Can Quantitative MRI Be Used to Differentiate Physiologic Changes Behind Muscle Weakness in Type 2 Diabetes Mellitus?

See also the article by Stouge et al in this issue.

Darryl Sneag, MD, is a fellowship-trained musculoskeletal MRI radiologist, director of peripheral nerve MRI at Hospital Special Surgery, and associate professor of radiology at the Weill Medical College of Cornell University. His clinical efforts focus on diagnostic imaging of neuromuscular conditions to help guide nonsurgical and surgical treatment as part of the Hospital for Special Surgery Center for Brachial Plexus and Traumatic Nerve Injury. Dr Sneag’s research efforts, in close collaboration with MR physicists, include the development and evaluation of vascular suppression and quantitative techniques for MR neurography.

Ek Tsoon Tan, PhD, is an associate scientist in the MRI Lab at the Hospital for Special Surgery and assistant professor of biomedical imaging in orthopedic surgery at the Weill Medical College of Cornell University. His research interests include quantitative MRI, including diffusion, for musculoskeletal applications. Dr Tan earned his doctoral degree in biomedical sciences from the Mayo Clinic, and his bachelor and master of engineering degrees from the National University of Singapore. He is also an adjunct assistant professor of medical physics at the Mayo Clinic College of Medicine and Science.

Peripheral neuropathies may cause muscle weakness from neuropraxia, a temporary loss of function, or from motor axon loss with denervation. Electrodiagnostic testing (comprising nerve conduction and electromyography studies) serves as an adjunct to the physical examination but is very much operator dependent (1) and semi-invasive. Dedicated noninvasive high-spatial-resolution MRI of the peripheral nerves or MR neurography supplements electrodiagnostic testing. However, standard approaches rely on subjective assessment of T2-weighted fat-suppressed images. Quantitative nerve imaging using diffusion techniques has been explored (2) but suffers from distortion, particularly off-isocenter distortion from the main magnetic field (B₀), and the spatial resolution is not high enough to visualize the small size of peripheral nerves (3). Muscle evaluation is sometimes a less regarded part of the MR neurography examination but is critical in helping localize and evaluate the severity of injury, as many neuropathies involve motor abnormalities. Muscle MRI, or “MR myography,” is complementary to MR neurography and is more conducive to quantitative assessment given the larger cross-sectional size of muscles compared with nerves.

In this issue of Radiology, Stouge and colleagues (4) use a panel of quantitative muscle MRI techniques in patients with type 2 diabetes mellitus. Diabetic polyneuropathy (DPN), characterized by symmetric sensory disturbances, is prevalent in approximately 50% of patients with chronic diabetes mellitus. Motor weakness is typically recognized only in severe cases of DPN (5) and less so in patients with type 2 diabetes mellitus without DPN. The authors hypothesized that patients with type 2 diabetes mellitus with DPN (+DPN) would exhibit microstructural changes proportional to the degree of neuropathy and muscle dysfunction, whereas patients with type 2 diabetes mellitus without DPN (-DPN) would display microstructural changes related to lower intrinsic muscle strength.

The authors performed a quantitative imaging protocol comprising T2-mapping, diffusion-tensor imaging, and fat fraction mapping sequences to compare groups of DPN-positive, DPN-negative, and healthy control participants of equal sample sizes (n = 20). The authors used isokinetic dynamometry to assess muscle strength at the knee and ankle. Results showed significantly higher fat fraction in the DPN-positive group (range, 13%–20%) compared with the DPN-negative (range, 8%–10%; P < .001) and healthy control (range, 6%–8%; P < .001) groups, suggestive of progressive fatty infiltration with disease. The authors also found prolonged T2 in the DPN-positive group (33 msec) relative to the DPN-negative (32 msec, P < .001 to P = .002) and control (31 msec, P = .03) groups, which they attributed to a higher percentage of muscle water composition. While the authors found no significant groupwise differences with diffusion-tensor imaging, fractional anisotropy (a measure of muscle fiber organization) along with T2 and fat fraction had positive correlation with the degree of neuropathy and impaired muscle strength in the DPN-positive group. Two important findings have immediate clinical relevance: (a) Similar to previous studies, participants in the DPN-negative group had preserved intrinsic muscle strength, suggesting that muscle strength relates to better glycemic control and disease duration. (b) Participants in the DPN-positive group had lower muscle strength explained predominately by muscle size, suggesting that muscle weakness is related to muscle atrophy rather than tissue quality, supporting the hypothesis that muscle excitability remains unaffected (6) and these individuals may recover motor function.

These results provide some basis for quantitative MRI to objectively determine the effects of various mechanisms behind muscle dysfunction. Fat fraction can depict the extent of fatty infiltration, T2 can depict the extent of extracellular edema, and diffusion-tensor imaging can depict changes in muscle microstructure related to atrophy. The authors addressed potential biases of these quantitative techniques by applying advanced methods, such as a bicomponent extended phase graph method for decomposing T2 attributable to water versus fat, and multiple b value diffusion to tease out perfusion effects. The authors presented significant differences and associations. However, the small effect sizes (eg, approximately 1–2-msec difference between DPN-positive, DPN-negaive, and control groups) may prove difficult to be detected with accuracy without first ensuring the reproducibility of the T2 mapping sequence, given the much larger interecho spacing of 7.6 msec used in this study (also typical of conventional T2 mapping).

Potential limitations include the absence of comparison against a clinical MR neurography protocol (that uses T2-weighted fat-suppressed sequences) and electromyography (only nerve conduction studies were used to diagnose DPN), which could be useful in translating quantitative MRI techniques to routine clinical use. Muscle biopsy was not performed for validation purposes, but this may not always be feasible, even in a research setting. While the 10-minute MRI protocol has been optimized in this work for quantitation, the low acquired spatial resolution (1.5 × 1.5 × 6 mm or 3 × 3 × 6 mm) relative to conventional morphologic sequences may make it difficult to reliably interrogate the small intrinsic foot muscles often involved in diabetes mellitus. Adding magnetization transfer ratio (7) and phosphorus 31 (³¹P) imaging (8) could bolster the quantitative panel. Magnetization transfer ratio and ³¹P imaging, respectively, provide information related to structural integrity or protein density and onset of ischemia. Ischemia may predispose patients with diabetes to early muscle atrophy.

Development of quantitative MRI markers of skeletal muscle remains an exciting research avenue toward providing objective physiologically sensitive metrics to detect and longitudinally monitor disease and treatment response. Such techniques could apply to many types of neuromuscular conditions; however, cost and efficiency concerns remain, particularly in the individuals with diabetic neuropathy who do not undergo routine MRI. Acquisition of quantitative MRI information may continue to compete with the need for high spatial resolution for detailed morphologic assessment of nerves and muscles, but recent technical advances, including accelerated MRI acquisitions, novel reconstruction schemes (9), and neural networks (10), could provide the means to meet both needs.