Development of Relaxometry Methods and Hardware for Routine Determination of Volume Status: Dialysis Pilot Study
Lina Avancini Colucci1, Matthew Li1, Kristin Corapi2, Andrew Allegretti2, Rayhnuma Ahmed2, Herbert Y. Lin2, and Michael J. Cima3

1Health Sciences and Technology (HST), MIT, Cambridge, MA, United States, 2Division of Nephrology, Massachusetts General Hospital, Boston, MA, United States, 3Materials Science and Enginering, MIT, Cambridge, MA, United States

Synopsis

Incorrect assessment of clinical volume status leads to increased mortality and healthcare costs yet there are no accurate, non-invasive, and quantitative methods to assess this health metric. We evaluated the ability of a portable NMR sensor and relaxometry techniques to detect fluid changes in hemodialysis (HD) patients during the course of HD treatment. There was a significant difference between relaxation values of HD patients compared to healthy subjects.

Background

Clinical volume status is closely tied to mortality in patients with heart, liver, and kidney disease1. Existing methods to determine volume status are imprecise, invasive, and/or easily confounded2,3. Nuclear magnetic resonance (NMR) relaxometry – the measurement of relaxation variables – is a non-invasive technique that measures in vivo fluid compartments and can quantitatively measure fluid shifts4–7. Hemodialysis (HD) patients are well suited for volume status studies because they regularly have large amounts of fluid removed with dialysis. We aimed to evaluate the ability of portable NMR sensors to detect fluid changes in HD patients during the course of a single HD treatment. We also compared baseline NMR results between our HD subjects and a sample of healthy controls.

Methods

HD patients (25+ years old, BMI 18.5-40, no amputations or metal implants) undergoing routine HD at the Massachusetts General Hospital were eligible for participation. Patients had serial NMR measurements taken of their finger during dialysis treatment. Weight change, fluid removal volume, vital signs and dialysis machine settings were recorded. A previously enrolled cohort of healthy subjects who had NMR measurements taken with the same device while undergoing aerobic exercise was used as a comparison group.

NMR Sensor: A custom NMR sensor based off a circular Halbach array magnet design was built for index finger measurements (B0=0.55T, 770mm3 cylindrical voxel; Figure 1). It utilized a CPMG sequence to measure the T2 relaxation time of the entire fingertip with 8,000 echoes and TE = 300ms.

MRI: T2-mapping was performed on the fingers (healthy pilot: n=1, age=24) using a 3T whole-body Siemens MR scanner with a wrist coil. The T2 maps of 6 slices with TR=5.0s and 32 echo times (TE1st scan=8ms, TE2nd scan= 25.5ms) and a resolution of 1x1x5mm (128 x 64 matrix) were acquired in 5.5 min per scan. Regions of interest (ROIs) were drawn on the MRI images to calculate the T2 relaxation time contributions of specific tissues.

Analysis: Multi-exponential fittings on the T2 decays of both sensor (5-exponential fit) and MRI (2-exponential fit) data were performed using the curve fitting toolbox in Matlab.

Results

The 21 HD patients who participated had a mean age of 65 ± 13 years (Figure 2). The mean ultrafiltration volume per dialysis session was 2267.1 ± 988.2mL. The 20 healthy volunteers had a mean age of 25 ± 2 years. Finger relaxation times in HD patients at the start of dialysis were significantly higher than those of healthy controls (T2,E: 428.6 ± 65.3 vs 327.7 ± 22.7ms; p<0.00001; Figure 3). There was a decrease in the HD patients’ finger relaxation times at the end compared to the beginning of treatment (T2,E: 413.7 ± 56.9ms vs. 380.1 ± 59.4ms; p = 0.1). The four T2 relaxation time components measured by the MRI (n=1) corresponded to four of the five relaxation components measured by the finger NMR sensor for the same subject (T2,B= 21.7 ± 11.7 vs 15.4 ± 0.9ms; T2,C= 47.9 ± 7.6 vs 45.2 ± 1.9ms; T2,D= 150.8 ± 29.4 vs 122.5 ± 6.1ms; T2,E= 635.2 ± 68.7 vs 325.9 ± 7.3ms; Figure 4).

