ESTRO 2024 - Abstract Book

S1390

Clinical - Head & neck

ESTRO 2024

Netherlands. 8 University of Amsterdam, Amsterdam, Amsterdam Center for Language and Communication, Amsterdam, Netherlands. 9 The Netherlands Cancer Institute, Biostatistics Unit, Amsterdam, Netherlands. 10 University of Southern Denmark, Department of Regional Health Research, Odense, Denmark. 11 Amsterdam UMC Location Vrije Universiteit Amsterdam, Department of Otolaryngology and Head & Neck Surgery, Amsterdam, Netherlands. 12 Amsterdam UMC location University of Amsterdam, Department of Otolaryngology and Head & Neck Surgery, Amsterdam, Netherlands

Purpose/Objective:

In conventional practice, Diffusion-weighted imaging (DWI) uses single-shot echo planar imaging (SS-EPI) for its fast read-out but struggles with distortion predominantly near bone-air interfaces. Turbo Spin Echo (TSE) DWI, like split acquisition of fast spin echo signal for diffusion imaging (SPLICE), offers a potential solution, particularly in the head and neck region.[1-5] This study conducts a comparative analysis between SPLICE-DWI and SS-EPI-DWI concerning head and neck lesions (HN-lesions), with the aim of assessing the geometric accuracy, comparability and usability of the techniques in the context of radiology and radiation therapy delineation.

Material/Methods:

From August 2020 to January 2022, a total of 57 potential MRI exams for head and neck lesions (HN-lesions) were conducted, using both SS-EPI and SPLICE techniques on a 3.0T MRI system. Lesions >0.5cm³ were delineated on both DWI-MRIs. The study assessed differences in whole lesion ADC-values through interclass correlation (ICC), Bland-Altman plots, and mixed-effect model regression. The geometric accuracy of the DWI-MRIs was approximated using the Dice score of the registered DWI-MRIs delineation to the T1-weighted post contrast (T1W+c) delineation and analyzed using paired T-test. Additionally, two blinded radiologists (B.W., L.B.) evaluated image quality and distortion levels using a standardized questionnaire.

Results:

(Preliminary findings)

The eligible 41 MRI exams of 41 patients included 40 benign and 20 malignant HN-lesions, visible on both DWI techniques. Strong interclass correlation coefficients were found for mean ADC values of the SPLICE-DWI and SS EPI-DWI (ICC agreement 0.864 (95%CI 0.782-0.917), consistency 0.862 (95%CI 0.780-0.915)). The Bland-Altman regression showed discrepancies between SPLICE and EPI for higher average ADC values (ß=0.16, p=0.02), but not when assessing benign and malignant lesions separately. Radiologist evaluations of visual distortion showed a trend towards less distortion and better image quality for SPLICE-DWI. The assessment of geometric accuracy is an ongoing aspect of the study, with current findings indicating a superior Dice score for SPLICE-DWI in comparison to SS-EPI-DWI.

Conclusion:

High agreement and consistency in mean ADC values were seen for HN-lesions measured with both techniques. Qualitative analyses of radiologists resulted in a trend towards better distortion and image quality scores for