Relationship between magnetic resonance imaging findings and histological grade in spinal peripheral nerve sheath tumors in dogs

1 INTRODUCTION

Primary neoplasms of the peripheral nervous system in dogs represent 26% of nervous system tumors in dogs.¹ Peripheral nerve sheath tumors (PNSTs) are a group of neoplasms originating from either Schwann cells, modified Schwann cells, intraneural fibroblasts, or perineural cells.^{2, 3} They typically exhibit endoneurial longitudinal spread, which is distinct from the transperineurial growth pattern of nonneural tumors that might only occasionally encase and efface peripheral nerves.⁴ The latest edition of the World Health Organization (WHO) Classification of Tumors, published in 2021,⁵ groups the cranial and paraspinal nerve tumors together with tumors of the central nervous system listing benign entities (BPNSTs) such as neurofibroma, schwannoma, and perineuroma and malignant tumors, called malignant PNST (MPNSTs).

Peripheral nerve sheath tumors of low (BPNST) and high (MPNST) grades have been observed in both cranial and spinal nerves in the dogs.^{6, 7} Histologically, the most reliable indicators of malignancy are high mitotic index, necrosis within the neoplasm and transperineurial growth. Benign peripheral nerve sheath tumors have a low mitotic index, even in densely cellular areas, whereas MPNSTs exceed 4 mitotic figures per 10 high-power fields. Nuclear morphology and hemorrhages are also increasingly associated with malignancy.^{4, 7}

Surgical resection and radiotherapy are currently the main therapeutic strategies for these tumors.⁸ Unfortunately, these options are often not feasible by the time a diagnosis is made, particularly in the MPNST, because these tumors tend to invade the surrounding tissue, consistently infiltrating multiple nerve roots, plexus branches, and potentially the spinal cord or brainstem.⁹

Given that MPNSTs commonly invade and breach the epineurium of the affected nerves,^{10, 11} a higher recurrence rate after surgical resection is expected in dogs because of incomplete resection.¹² In people, the only effective treatment for MPNST is complete surgical resection to obtain negative surgical margins; patients with negative margins have a relatively higher survival rate than those with positive margins.¹³

Previous reports described the imaging findings and outcome of PNSTs in small cohorts of dogs.^{14, 15} Despite accurate spatial resolution, imaging techniques cannot definitively support a clinical diagnosis of PNST.^14-18 Differential diagnoses include other tumors, such as meningioma in the intradural compartment,¹⁹ lymphoma and sarcomas, or inflammatory lesions such as focal hypertrophic neuritis^{20, 21} and chronic hypertrophic ganglioneuritis,²² all presenting with comparable imaging findings.

Despite its lack of specificity, magnetic resonance imaging (MRI) is the diagnostic modality of choice to diagnose PNSTs,¹⁵ presumptively, because it allows for non-invasive early visualization of this entity.^23-25

The frequent occurrence of malignant behavior in dogs emphasizes the importance of predicting the histologic grade of the tumors in pre-surgical settings. In veterinary medicine, few studies^{14, 15} have described imaging findings on computed tomography (CT) and MRI in small cohorts of dogs without investigating possible associations with histologic phenotype.

Our objective was to assess MRI findings of a large group of dogs with a tailored histologic classification and grading for PNSTs to identify imaging features predictive of their malignancy.

2 MATERIALS AND METHODS

2.3 Histopathological diagnosis

Biopsy samples and tumor resections collected at surgery at referral centers were fixed in 10% neutral-buffered formalin and shipped to the laboratory. After gross inspection, samples were trimmed in longitudinal and transverse orientation, and subjected to paraffin-embedding and sectioning. Whenever margins and orientation of resected tissue were marked by the submitting surgeon, sections of distal and proximal margins were sampled and followed the same paraffin-embedding and sectioning protocol. All slides were stained with hematoxylin-eosin (H&E), as well as Giemsa stain according to standard protocols. Immunohistochemistry for selected markers including, but not limited to S100, vimentin, glial fibrillary acidic protein (GFAP), and smooth muscle actin (SMA) was requested only if investigating pathologists (KM, MR) deemed doing so necessary to reach the final diagnosis of PNST.

