Comparison of Diffusion-Weighted MRI Using Single-Shot Echo-Planar Imaging and Split Acquisition of Fast Spin-Echo Signal Imaging, a Non-EPI Technique, in Tumors of the Head and Neck

DISCUSSION

This study offers a comprehensive evaluation of 2 DWI techniques (DWI-EPI and DW-SPLICE), by examining their differences in measured ADC values, geometric similarity to T1WIc MRI, and qualitatively scored distortion and image quality. Additionally, we evaluated their effectiveness in detecting recurrences of SCC in the HN region. Our findings indicate that DW-SPLICE surpasses DWI-EPI on 3T MRI in several areas, including reduced overall image distortion, higher image quality, and diagnostic accuracy for detecting recurrent and residual SCC. Thus, integrating DW-SPLICE into standard HN MRI protocols can significantly improve tumor assessment.

DW-SPLICE scans showed significantly better image quality and less overall distortion compared with DWI-EPI scans (Table 2). Although the tumor-specific distortion score was also better for DW-SPLICE, this difference was not statistically significant, possibly due to the smaller number of patients evaluated for this metric. These findings are in line with the results of Hirata et al¹⁷ and Fu et al,¹⁹ showing that DW-TSE sequences provide qualitatively better image quality and fewer artifacts than DWI-EPI in the oral and ocular regions, respectively.

Our mixed-effect linear models, which account for the random effects of the raters, show overall better scores for DW-SPLICE. However, a potential rater bias, as shown by the lack of agreement in overall distortion and image-quality weighted κ scores of DWI-EPI. shown in Table 2 and further visualized in the Supplemental Data, needs to be addressed. Several standardization and training measures such as a kick-off meeting and an evaluation after the first cases were conducted, yet more training may have been necessary or the score should have been simplified. Tumor-specific κ scores improved after simplifying the scoring into 2 categories: “no distortion” (including none and mild distortion) and “distortion” (encompassing moderate-to-severe distortion). Possibly, the time between the standardization measures and the actual evaluations also influenced the results, because disagreements were observed only in the overall distortion and quality scores and less in the tumor-specific distortion score, which was rated only in the first subgroup (Supplemental Data). Unlike DWI-EPI, DW-SPLICE did not encounter the same rater disagreement. This discrepancy could stem from the raters’ limited familiarity with DW-SPLICE, reducing unconscious biases and, thus, improving adherence to the original scoring guidelines. However, this could also indicate that DW-SPLICE may provide more consistent scores across raters due to better readability and just less reliance on experience with the sequence.

Because DWI is used as a functional technique, reproducibility of the already-established ADC values of DWI-EPI is of high importance.^17,20 Within this study, a high degree of ICC reliability is seen between the ADC values of the 2 diffusion-weighted techniques. Yet, a proportional bias was observed in ADC measurements of DW-SPLICE compared with DWI-EPI within the overall and benign lesion analyses (Table 3 and Fig 4). This indicates that DW-SPLICE ADC values of these groups are higher than the DWI-EPI ADC values, with this difference increasing for higher ADC values. Previous tests by Schakel et al⁴ that compared ADC values of DW-SPLICE and DWI-EPI using ice water phantoms did corroborate the established literature values. Additionally, a Bland-Altman analyses of Panyarak et al¹⁰ did not find a proportional bias in the salivary glands or lesions in the HN region of DW-TSE compared with DWI-EPI. In our study, variations in delineation might have contributed to the observed differences in ADC values. Delineations for DWI-EPI and DW-SPLICE were performed independently and blinded, with matching done retrospectively, which could have led to slight inconsistencies. This issue is particularly pronounced in benign lesions, which are harder to delineate due to variations in ADC or b = 1000 contrast and the variety in locations. This is strengthened by the largest discrepancies being observed in lesions located lower in the neck, with less signal from the coil, or that were incompletely depicted (Fig 4A, -B). Moreover, the presence of 2 lesions visible on DWI-EPI but not on DW-SPLICE and 3 lesions visible on DW-SPLICE but not on DWI-EPI indicate some variation in lesion visibility between the DWI techniques.

