-compartment fitting models for quantitative T2 data.

Part of text adding into the Background section:
There is extensive evidence of skeletal muscle quantitative T2 relaxation being better represented by a bi-exponential model as compared to a mono-exponential model.There is open discussion as to the specific physiological context of the two exponential decays, with support for the two relaxation dynamics originating from either water and lipids, or from differing water compartments within a tissue [12][13][14] .Multi-compartment analysis has demonstrated higher specificity in differentiating muscle tissues with inflammatory pathologies, dystrophic pathologies, differing fat fractions, and differing water content, regardless of the origin of bi-exponential signal 12,13 .We support the conclusions of previous work that the two relaxations represent intercellular (shorter component) and intracellular (longer component) water compartments within a singular tissue.While both muscle and subcutaneous tissues exhibit biexponential T2 decays, the shorter 'intracellular' time constant is largely conserved between tissue types, while the longer of the time constants differs between tissues and can be distinguished from one another 14 .

-Does the work support the conclusions and claims, or is additional evidence needed?
The additional evidence is needed to explain why the biexponential fit of the T2 decays is suitable for the detection of the skeletal muscle.The relationship between calculated T2 by biexponential fit of the CPMG decay and skeletal muscle evaluation is needed to provide.There is no theoretical description for this relationship, which is vital for the applicability of the designed probe.

Response: Same as above. Further information informing the decision to use a bi-exponential fit has been included
in the manuscript.
-Are there any flaws in the data analysis, interpretation and conclusions?-Do these prohibit publication or require revision?There are some errors or doubts that will be detailed in the following questions.These require revision.
-Is the methodology sound?Does the work meet the expected standards in your field?The method still needs further explanation.This work can be further improved to meet the expected standards in our field.
Response: Several parts have been supplemented in the Methods section, including an additional subsection on FEA design and outcomes, and additional information pertaining to fitting and statistical analysis.
-Is there enough detail provided in the methods for the work to be reproduced?In the design method of sensor, as the parameter scanning method used is very common, these details are sufficient.From the method of biexponential fit of the T2 decays, the method is also commonly used and the details are sufficient.The processing of images and data statistics can provide some more details.
Response: More details have been added in the Background and Methods sections on the data processing and analysis.A section on statistical analysis has been added.major points: 1.In this manuscript, the authors propose a low-field single-sided MR sensor, aiming for skeletal muscle in vivo.Although it was reported that similar MRI based systems and methods have been used for skeletal muscle evaluation, the low-field single-sided MR sensor designed in this manuscript presents a different way for skeletal muscle measurements.The manuscript is well written and the concept of sensor design is reasonable, but the method still needs further explanation.Some major concerns are needed to be addressed before it is acceptable.
2. In line 172, it states that " Data was collected from the phantoms at 8. 48, 8.43, 8.38, and 8.29 MHz, In line 113, it  states that " The mapped homogeneous region following fabrication has a field strength of 0.2 T and sits 8mm above the surface of the array with a natural descending gradient in the Z direction of 1 T/m…".These corresponding to depths of approximately 2-, 5-, 8-, and 10-mm from the surface of the sensor…".two paragraphs seem to contradict each other.Because the spin magnetic ratio of hydrogen protons is approximately 42.58 MHz/T, so the frequency changed 0.04258 Mhz/mm.The phantoms at 8. 48, 8.43, 8.38, and 8.29 MHz seem not corresponding to depths of approximately 2-, 5-, 8-, and 10-mm.Please check it.In addition, the field strength of 0.2 T is corresponding to 8.516 MHz, then there will be an undeniable millimeter level error.
Response: This is an excellent point.The wording pertaining to the field strength has been clarified in the text.The maximum field strength of the array is 0.2T at the surface of the sensor.The homogeneous region has a slightly lower field strength (0.196T) that sits 8mm above the surface of the sensor.With this context clarified, the frequencies use for phantom experimentation correspond to the given depths.
3. In Line 179 to 184, it states that "The signal from the layered phantom acquired at and below 8.38MHz is statistically the same as the muscle tissue phantom.This demonstrates we are only capturing the signal from the muscle portion of the layered phantom.The signal reflects amplitudes between the muscle and adipose phantoms at 8.43 MHz.The signal at this frequency contains contributions from both phantom types near the layer junction.The signal acquired at 8.48 MHz, however, statistically reflects the adipose phantom; verifying that we are fully below the phantom layer junction, we can achieve an accurate signal from muscle phantom above a 6-mm-thick layer of adipose phantom at 8.38 MHz using a biexponential fit of the decay…" In fact, for unilateral magnets, due to the presence of natural gradients, it is relatively easy to choose a specific RF excitation frequency to determine which height the sample is excited to above the surface of the magnet.This has nothing to do with the biexponential fit of the decay method.In addition, from the spectral analysis of Transverse relaxation time(T2), the composition of skeletal muscle or adipose layer containing hydrogen atom groups is also relatively complex, and they do not correspond to a specific Transverse relaxation time(T2).So the method of biexponential fit of the decay is also an approximate method.The T2 spectra by ILT from the acquired CPMG echo curves could be used to as a further proof.
Response: We agree with the reviewer that the natural gradient of single sided magnets makes the selection of RF frequency for slice selection relatively easy.The use of bi-exponential fitting models does not contribute to the slice selection or determination of which tissue is being detected.Since skeletal muscle tissue is better modeled (detailed further in the Background in response to other comments) and since pathology detection in muscle is more sensitive with a bi-exponential fit, we use this fitting methods as our analysis for all data presented in this manuscript.
We agree that and ILT analysis of the curves could be used as further proofcurrently our SNR is low for the use of ILT.In previous literature where ILT was implemented,150 was the minimum SNR and even at an SNR of 500 it was 80% accurate detection of peaks (Ioannidis et al., 2020, Berman et al, 2013).
Berman, P., Levi, O., Parmet, Y., Saunders, M. and Wiesman, Z. (2013), Laplace inversion of low-resolution NMR relaxometry data using sparse representation methods.Concepts Magn.Reson., 42: 72-88.4. How to define the relationship between calculated T2 by biexponential fit of the CPMG decay and skeletal muscle evaluation?There is no theoretical description for this relationship, which is vital for the applicability of the designed probe Taking MRI for example, it directly gives a visual image showing the skeletal muscle status.
Response: Further details pertaining to the justification for applying multi-compartment models (bi-exponential fitting) to evaluation of skeletal muscle as well as further information on the applicability of the instrument has been included in the Background section.There is additionally now a caveat included of specific analysis techniques varying with the clinical application in question.