Histopathology slides from all tumors were reviewed and graded, applying diagnostic algorithms adapted from a modified French system (Fédération Nationale des Centres de Lutte Contre Le Cancer, FNCLCC) for grading human²⁶ and canine peripheral nerve sheath tumors⁷ for relevant morphologic features^{5, 27} by a board-certified veterinary pathologist (MR). The score is reported in Table 1 and an example of the different grades is depicted in Figure 1.

TABLE 1. Modified histological grading according to a modified French system (Fédération Nationale des Centres de Lutte Contre Le Cancer, FNCLCC) and additional features recorded for comparative analysis of histologic and MRI findings.

Tumor differentiation
Score 1	Closely resembling normal tissue
Score 2	Histological typing is certain
Score 3	Embryonal or undifferentiated sarcomas
Mitotic count (per 2.37 mm2)
Score 1	0-9 mitoses per 2.37 mm2
Score 2	10-19 mitoses per 2.37 mm2
Score 3	>19 mitoses per 2.37 mm2
Tumor necrosis
Score 0	No necrosis
Score 1	<50% tumor necrosis
Score 2	≥50% tumor necrosis
Histological grade	Grade 1: total score 2, 3 Grade 2: total score 4, 5 Grade 3: total score 6, 7, 8

Additional histologic features
Hemorrhage
Score 0	Absent
Score 1	<50% of tumor area
Score 2	>50% of tumor area
Stroma
Score 1	Delicate
Score 2	Moderate
Score 3	Abundant
Myxoid changes
Score 0	Absent
Score 1	<50% of tumor area
Score 2	>50% of tumor area
Cystic Changes
Score 0	Absent
Score 1	<50% of tumor area
Score 2	>50% of tumor area
Mineralization
Score 0	Absent
Score 1	Present
Inflammation
Score 0	Absent
Score 1	Mild
Score 2	Moderate
Pigments
Score 0	Absent
Score 1	Present
Margins
Negative	Free
Positive	Infiltrated
Epineurial Invasion
Score 0	Absent
Score 1	Present

In summary, tumors were histologically graded as grade 1 (score 2-3), grade 2 (score 4-5), or grade 3 (score 6-8) according to histological features of malignant behavior. Additional semiquantitative scoring of morphologic features included amount of stroma, cystic changes, myxoid changes, inflammation, hemorrhage, pigments, calcification, tumor margins (positive with tumor cell infiltration; negative without tumor cell infiltration), and epineurial invasion (Table 1).

The relationship between individual histologic and MR features and tumor grade were assessed one-by-one to identify imaging findings predictive of tumor malignancy.

3 RESULTS

Forty-four cases met the inclusion criteria. Breeds represented in the study included Labrador retriever (n = 9), cross-breeds (n = 8), West Highland White Terrier (n = 3), Jack Russell Terrier (n = 2), Border Terrier (n = 2), Beagle (n = 2), Border Collie (n = 2), Miniature Pinscher, American Staffordshire Terrier, Staffordshire Bull Terrier, Golder Retriever, Cavalier King Charles Spaniel, Fox Terrier, Magyar Vizsla, Doberman, Breton, Shi Tzu, Poodle, Magyar Vizsla, Yorkshire Terrier, Chihuahua, Coton de Tulear and Alaskan Malamute (n = 1 each). Of the 44 dogs, 30 were males (68.1%), of which 6/30 (20%) were spayed, and 14 were females (31.8%), of which 9/14 (64.2%) were spayed. The median age of the dogs was 9 years (IQR, 7-11 years; range, 2-13 years) and the median weight was 20 kg (IQR, 13-30 kg; range, 4-40 kg).

Dogs were referred because of lameness (20/44), paraparesis (8/44), tetraparesis (4/44), monoparesis (4/44), hemiparesis (4/44), ataxia (2/44), inability to jump (1/44), cervical hyperesthesia (1/44) and with the neurological examination were localized in the cranial cervical tract (11/44), cervicothoracic tract (18/44), thoracolumbar tract (4/44) and lumbosacral tract (11/44).

Tumor grade was not found to be associated with dog age (Kruskal-Wallis P-value, .67), body weight (Kruskal-Wallis P-value, .27) or sex (Fisher's exact P-value, .64).

