Readout-segmented diffusion tensor imaging for assessing sciatic nerve invasion by soft tissue tumor

There are few publications regarding feasibility of readout-segmented diffusion tensor imaging (rsDTI) in assessing nerve invasion by soft tissue tumor. 64 patients with soft tissue mass in upper leg suspected of sciatic nerve invasion underwent rsDTI. Nerve invasion was conrmed in 28 cases by operation or electromyogram. The sciatic nerve was better depicted with diffusion weighted map versus b = 0 map of rsDTI. Inter-reader agreement in using rsDTI to rate nerve invasion was excellent. Sensitivity and specicity of rsDTI in identifying nerve invasion were 93% (26/28) and 92% (33/36) respectively. Apparent diffusion coecient (ADC) was signicantly higher in invaded nerves versus normal nerves (1.45 ± 0.67 × 10 − 3 mm 2 /s vs. 1.39 ± 0.46 × 10 − 3 mm 2 /s, P(cid:0)0.05). DTI derived FA was signicantly lower in invaded nerves versus normal nerves (0.22 ± 0.11 vs. 0.37 ± 0.13, P(cid:0)0.05). Readout-segmented DTI was feasible in assessing sciatic nerve invasion by soft tissue tumor in selected patients.


Introduction
Soft tissue mass of leg usually invades adjacent vessels and nerves. 1,2 Computed tomography angiography (CTA) is the rst-line test in revealing arterial invasion by tumor. 2 Magnetic resonance angiography (MRA) serves as an alternative to CTA. 2 However, CTA and MRA have di culty in identifying nerve invasion by tumor. 3 In recent years, fast spin echo (FSE) T2 weighted imaging (T2WI) was found useful in depicting peripheral nerve. 4,5 However, conventional T2 MR provides no quantitative parameter of nerve besides signal intensity. 6 Quantitative MR parameters are still urgently required to re ect the intrinsic character of nerve.
Diffusion weighted (DW) MR is considered as a quantitative sequence. Diffusion weighted imaging (DWI) and diffusion tensor imaging (DTI) are widely used in clinical practice. However, conventional DW MR was limited in extremities due to image distortion and susceptible artifacts. 7 In recent years, readoutsegmented method was frequently used for diffusion weighted MR to overcome distortion and artifacts. 8, 9 By using multiple readout segments, depiction of lesions of limb can be signi cantly improved. 10,11 It is well established water molecular motion is more restricted in nerve versus muscle. DWI and DTI thus have the potentiality to separate peripheral nerve from background muscles. 12,13 Compared to DWI, DTI uses more diffusion directions, and provides fraction of anisotropy (FA) besides apparent diffusion coe cient (ADC). FA was widely used in central nervous system; however, there are few publications regarding FA of peripheral nerve invaded by tumor. We here hypothesize sciatic nerve could be identi ed in DTI images. The purpose of the study is therefore to investigate the feasibility of readout-segmented DTI (rsDTI) in assessing sciatic nerve invasion by soft tissue mass.

Methods
All methods were carried out in accordance with relevant guidelines and regulations.

Study Design And Participants
This prospective study was approved by the Ethics Commission of Hospital (TJ-2015-075). Each patient signed an informed consent before inclusion of study.
Inclusion criterion was patients with soft tissue mass in upper leg suspected of sciatic nerve invasion. Exclusion criteria were: 1) standard contraindication to MR examination (such as claustropibia); 2) metal internal xator in upper leg; 3) underwent chemotherapy or radiotherapy before MR; 4) nerve invasion not con rmed by surgical operation or electromyogram (EMG). From September 2016 to October 2020, 88 patients with soft tissue mass in upper leg suspected of nerve invasion were enrolled in the study. 24 patients were excluded due to the following reasons: claustrophobia (n = 1); metal internal xator (n = 3); chemotherapy or radiotherapy (n = 7); no con rmation of nerve invasion (n = 13). The remaining 64 patients who underwent rsDTI were nally analyzed.

Data Collection
All MR examinations were performed on a 3T whole-body scanner (Siemens, Skyra). The region of interest (ROI) in the portion of leg was covered with an 18-element body coil. Fat-suppressed FSE T2WI was performed with the following parameters for the purpose of localizing mass: TR/TE, 4000/80 ms; matrix, 320; eld of view, 320 mm or greater; slice thickness, 4 mm; slice number, 20 or more. Readoutsegmented DTI was performed in the transverse plane covering the soft tissue mass with the following parameters: TR/TE, 5500/78 ms; matrix, 240; eld of view, 240 mm × 240 mm or greater to t of mass size; phase direction, anterior-posterior; slice thickness, 4 mm; slice number, 24 or more; b-value, 0 and 800; average for b = 0, 3; average for b = 800, 1; directions, 20; readout segments, 5; scanning time, 6 ~ 8 minutes. ADC map and FA map were automatically generated by the scanner.

