This study aims to use quantitative metrics to compare magnetic resonance imaging (MRI) segmentation utilizing Double Inversion Recovery (DIR) and Fluid-Attenuated Inversion Recovery (FLAIR) sequences for Tumor Treating Fields (TTFields) treatment planning.

TTFields therapy is an established adjuvant treatment for glioblastoma patients in combination with surgery, radiation and temozolomide chemotherapy. Treatment planning requires modeling TTFields delivery inside the brain to determine the dose targeting the tumor, and segmentation of the white matter is the most time-consuming process in the workflow. DIR sequence appears to provide better contrast between white and gray matters than FLAIR.

Semi-automated segmentation of the white matter from 3 glioblastoma patients was performed using DIR and FLAIR sequences according to established methods (PMID: 29023236 and 28544575). Dice coefficient and Interaction over Union (IoU) indices were computed for the segmented white matters, together with the volume of the structures and time required for segmentation.

Median Dice coefficient was 0.704 (range 0.678-0.741) and IoU index was 0.543 (range 0.512-0.589).  The median volume of white matter was 327.9 cc (range 317.8-390.5) when segmented using DIR and 343.9 cc (range 324.3-532.8) when segmented using FLAIR.  The time required to complete the segmentation was about 30-45 minutes less for DIR compared to FLAIR.

The Dice coefficient and IoU indices indicate that segmentation using DIR is less prone to over-contouring compared to FLAIR, probably because DIR provides better contrast between white and gray matters. The volumes are comparable but segmentation using DIR requires less time and fewer manual corrections.

Supported in part by The Rhode Island Life Science Hub and The Musella Foundation.