3.2 Magnetic resonance imaging

The brachial plexus (C6-T2) was commonly affected (18/44, with 6/16 of grade 1, 5/19 of grade 2, and 7/9 of grade 3), followed by the cervical tract (C1-C5; 11/44, with in 4/16 of grade 1, 6/19 of grade 2, and 1/9 of grade 3) and lumbar plexus (L4-S3; 11/44, with 5/16 of grade 1, 6/19 of grade 2, and 0/9 of grade 3). The thoracolumbar tract (T3-L3) was the least common site (4/44, with 1/16 of grade 1, 2/19 of grade 2, 1/9 of grade 3; Figure 2A).

Most PNSTs involved both the intracanal and peripheral compartment, particularly grade 1 and grade 2. None of the grade 3 tumors involved only the intracanal tract (Figure 2B). Only 8/44 (18.1%) cases had a dumbbell shape (Figure 3).

The MRI findings of the different grades and of the PNSTs are presented in Table 4.

TABLE 4. Standardized grading scheme for magnetic resonance imaging (MRI) criteria used in 44 dogs with confirmed PNST.

	Score	Grade 1 (n = 16)	Grade 2 (n = 19)	Grade 3 (n = 9)	P
Numbers of nerve	1	10 (62.5%)	16 (84.2%)	6 (66.7%)
	2	3 (18.8%)	3 (15.8%)	2 (22.2%)
	3	3 (18.8%)	0 (0.0%)	1 (11.1%)	.35
Shape	Mass	0 (0.0%)	2 (10.5%)	2 (22.2%)
	Rounded	2 (12.5%)	6 (31.6%)	1 (11.1%)
	Tubular	14 (87.5%)	11 (57.9%)	6 (66.7%)	.18
Margins	Well-defined	12 (75.0%)	17 (89.5%)	8 (88.9%)
	Unclear	4 (25.0%)	2 (10.5%)	1 (11.1%)	.56
Foramen and fat obliteration	No	2 (12.5%)	6 (31.6%)	4 (44.4%)
	Yes	14 (87.5%)	13 (68.4%)	5 (55.6%)	.23
Enlargement of the foramen with bone atrophy	No	2 (12.5%)	6 (31.6%)	2 (22.2%)
	Yes	14 (87.5%)	13 (68.4%)	7 (77.8%)	.42
Dumbell shape**	Yes	3 (18.75%)	4 (21.05%)	1 (11.1%)
	No	13 (81.25%)	15 (78.95%)	8 (88.8%)	1
Muscle atrophy***	No	3 (18.8%)	1 (5.6%)	0 (0.0%)
	Yes	13 (81.3%)	17 (94.4%)	9 (100.0%)	.33
Grade atrophy	None	3 (18.8%)	1 (5.6%)	0 (0.0%)
	Mild	4 (25.0%)	8 (44.4%)	5 (55.6%)
	Moderate	5 (31.3%)	5 (27.8%)	2 (22.2%)
	Strong	4 (25.0%)	4 (22.2%)	2 (22.2%)	.73
Signal intensity in T1	Hyperintense	2 (13.3%)	3 (16.7%)	0 (0.0%)
	Isointense	13 (86.7%)	15 (83.3%)	8 (100.0%)	.70
Signal intensity in T2	Hyperintense	14 (87.5%)	14 (73.7%)	9 (100.0%)
	Hypointense	1 (6.3%)	1 (5.3%)	0 (0.0%)
	Isointense	1 (6.3%)	4 (21.1%)	0 (0.0%)	.49
Signal intensity in STIR	Hyperintense	14 (100.0%)	14 (93.3%)	9 (100.0%)
	Isointense	0 (0.0%)	1 (6.7%)	0 (0.0%)	1.00
Heterogeneity of the signal in T1	No	15 (100.0%)	16 (88.9%)	7 (87.5%)
	Yes	0 (0.0%)	2 (11.1%)	1 (12.5%)	.41
Heterogeneity of the signal in T2	No	8 (50.0%)	9 (47.4%)	4 (44.4%)
	Yes	8 (50.0%)	10 (53.6%)	5 (55.6%)	1.00
Contrast intensity	Mild	2 (12.5%)	3 (15.8%)	3 (37.5%)
	Moderate	2 (12.5%)	4 (21.1%)	1 (12.5%)
	Severe	12 (75.0%)	12 (63.2%)	4 (50.0%)	.63
Distribution of contrast medium	Homogenous	12 (75.0%)	16 (84.2%)	3 (37.5%)
	Intracanal	1 (6.3%)	1 (5.3%)	0 (0.0%)
	Peripheral	3 (18.8%)	2 (10.5%)	5 (62.5%)	.05*
Cystic portion	No	16 (100.0%)	19 (100.0%)	7 (87.5%)
	Yes	0 (0.0%)	0 (0.0%)	1 (12.5%)	.19
Necrosis	No	14 (87.5%)	13 (68.4%)	5 (55.6%)
	Yes	2 (12.5%)	6 (31.6%)	4 (44.4%)	.23
Fat split sign	No	12 (75.0%)	11 (57.9%)	8 (44.4%)
	Yes	4 (25.0%)	8 (42.1%)	5 (55.6%)	.32