Additional text in the Background section includes:
Techniques including T2 relaxometry and T2-weighted diffusion can be performed on single-sided sensors to provide clinicallyactionable information [15][16][17][18] .Uses include assessment of liver disease, inflammation, tumor characteristics, iron overload, and cartilage diseases, among others 7,11,19 .Within skeletal muscle tissue specifically, relaxometry can provide insight into fluid status, progressive disease musculoskeletal disease monitoring (sarcopenia, muscular dystrophies, etc.), vascular kinetics and oxygenation tracking, among other applications 10,13,20 ….No one clinical application is evaluated for the use of this tool.Outcome metrics, acquisition parameters, and analysis techniques for specifical applications will vary based on clinical application.
5. In Line 69, it states that "Previous clinical studies with SSMR sensors were limited by the penetration depth (< 6 mm) and signal sensitivity…".In fact, there are many SSMR sensors with penetration depth more than 6 mm and relatively high signal sensitivity.Previous clinical studies with SSMR sensors maybe limited by other reasons such as engineering issues.
Response: This is an excellent point, the wording has been clarified and references included to point to further limitations, including engineering challenges, of SSMR sensors and their clinical applications.
6.In Line 154, it states that "Slice selection characterization was performed by adjusting the pulsed (B1) frequency from 8.32 -8.42 MHz in 0.1 MHz increments… " The 0.1 MHz increments should be 0.01 MHz increments from Figure 3 B.
Response: Thank you for the correction.This mistake has been adjusted in the text.
7. For CPMG pulse, how about is the excitation bandwidth of the RF excitation pulse in kHz?
Response: The methods sections describing the pulse used now include the pulse bandwidth (43kHz).
Reviewer #2: This manuscript described a single-sided MR sensor for measuring calf muscle, with a main focus on the tailored magnet design.While each component of the system did not show any significant breakthroughs in terms of novelty, it did enable a particular yet limited in-vivo application.The verification experiments were reasonably designed and the conclusions were generally sound.

Specific comments:
1. FEA design.If the goal was to increase SNR, via trade-offs between sweet spot volume/B0, and depth/weight etc., how exactly were the decisions made?Was there a quantifiable metric (e.g.SNR itself as a function of the outcome measures) that supported the statement and were any outcome measures play more important roles?These are important information currently missing.
Response: This is an excellent pointan additional section has been added to the Methods section further detailing design process, outcome metrics, and key decisions influencing the final array design

Additional text incudes:
A two-step approach was used to determine array configuration.Initially 2-D models were used to approximate the net magnetic profile of several basic magnet array geometries including unilateral Halbach, semi-cylindrical Halbach, U shaped, and L shaped For these geometries, an optimization score was calculated based on the B0 strength and gradient: O  =  0 7 4  *  0 35 .The Optimization score and depth of homogeneous region was used to select a semi-cylindrical design as the basis for further design optimization.A 3-D model of the geometry was then used to evaluate several parameters including magnet size, shape,