Data analysis
Two radiologists with 7 and 9 years' experience in diagnosing peripheral nerve diseases in consensus assessed the depiction of sciatic nerve on b = 0 map, diffusion weighted (DW) map, ADC map and FA map. Subjective score was given to each map according to the following scale: 0 = poor depiction of nerve, identi cation from background impossible; 1 = moderate depiction, with nerve identi cation and ROI assessment possible; 2 = good depiction of nerve, adequate nerve-to-background contrast; 3 = excellent depiction of nerve.
Two radiologists with 8 and 11 years' experience in diagnosing peripheral nerve diseases separately assessed the invasion of sciatic nerve using all rsDTI maps. A subjective score was given to each case according to the following scale: 0 = nerve free from tumor; 1 = tumor very close to nerve, nerve boundary clear; 2 = no gap between tumor and nerve, part of nerve boundary not clear; 3 = nerve boundary not seen, separation of nerve from tumor impossible. In the current study, invasion score 0 and 1 were considered as nerve non-invasion, while score 2 and 3 as nerve invasion. The two radiologists separately measured FA and ADC of sciatic nerve by drawing ROI. ROI was drawn at ve consecutive slices where soft tissue mass and sciatic nerve were closest. ROI was drawn on DW map rst, and then transferred to FA map and ADC map. ROI analysis of nerve was abandoned if separation of sciatic nerve from tumor impossible. The two readers also measured FA and ADC of soft tissue mass by drawing ROI at the slices where mass size was the largest. Necroses, vessels and fat were asked to be avoided when drawing ROI on mass. Values from multiple slices were averaged.

Statistical analysis
All statistical analysis was performed using SPSS 22.0. The score of nerve depiction was compared between rsDTI maps using a Wilcoxon sign-bank test. Cohen kappa coe cient was used to determine the inter-reader agreement in rating nerve invasion. Intra-class correlation coe cient (ICC) was calculated to determine the inter-reader reproducibility in measurement of DTI parameters. ICC 0.7 was considered excellent inter-reader reproducibility, thus data from two readers could be averaged. A Mann Whitney test was used to compare DTI parameters of sciatic nerve between nerve invasion cases and non-invasion cases. Receiver operator characteristic (ROC) curve was constructed for further evaluation. P value less than 0.5 was considered as statistical signi cance.

Results
From September 2016 to October 2020, 64 patients (male: female = 34:30; mean age = 43.4 years; age range = 17 ~ 71) underwent rsDTI for assessing sciatic nerve invasion by soft tissue mass. Sciatic nerve invasion was con rmed in 28 out of 64 cases (15 by operation, 13 by EMG), while non-invasion was con rmed for the other 36 cases (22 by operation, 14 by EMG).
ICC for ADC and FA of sciatic nerve were 0.91 and 0.88 respectively. ICC for ADC and FA of soft tissue mass were 0.93 and 0.91 respectively. As all ICC above 0.7, data from the two readers were averaged. ADC of sciatic nerve was signi cantly different between non-invasion cases and nerve invasion cases ADC and FA were signi cantly lower in soft tissue sarcoma versus vascular anomalies (see Table 4). AUC for ADC and FA were 0.72 and 0.84 respectively in discrimination of tumor. Figure 1 shows a comparison among DTI maps for depiction of sciatic nerve. Figures 2, 3, 4 and 5 show nerve invasion score 0, 1, 2 and 3 respectively.