• * P < .05.
• ** 30 dogs included.
• *** 1 of grade 2 was not available (amputation of the affected limb).

In general, PNSTs had overlapping imaging findings despite distinct histologic grades. Grade 3 PNSTs had larger volume (median, 4.18 cm³; IQR, 0.97-33.69 cm³) compared to grade 1 tumors (median, 0.97 cm³; IQR, 0.37-2.04 cm³; P = .03) although tumor volume was not significantly different between grade 1 and grade 2 tumors nor between grade 2 and grade 3 tumors (Figure 4).

On T1 post-contrast studies, grade 1 and 2 tumors had the highest frequency of homogenous distribution of contrast medium (75.0% and 84.2%, respectively) and grade 3 the lowest (37.5%; (P = .04; Figure 5A,B,D,E). Conversely, grade 3 tumors had the highest frequency of peripheral distribution (62.5%) and it was lower in grade 1 (18.8%) and grade 2 (10.5%) tumors (Figure 5C,F).

Two cases showed intracanal and peripheral locations with contrast enhancement only in the intracanal component and not the peripheral component (Figure 6). The intensity of the contrast enhancement was strong in 12/16 grade 1, 12/19 grade 2, and 4/9 grade 3 tumors; it was moderate in 2/16 grade 1, 4/12 grade 2, and 1/8 grade 3; and mild in 2/16 grade 1, 3/19 grade 2, and 3/8 grade 3 tumors. Significant relationships between tumor grade and the remaining MRI features were not identified.

The internal fat component was not included in the statistical analysis because of lack of consistent sequences from MRI protocols necessary to evaluate these findings. Fascicular and target signs, described in human patients,²⁶ were excluded because of the absence of consensus on these findings.

4 DISCUSSION

We analyzed the MRI features of PNSTs and their relationship with histological grade to identify imaging features predictive of malignancy. We found that different malignancy grades of PNSTs shared several MRI features. However, some qualitative MRI characteristics were significantly associated with the grade of PNSTs such as large tumor volume and distribution of contrast enhancement. Large tumor volume was significantly associated with high tumor grade. In contrast, homogenous contrast enhancement was significantly associated with tumors of low grade, whereas peripheral contrast enhancement was most consistently reported in high-grade tumors.

In our cohort of patients, grade 3 tumors were significantly larger than grade 1. In human medicine, tumor size is a discriminating factor between benign and malignant PNSTs and 5 cm in diameter is used as cut-off.²⁸ Rapid growth in malignant tumors is probably the reason behind this size dichotomy²⁹ and it likely reflects a more rapid onset and progression of clinical signs compared with low-grade tumors. Despite wider variations in the overall size of dogs compared to humans, we still could appreciate a significant contribution of tumor size to grade prediction and prognosis. As a general rule, the size of soft tissue sarcomas and the anatomic location impact the feasibility of complete surgical excision and tumor-free margins.³⁰ Large tumor size generally is considered a poor predictive factor for surgery and radiotherapy.³¹

In our study, homogenous contrast enhancement was significantly associated with tumors of low grade (grade 1), whereas peripheral contrast enhancement was most consistently reported in high-grade tumors (grade 2). In humans, peripheral enhancement is noted in malignant PNSTs, and central focal enhancement is observed in their benign counterparts.^{29, 32} Most likely, the degree of differentiation of benign tumors from normal tissue is such that proliferated cells and supporting fibrovascular stroma grow at a similar rate resulting in homogeneous growth of new tissue. The opposite might occur in high-grade tumors where cell proliferation, metabolic demand and nutritional and oxygen supply are out of balance as demonstrated by tumor necrosis and hemorrhage. According to previous studies in people,^{33, 34} peripheral enhancement also may reflect poor central vascularization and increased compactness of the tumor. Decreased extracellular space and abnormal vascular changes can result in thrombosis, hemorrhage, necrosis, and cyst formation.³³ Histopathology identified substantially increased frequency of intra-tumoral hemorrhage and amount of stroma deposition in high-grade PNSTs that would explain differences of contrast distribution among grades.