Discussion and Conclusion

Finger relaxation times of volume-overloaded HD patients were significantly higher (differenceE = 100.9ms, p<0.00001) than those of healthy controls. The increased relaxation times for HD patients compared to healthy subjects suggests a portable NMR device can distinguish between normal versus volume-overload status with a <10 minute finger measurement. Relaxation time may correlate with volume removal in real time, though additional enrollment is needed to confirm this finding. Modeling to understand how age, gender, BMI, medications, and fluid kinetics affect NMR readings is necessary. Age difference between the HD and healthy groups is a limitation of this study. Measurements with age-matched healthy subjects will be performed in the future.

MRI results validated the multi-exponential fitting of the NMR finger sensor data by yielding similar values and elucidating which tissues contribute to which relaxation time compartments. Our phantom studies on the MRI and NMR finger sensor show that T2 dependence on B0 is very limited for relaxation times <300ms (not shown). The highest relaxation time (~300ms) in the finger originates from marrow. The T2 decay for the bone marrow ROI in the MRI data was not fully relaxed and therefore had a bad fitting. This is likely why the MRI relaxation E value is significantly different from the finger sensor’s relaxation E value. Relaxation components B-D match between the MRI and finger sensor within their confidence interval.

Portable NMR sensors and relaxometry techniques are a promising approach to quantitatively assess a person’s volume status.

Acknowledgements

This work was supported by the MGH-MIT Strategic Partnership Grand Challenges Grant (Diagnostics Round), the Institute for Soldier Nanotechnologies (W911NF-13-D-0001), and the Koch Institute for Integrative Cancer Research (P30-CA14051, NCI). The authors thank the MGH Hemodialysis Unit as well as Dr. Martin Torriani and the MGH Metabolic Imaging Core for their assistance.

References

1. US Renal Data System, “Chapter 3: Morbidity and mortality in patients with CKD,” in USRDS 2013 Annual Report: Atlas of Chronic Kidney Disease and End-Stage Renal Disease in the United States, Bethesda, MD, 2013, pp. 63–72.

2. W. Frank Peacock and K. M. Soto, “Current technique of fluid status assessment.,” Congest. Hear. Fail., Supplement, pp. S45–51, Aug. 2010.

3. R. Carter, S. N. Cheuvront, M. A. Kolka, and M. N. Sawka, “Hydration Status Monitoring,” in Monitoring Metabolic Status: Predicting Decrements in Physiological and Cognitive Performance, Washington, D.C.: The National Academies Press, 2004, pp. 270–280.

4. M. Li, C. C. Vassiliou, L. A. Colucci, and M. J. Cima, “1H nuclear magnetic resonance (NMR) as a tool to measure dehydration in mice,” NMR Biomed., vol. 28, no. 8, pp. 1031–1039, 2015.

5. K. J. Hackney, S. B. Cook, T. J. Fairchild, and L. L. Ploutz-Snyder, “Skeletal muscle volume following dehydration induced by exercise in heat,” Extrem. Physiol. Med., vol. 1, no. 1, p. 3, Jan. 2012.

6. B. Kunnecke, P. Verry, A. Benardeau, and M. von Kienlin, “Quantitative Body Composition Analysis in Relaxometry,” vol. 12, no. 10, 2004.

7. M. Bruvold, J. G. Seland, H. Brurok, and P. Jynge, “Dynamic water changes in excised rat myocardium assessed by continuous distribution of T1 and T2,” Magn. Reson. Med., vol. 58, no. 3, pp. 442–7, Sep. 2007.

Figures

Figure 1. (a) The NMR finger sensor is a custom sensor with a circular Halbach array magnet design that measures the entire fingertip, Bo=0.55T, 770mm3 cylindrical voxel, (b) abstract authors demonstrate the clinical setup in the Massachusetts General Hospital Hemodialysis Unit.

Figure 2. Summary of the basic demographic information and comorbidities of the study’s HD patient population (n=21).

Figure 3. Four out of five T2 relaxation time components are plotted for healthy patients vs. HD patients at the beginning of dialysis. The raw finger sensor data is fit with five exponentials as shown in the equation in the figure. The volume-overloaded HD patients have a significantly higher relaxation time across all five components (shortest relaxation component A is not shown).

Figure 4. T2 relaxation times measured by MRI correspond to those measured by the NMR finger sensor for the same subject (n=1, pilot study). Three ROIs were drawn on the index finger corresponding to different tissues: marrow, fat, and connective tissue and nerves. The MRI T2 decays could be fit by up to 2-exponentials due to the relatively small number of points. The longest relaxation component (E) comes from bone marrow. The MRI T2 decay curve for bone marrow was not fully relaxed.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3876