Can imaging be the new yardstick for diagnosing peripheral neuropathy?—a comparison between high resolution ultrasound and MR neurography with an approach to diagnosis

A prospective, hospital-based study was performed over a period of 11 months, from February 2018 to December 2018 after taking ethical approval on a heterogenous group of 131 patients and 180 peripheral nerves. Consenting patients with clinical suspicion of peripheral mononeuropathy referred for ultrasound and MRI examination at a tertiary hospital without any contraindications to MRI. Any patient with symptoms, neurological assessment or radiological imaging (plain radiograph or MRI) suggestive of plexopathy or cord injury was excluded from the study. Electrodiagnostic findings (NCV) of each patient were recorded which was done by a technician with 8 years’ experience.

Data analysis was carried out using the commercially available SPSS program, Version 20.0.

Peripheral nerve has a characteristic honeycomb appearance on ultrasound, reflecting the fascicular composition of nerve on histopathology. Nerves can be identified in the transverse section at pre-determined landmarks and then followed proximally and distally to identify the site of pathology. Normal nerves tend to have a higher echogenicity when compared to the adjacent muscles and lesser when compared to the tendons. The pathological segment of the nerve will show decreased echogenicity, loss of the normal honeycomb structure, discontinuity of nerve, focal enlargement or neuroma formation. With imaging abnormalities in the surrounding soft tissue like compressive lesions, oedema and joint effusions can also be identified. We used a 14 Hz linear transducer for ultrasound examination and recorded the total time taken in each examination. Longitudinal scans were done to ascertain continuity and contour of the nerve and compare the thickness, while the area and echogenicity was determined on the transverse scans (Fig. 1a). All ultrasound examinations were done by the same radiologist using the same equipment having more than 6 years’ experience in the field. NCV findings of all patients were not disclosed to the operator before the examination [8, 11, 12, 14].

On the same day, a focused MRI using a 3T scanner, depending upon the suspected nerve involved was carried out with the protocol elaborated in Table 2. Due to a resource-limited setting, body coils were used in all settings. Gadolinium-based contrast (Gadodiamide at 0.2 mmol/kg) was used for assessment of enhancement if a neuroma was identified or in the presence of soft tissue masses in the course of the nerve of interest. A honeycomb pattern, similar to that on ultrasound, is visible on axial sections of high-resolution MRI (Fig. 1a). Anatomical map of the concerned nerve must be well versed in order to identify it in its short axis. Normal nerves have a hyperintense peripheral rim which corresponds to the fat in the nerve sheaths and hypointense ‘dots’ which correspond to the fascicles when seen in the short axis. This allows easy identification of nerves. Pathologies are identified by increased intraneural signal on T2 images. Fat-suppressed sequences are imperative to allow identification of this high signal which can easily be masked by the epineural or perineural fat [15, 16]. Focal enlargement, hyperintensity and altered fascicular patterns are signs of nerve pathology. The denervated muscles first begin to appear hyperintense on T2-weighted images only 48 h after the nerve injury. This is seen in axonotmetic and neurotmetic type of injuries. Disruption injuries cannot be reliably differentiated from contusion of the nerve on MRI [17]. When there is failure of regeneration of the fascicles, muscle atrophy and persistent hyperintensity is seen [16] while fatty infiltration is seen in chronic denervation. Susceptibility artefacts due to orthopaedic implants limit the use of MRI to some extent. The MR images in our study were read by another radiologist, with 9 years’ experience, blinded to the findings of ultrasound and NCV.

Presence or absence of neuroma, defined as a focal fusiform nerve swelling, was documented for each examined nerve. Neuroma can either be a pseudoneuroma or a true neuroma, both of which can be reliably differentiated by imaging studies. Compression or entrapment of the nerve corresponds to grade III Sunderland injury results in focal nerve enlargement known as pseudoneuroma. In its true sense, neuromas refer to nerve scarring with fibrosis and resultant loss of fascicular architecture, which in Sunderland’s terms corresponds to a grade IV injury. An intact intraneural architecture in an enlarged nerve is seen on HRUS and MRI in a pseudoneuroma while discontinuity of fascicular architecture is seen in a true neuroma [20, 21].

Thorough understanding of the anatomical relations of the target nerve is of utmost importance to identify the nerve and possible aetiology for the underlying pathology. Additional file 1 provides detailed anatomical information along with relevant figures to identify the nerves included in the study.

In our study, majority (28.24%) of the subjects belonged to the age group of 31–40 (Table 3) with a male to female ratio of 1.42:1. The median nerve was the most common nerve tested, comprising of 35.56% of total nerves tested (Table 4) while wrist was the most common site examined for neuropathy (36.11%) followed by the elbow (24.44%) and leg (17.22%) (Fig. 2). Overall, the most common aetiology of neuropathy was carpal tunnel syndrome (45 cases) while post-traumatic neuropathy was diagnosed in 37 cases.

Out of the 180 nerves tested, 150 tested positive on the gold standard, electrodiagnostic study (NCV) while 30 were negative on NCV. HRUS detected 137 of the positive cases while MRI proton density (PD) fat-saturated (FS) sequence detected 147 of the total positive cases. All 137 cases detected by HRUS were also detected by MRI except one in which the superficial branch of the common peroneal nerve was entrapped in the superficial fascia of the leg resulting in compressive neuropathy. Probably due to the millimetric size of the nerve, the alteration in signal intensity was not well perceived on MR while HRUS depicted focal change in calibre and echogenicity.

