ESTRO 2024 - Abstract Book

S3850

Physics - Image acquisition and processing

ESTRO 2024

Hilde J.G. Smits 1 , Cornelis P.J. Raaijmakers 1 , Zeno A.R. Gouw 2 , Lilian N. Ruiter 3 , Koen Kuijer 1 , Mischa de Ridder 1 , Patricia A.H. Doornaert 1 , Joyce E. Lodeweges 1 , Frank A. Pameijer 4 , Tim Schakel 1 , Remco de Bree 5 , Chris H.J. Terhaard 1 , Gerben E. Breimer 3 , Jan W. Dankbaar 4 , Stefan M. Willems 6 , Marielle E.P. Philippens 1 1 University Medical Center Utrecht, Department of Radiotherapy, Utrecht, Netherlands. 2 Netherlands Cancer Institute, Department of Radiotherapy, Amsterdam, Netherlands. 3 University Medical Center Utrecht, Department of Pathology, Utrecht, Netherlands. 4 University Medical Center Utrecht, Department of Radiology, Utrecht, Netherlands. 5 University Medical Center Utrecht, Department of Head and Neck Surgical Oncology, Utrecht, Netherlands. 6 University Medical Center Groningen, Department of Pathology and Medical Biology, Groningen, Netherlands

Purpose/Objective:

As the precision of radiotherapy dose delivery has increased with modern conformal techniques, the accuracy of the target definition becomes more and more important for successful radiation treatment. A recently implemented geometrically accurate diffusion-weighted (DW-) MRI sequence 1 , referred to as DW-SPLICE, can potentially aid in tumor delineation, as it provides high contrast between the tumor and its surrounding tissue. The aim of this study was to determine the added value of DW-MRI on the inter-observer variability (IOV) and the accuracy of GTV delineations, using pathological tumor delineations as a ground truth.

Material/Methods:

Sixteen patients with laryngeal or hypopharyngeal carcinoma were included. After total laryngectomy, the specimen was cut into 3mm slices, from which 4μm sections were taken and stained with hematoxylin and eosin (HE). The 3mm slices were photographed and digitally stacked to create a 3D digital reconstruction of the specimen. This reconstruction was registered to the in vivo imaging via an ex vivo CT of the specimen 2 . The pathological tumor (tumor HE ) was delineated on the specimen reconstruction using the HE sections as a reference. Six observers delineated all tumors twice. In the first session, only anatomical MR imaging (T1, T1 with gadolinium, and T2) was provided. In the second session, the DW-SPLICE MR image with a b-value of 800 s/mm 2 and the apparent diffusion coefficient map were added. The IOV was determined by comparing the delineations of all observers in the general conformity index (CI gen ) 3 . The individual delineations were combined into majority voting delineations, comprising the volume that was delineated by four or more observers. The majority voting delineation of session one (GTV MRI ) and session two (GTV DW-MRI ) were compared to tumor HE , by calculating the positive predictive value (PPV) and the sensitivity. The PPV is the percentage of the GTV that contains tumor HE and the sensitivity is the percentage of tumor HE that is covered by the GTV. Additionally, the type I 95% Hausdorff distance (HD95) was calculated, representing the underestimation of the tumor boundaries by the GTV.

Results:

3 . The mean GTV

3 , compared to a mean

On pathology, the mean tumor HE volume was 11.0 cm

MRI volume was 18.2 cm

3 , Figure 1.

GTV DW-MRI volume of 15.5 cm