Overall, this issue suggests that ADC values derived from these 2 techniques should not be considered interchangeable. Instead, a single technique should be consistently used throughout the staging and follow-up timeline.

While DWI is used as a functional imaging technique relying on a perceived ADC value, enhancing geometric accuracy is essential for accurately detecting small lesions, particularly in areas with variable magnetic susceptibility. Reducing susceptibility-induced distortion common in DWI-EPI and improving geometric similarity helps prevent small lesions from being missed or misidentified as artifacts.^1,4,5 To measure the level of distortion across the 2 DWI techniques, we compared the overlap between the DWI delineations and the more anatomically accurate T1WIc delineations. Because T1WIc and DWI depict different aspects of lesions, a perfect match was not anticipated. However, we postulated that most of the areas discernible on DWI would fall within the T1WIc delineation. In our data, the DW-SPLICE showed a better delineation match compared with DWI-EPI before registering the image to the T1WIc (Supplemental Data), corroborating the results of Panyarak et al,¹⁰ which superimposed their DW-TSE and DWI-EPI on the T2-weighted imaging to calculate a distortion ratio. After registering the DWI to the T1WIc images, this effect dissipates as match percentages equalize, underlining that the same lesion information is likely available in both scans but just hampered by distortion. In clinical radiology practice, where no image registration is applied, radiologists may benefit from using the more accurate DW-SPLICE for locating and assessing lesions.

DWI is beneficial for managing patients with SCC because it is less affected by postradiotherapy effects and inflammation.^1,20 Therefore, this study focuses on comparing the diagnostic capabilities of DWI-EPI and DW-SPLICE in detecting recurrent or residual SCC after treatment. Our results indicate that DW-SPLICE was more effective in detecting residual or recurrent SCC at 1 year follow-up after the assessed MRI scan (P = .048). Raters agreed to base their assessments primarily on what was visible on DWI, though the pretreatment tumor location and stage and an additional sequence (T1WIc) were available for a limited clinical context. Raters lacked access to prior and/or baseline scans and full clinical and treatment details, which are typically considered in standard clinical practice. This omission shows that DW-SPLICE potentially is a superior technique to assess response.

To optimize diffusion-weighted image quality, we chose to include only scans obtained on 3T MRI scanners. Clinical MRI scanners with lower field strengths require longer diffusion gradients, which result in higher TE and noisier DWI. This requirement is especially a problem for DW-TSE like DW-SPLICE, which already has a longer acquisition.^4,17 Considering that clinical MRI scanners are now increasingly equipped with stronger-gradient hardware, we opted to test the DW-SPLICE under ideal circumstances.³ However, as susceptibility artifacts scale with field strength, scanning at higher field strengths may be less advantageous for DWI-EPI. This possibility suggests that our results might not be as pronounced at lower field strengths.¹⁷ Furthermore, as sensitivity encoding (SENSE) was applied to all DWI scans, standard SNR calculation was not feasible because the distribution of noise was not uniform over the scan.²¹

Other limitations include the retrospective study design, resulting in variations in lesion types, time between scan and treatment, and treatment types. In clinical practice, radiographers may adjust the field of view during scans. Although the dimensions of the DWI-EPI and DW-SPLICE scans are typically linked, in some cases, the field of view of the DWI-EPI scan was altered without applying the same adjustment to the corresponding DW-SPLICE scan. This step led to more variations in scan duration, resulting in a more varied TR for the DWI-EPI scans. While these TR variations are not expected to affect the ADC,²² the differences in FOV could potentially have a minor impact on how radiologists interpret the data. A prospective comparison of the 2 techniques in a more standardized group of patients would be of interest. The reliability of DW-SPLICE for detecting a recurrent and residual HNSCC mass at a lower field strength should be assessed in future work.

Aside from the DW-SPLICE, several other DWI methods exist to mitigate the artifacts seen for single-shot DWI-EPI scans, either by using a multishot EPI sequence or by applying the PROPELLER technique for reconstruction.^23,24 Other non-EPI techniques such as HASTE likewise show value in discerning cholesteatoma but do not always outperform DWI-EPI for HN SCC assessment.^7,25,26 Although DW-SPLICE outperformed the DWI-EPI sequence, a comprehensive comparison with all other alternative techniques would be of interest.