Enhancement of the intracanal compartment of the tumor without enhancement of the peripheral part was observed in 2 cases. This characteristic may be related to the absence of invasion of the tumor within the foraminal tract of the nerve, and the distal thickening of the nerve could be caused by edema from compression of the intradural mass. This hypothesis also may explain the thickening of >1 nerve that we observed.

In our cases, the brachial plexus was the most commonly affected site, particularly for high grade PNSTs because no grade 3 was encountered in the lumbar plexus. This result is in agreement with previous literature where few case reports described lumbosacral location of PNSTs.^{6, 35}

Surprisingly, no correlation was identified between spinal cord compression and grade of malignancy.

Grade 1 and grade 2 tumors mainly involved the intracanal compartments, whereas grade 3 mostly affected the peripheral segment of the spinal nerves. Regardless, the finding may not correlate with better prognosis, because gross tumor margins are difficult to identify on visual inspection during surgical procedures, and involvement of neighboring nerves or nerve roots limits the chance of complete resection within the space of the vertebral canal.¹⁵

Tubular shape on MRI was the most common imaging finding of PNSTs independent of grade according to previous studies in both humans²⁹ and veterinary patients.¹⁵ However, in our cases, only grade 2 and grade 3 had a mass-like appearance whereas in grade 1 a mass-like appearence was not observed. Regardless of the shape associated with more benign grades in our cohort, elongated shape and anatomic contiguity of a space occupying lesion adjacent to peripheral nerves and nerve roots is considered highly suggestive of PNST.³⁶

The term “dumbbell-shape” applies to a group of tumors arising along the vertebral column featuring 1 part of the tumor inside the spinal canal and the other part in the paravertebral space connected with each other by a portion of tumor traversing the neural foramen.³⁷ In our study, few cases (8/44, 18.1%) had this feature and no correlation was found with tumor grade.

In our cases, no relationship was found between ill-defined margins on MRI studies and high-grade tumors. In recent human medical literature, benign PNSTs had with well-defined margins, whereas malignant PNSTs had a tendency toward ill-defined margins or had an abnormal signal intensity relative to adjacent soft tissue.³⁶ Other authors considered these 2 features manifestations of aggressive neuroinvasive lesions, including malignant tumors, but other etiologies such as infectious or inflammatory lesions could not be reliably ruled out.²¹ Microscopic assessment of margin invasion did not identify correlations between positive invaded margins and tumor grade in our study, possibly reflecting anatomic restrictions to complete surgical resection rather than providing further information on tumor grade, behavior or both.

In humans, malignant PNST can cause lysis of the vertebrae, and this finding appears to correlate with grade of malignancy. In our study, none of the tumors caused detectable lysis of adjacent vertebrae on MRI, as previously reported in veterinary medicine.¹⁷ However, we observed mild bony changes consistent with pressure atrophy of the affected vertebrae (34/44, 77.2%). Pressure atrophy is a consequence of local impairment of blood supply resulting in thinning and rarefaction of bones in contrast to osteolysis derived from neoplastic infiltration and substitution of bone tissue.

Muscle atrophy is a frequently described secondary change that has been associated with PNSTs and it is reported in 23% of affected human patients.¹⁶ On the contrary, other soft tissue tumors only rarely are accompanied by muscle atrophy and compression or constriction of the peripheral or central nervous system is required to induce such changes. Chronic denervation of the target organ from motor nerve fiber loss adds to disuse of the affected limb, ultimately resulting in muscle atrophy.¹⁸ For some body areas and skeletal muscles, such as those innervated by the brachial plexus, volumetric changes can be subtle and may require measurements and comparison with the normal, clinically unaffected, side to become evident. For this reason, and concordance with previous literature,¹⁵ we recommend increasing the size of the field of view (FOV) by starting the MRI examination with the dorsal plane in order to visualize a large portion of axial and upper appendicular skeletal muscles for side comparison. In our study, muscle atrophy was a frequent finding (40/44), ranging from mild to severe but showed no correlation with tumor grade.