Discussion
In the current study, we investigated the feasibility of readout-segmented DTI in assessing sciatic nerve invasion by soft tissue mass. The most important ndings were as follows: 1) sciatic nerve was best depicted with diffusion weighted map; 2) rsDTI had 93% sensitivity and 92% speci city in identi cation of sciatic nerve invasion; 3) FA value of invaded nerve decreased.
It is necessary to assess peripheral nerve invasion by tumor before producing treatment plan. [14][15][16] In recent years, several MR sequences have been tried for peripheral nerve diseases. [17][18][19] However, they fail to provide objective parameters to re ect intrinsic characteristics of nerve. Readout-segmented DTI was thus introduced by us as a new way to assess peripheral nerve, as it could provide quantitative parameters including FA and ADC. [20][21][22][23] The b = 0 map was T2 weighted. We actually found b = 0 map of rsDTI was similar to FSE T2WI image, which was valuable in assessing peripheral nerve diseases. [24][25][26] However, we found FSE T2WI, as well as b = 0 map, failed to well depict sciatic nerve when nerve and background tissue had similar signal intensity (SI). In contrast, sciatic nerve and background generally had substantially different signal in diffusion weighted map. A possible explanation is that water molecular diffusion is more restricted in nerve versus background tissue. 27,28 From b = 0 to b = 800, the signal decrease was less for nerve, so it was brighter than background on b = 800 map. Compared to T2 weighted image, diffusion weighted MR seemed more suitable for depicting peripheral nerve. On diffusion weighted map, nearly all soft tissue masses and nerves were hyperintensity, while background tissues were hypointensity or isointensity, making it possible to assess the relationship of nerve and tumor. In fact, the inter-reader agreement of rsDTI in assessing nerve invasion was excellent (kappa = 0.911).
Water molecular motion could be prevented by myelin sheath of nerve, resulting in diffusion difference in transverse and longitudinal directions, which is described with FA. For diseased nerve, destroy of myelin sheath is expected to cause FA to decrease. 29,30 Our data seemed to support this hypothesis. We found FA was lower in invaded nerves versus normal nerves. Invaded and normal nerves also differed in ADC, which is not easy to explain. It is worth noting that AUC of FA was higher than that of ADC. DTI derived FA seemed more suitable than DWI derived ADC in identifying nerve invasion.
Soft tissue sarcoma is the most common malignant soft tissue tumor of leg, while vascular anomalies the most common benign soft tissue tumor. We performed a comparison in DTI parameter between the two tumors, and found ADC was lower in soft tissue sarcoma versus vascular anomalies, in consist with previous publications. 31 Interestingly, we found FA was also lower in soft tissue sarcoma. A possible explanation is that tissue structure of soft tissue sarcoma is extremely disordered, 32 resulting in similar diffusion in all directions. DTI derived FA seemed suitable for discrimination of soft tissue sarcoma and vascular anomalies, with AUC above 0.8.
Our study had some limitations. First, the sample size was small. Soft tissue tumor in leg was not a common disease. Large tumor with sciatic nerve invasion was not easy to collect. We only collected 64 cases during nearly four years. Multi-center large-size study is required to validate our results. Second, nerve depiction on diffusion weighted map was compared to that on b = 0 map, but not directly to FSE T2WI. We found b = 0 map of rsDTI was very similar to FSE T2WI, and nerve depiction was nearly the same on the two maps. Nerve depiction on rsDTI should be compared to other sequences in future study. Third, EMG is not a gold standard in determining nerve invasion. The accuracy of rsDTI in identifying nerve invasion may be in uenced by this imperfect standard. Fourth, FA and ADC of nerve could not be obtained when separation of nerve and tumor was unavailable. Other MR sequences (such as contrast enhanced T1) are still required to separate nerve from tumor for such cases.
In conclusion, readout-segmented DTI was feasible to assess sciatic nerve invasion by soft tissue tumor in selected patients.

Declarations Data availability
After publication, the data will be made available to others on reasonable requests to the corresponding author.

Competing interests
We declare no competing interests.  Figure 1 Comparison of nerve depiction among maps of readout-segmented DTI. a, b, c and d were b=0 map, diffusion weighted map, ADC map and FA map respectively. T2 weighted b=0 map provided poor depiction of sciatic nerve. The sciatic nerve was well depicted on diffusion weighted map (b, arrows), and was brighter than background tissue. ADC map provided acceptable depiction of nerve (c, arrows), but FA map did not (d).
Neither axial FSE T2WI (b) nor b=0 map of readout-segmented DTI (c) well depicted sciatic nerve. It was di cult to determine whether sciatic nerve was invaded by tumor on b or c. Diffusion weighted map (d) could well depict sciatic nerve (d, arrows), which was free from tumor. ADC map (e) and FA map (f) provided poor depiction of nerve. The electromyogram for this patient was normal.

Figure 3
Score 1 of nerve invasion. a, b, c and d were FSE T1WI, fat-suppressed FSE T2WI, b=0 map and diffusion weighted map respectively. The lipoma was hyperintensity on T1WI, and hypointensity on fat-suppressed T2WI. Sciatic nerve was similar to adjacent muscle in T1 signal (a, arrows). On b=0 map (c) and diffusion weighted map (d), sciatic nerve was brighter than background (c, d, arrows). The clear nerve boundary indicated non-invasion, which was con rmed by operation.
Page 15/16 Figure 4 Score 2 of nerve invasion. a, b and c were b=0 map, diffusion weighted map and ADC map of readoutsegmented DTI respectively. Tumor and nerve were both hyperintensity on diffusion weighted map (b, arrows). There was no gap between tumor and nerve. Part of nerve boundary was not clear, indicating nerve invasion, which was con rmed by operation. The pathology result was schwannoma (d).

Figure 5
Score 3 of nerve invasion. a, b, c and d were respectively axial b=0 map, axial diffusion weighted map, coronal b=0 map and coronal diffusion weighted map. The tumor was hyperintensity on diffusion weighted map. Normal sciatic nerve in the right side could be seen on axial DTI (a, b, arrows). The beginning part of sciatic nerve in the left side was well depicted on coronal DTI (c, d, arrows), but it was impossible to separate other part of nerve from tumor. Nerve invasion was con rmed by operation. The pathology result was plexiform schwannomas (e).