The diagnostic confidence of PD FS MR sequence was seen to be highest with a sensitivity of 95.34%. Only in seven out of 150 positive nerves, PD FS MR sequence failed to detect the abnormal nerve. T2 MR sequence showed a sensitivity of 73.33%. The sensitivity of HRUS in our study was 87.33%. HRUS could not identify the abnormal nerve in 19 out of 150 cases.

T1 MR sequence had the highest specificity of 96.67% followed by the T2 and PD FS sequences which had equal specificity of 86.67% each. HRUS was the least specific with a specificity of 80%. The positive predictive value was highest for T1 MR sequence (98.48%) while HRUS had the least positive predictive value (95.62%). A 78.79% negative predictive value was calculated for MRI PD FS sequence while T1 MR sequence had the lowest negative predictive value of only 25.44%.

Overall, the accuracy was highest for PD FS MR sequence (93.89%) followed closely by HRUS (80%). These findings are summarised in Fig. 3. The combined accuracy of both modalities was calculated to be 93.33% implying that MR examination with PDFS sequence alone will identify most pathological nerves. HRUS is more widely available, cost effective, quicker, feasible for evaluation of patients with MR unsafe implants and has a comparable diagnostic accuracy. These properties of HRUS can be employed in screening patients with suspected nerve pathologies.

In our study, six out of 37 cases of post-traumatic neuropathy had nerve discontinuity, all of which were diagnosed by ultrasound and MRI. In the remaining causes of non-traumatic neuropathy (infective, inflammatory, mass lesion, unclassified [24]), continuity was seen in 100% nerves. On comparing MRI and HRUS in assessment the continuity of the nerve, both were equally sensitive (100%) when compared to the surgical findings (obtained in 28/37 patients).

Neuroma formation was seen in 25 cases on MRI out of which HRUS did not detect three cases correctly. HRUS could detect 88% nerves with neuroma when compared with MRI (100%). Eighteen out of the 25 cases diagnosed to have a neuroma on imaging testified the same on surgical exploration. The remaining patients refused surgery.

Fifteen affected nerves had concomitant altered muscle intensities on MRI while HRUS detected increase in muscle echogenicity in only four cases.

MRI, on an average, took five times longer for a complete nerve study when compared to HRUS. Overall, the mean time taken for a complete MRI study was 37.32 min with a standard deviation of 5.36. HRUS required considerably less amount of time with a mean of 7.91 min and standard deviation of 2.05 min.

Idiopathic carpal tunnel syndrome (CTS), the most common aetiology diagnosed in our study, has been diagnosed, like all other neuropathies, conventionally by NCV testing. Table 5 summarises the findings in CTS with reference cut-off values for all the parameters assessed in this study. One added advantage of HRUS over MRI in this clinical entity is dynamic scanning. HRUS allows examination of the median nerve both at rest and during flexion of the wrist. In normal subjects, there is smooth translation of median nerve towards ulnar side on complete flexion with return to neutral position on extension. This movement is restricted in patients with CTS and is one of the parameters, albeit subjective, assessed by us in our study [25] (Fig. 4a–c).

Apart from this, radial tunnel syndrome (Fig. 5a–c) and cubital tunnel syndrome (Fig. 6a–d) were the other entrapment syndromes encountered in our study.

Compressive neuropathies include nerve compression by perineural or intraneural ganglion cysts (Fig. 7a–h). Post-traumatic compressive neuropathies can also result due to entrapment in local scar tissue, inflammatory tissue (Fig. 8a–c), compression by a hematoma (Fig. 9a, b) or entrapment in clot, as in one case of a patient with haemophilia (Fig. 10a, b).

Penetrating injuries can cause direct trauma to the nerves resulting in various grades of neuropathy as described by Sunderland and Seddon (Table 1). This may lead to a loss of continuity of the nerve with end neuroma formation (Fig. 11a–e), formation of a neuroma in continuity (Fig. 12a–d) or simply hyperintensity (Fig. 13a, b) depending upon the extent of the injury.

Leprotic involvement of peripheral nerves is not unheard of and may manifest in a variety of ways. It may present as an intradermal nerve involvement in a visible cutaneous patch or may occur as an isolated lesion in the distribution of the affected nerve [26]. The patient may or may not have neurogenic symptoms. On imaging, the nerve may appear hypoechoic with loss of fascicular architecture or have frank abscess formation, cystic change with intraneural hypervascularity (Fig. 14a, b).

PNSTs, either benign or malignant, include two entities—neurofibroma and schwannoma. These may present as soft tissue masses or may be incidentally detected depending upon their location, superficial or deep. PNSTs usually have a normal plain radiograph or may show enlargement of the soft tissue shadow. On ultrasound, a hypoechoic fusiform mass may be seen in continuity with the parent nerve which points towards its neurogenic origin. Differentiation between the two main differentials, PNST and neuroma is possible by MRI. On MRI, a T1 iso-hypointense, T2 hyperintense fusiform mass is seen in continuity with the parent nerve with central low signal on T2WI. This central T2 hypointensity corresponds to the higher fibro-collagen content which also co-relates with the organised histological architecture and is known as the ‘target sign’. Another sign is the split fat sign wherein the hyperintense fat on T1 images around the fusiform mass suggests its intermuscular location. Detection of fascicular pattern within the mass also supports its neurogenic origin [27]. Literature review shows discrepancy in the utility of post-contrast scans to differentiate between PNST and neuromas. While some [27] suggest that enhancement is seen in PNST and not in neuromas, others like Ahlawat et al. suggest that due to a broken blood-brain barrier, even 88% of traumatic neuromas are enhance. According to them, the absence of target sign and history of trauma are the most reliable signs of a traumatic neuroma [28] (Fig. 15a–f).