According to a previous study,¹⁵ T2- and T1-weighted (W) pre- and post-contrast pulse sequences were essential for PNSTs detection. In our study and according to the previous study, the STIR sequence was necessary because it increased the radiologist's ability to identify the affected nerves, suppressing signal from fat and making the hyperintense neoplastic lesions more noticeable.¹⁵ The STIR was available in 38/44 (86.3%) cases and it was used as the first sequence to decrease the time for identifying the nerve changes, because it is a fat-suppressing technique with high sensitivity for fluid and pathology³⁷ and it was particularly useful in studies when low-field MR is used, in which spectral fat suppression is not possible.³⁸

The “fat split-sign” (a fine ring of fat signal surrounding the bulk of the tumor) is a radiologic sign reported in radiology in human medicine and it is correlated with benign behavior because of the lack of infiltration of surrounding tissues. It is best appreciated on the peripheral nerves, such as the distal portion of the brachial or lumbar plexus, when cleavage is present between tumor and soft tissues. None of the patients in our study had the “fat split-sign”, which might be because of overrepresentation of high-grade tumors as well as higher frequency of PNSTs location within the vertebral canal or the foraminal tract. Nevertheless, radiologists should be aware of this sign and look for it.

Fascicular and target signs described in human patients²⁸ were excluded from analysis because of the absence of consensus among radiologists on these findings. These signs have never been described in veterinary medicine and there was no confidence in their identification between radiologists. The fascicular sign is detectable on T2W MRI images and is characterized by multiple small ring-like structures with peripheral hyperintensity representing bundles of nerve fibers within the peripheral nerves. The target sign consists of a central area of low intensity surrounded by a T2W hyperintense rim. Both signs correlate with benign tumor behavior according to the human medical literature,²⁸ but according to our observations, they appear to be of little value for grade prediction.

As a retrospective multi-center study, our study had some limitations. Because both low- and high-field MRI equipment was used in the study and the protocols were not standardized, it was not possible to differentiate among extradural, intradural and intramedullary location in some cases. Detection of hemorrhage, necrosis, cystic lesions, and metaplastic changes that contribute to establishing tumor grade also was also impaired for the same reason. The lack of the gradient sequence (T2*W and susceptibility weighted imaging [SWI]), the gold standard for detecting hemorrhage and venous structures,³⁹ and fluid-attenuated inversion recovery (FLAIR), the gold standard for detecting cystic lesions,⁴⁰ might have decreased the sensitivity of image analysis for further characterizing these specific changes and decreasing the capacity to identify a correlation between high tumor grade and the presence of necrosis, cystic lesion, and a hemorrhagic component. Hence, implementation of T2*W, SWI, and FLAIR sequences should be considered and tested in prospective studies for clarification of their contribution to radiographic prediction of tumor grade. Another limitation relates to on the composition of the studied PNSTs being clearly skewed toward high-grade tumors, that might have limited detection and description of imaging features typical of low-grade tumors. This result might simply reflect the natural selection of patients from secondary and tertiary referral centers for neurosurgical procedures and specialized neuropathology evaluation.

In conclusion, MRI provides features predictive of high-grade PNSTs including tumor size and peripheral contrast enhancement that might better guide patient management in terms of surgical planning and ancillary treatments. Introduction of a grading system for PNSTs in dogs in larger prospective studies ultimately will provide useful information on the prognostic utility of such a system, response to different treatments, disease progression and overall survival.

CONFLICT OF INTEREST DECLARATION

Luisa De Risio is employed by Linnaeus veterinary limited, a provider of veterinary services. No other authors declare a conflict of interest.

OFF-LABEL ANTIMICROBIAL DECLARATION

Authors declare no off-label use of antimicrobials.

INSTITUTIONAL ANIMAL CARE AND USE COMMITTEE (IACUC) OR OTHER APPROVAL DECLARATION

Approved by the Research Ethics Committee of the Animal Health Trust (Ref19/2019).

HUMAN ETHICS APPROVAL DECLARATION

Authors declare human ethics approval was not needed for this study.