Klin Padiatr 2020; 232(02): 112
DOI: 10.1055/s-0040-1701940
16 May 2020
S-IX
Session IX: Staging Evaluation and Response Criteria Harmonization (SEARCH) for CAYAHL
© Georg Thieme Verlag KG Stuttgart · New York

CT-morphologic vs. metabolic tumor volume in pediatric Hodgkin-Lymphoma - A comparison of 19 different semiautomatic delineation methods

G Friedrich
1   Clinic for nuclear medicine, University Leipzig, Leipzig, Germany
,
D Stoevesandt
2   Clinic for radiology, University Halle, Halle, Germany
,
L Kurch
1   Clinic for nuclear medicine, University Leipzig, Leipzig, Germany
,
TW Georgi
1   Clinic for nuclear medicine, University Leipzig, Leipzig, Germany
,
D Körholz
3   Clinic for pediatric oncology, University Gießen, Gießen, Germany
,
C Mauz-Körholz
3   Clinic for pediatric oncology, University Gießen, Gießen, Germany
,
R Kluge
1   Clinic for nuclear medicine, University Leipzig, Leipzig, Germany
› Author Affiliations
Further Information

Publication History

Publication Date:
18 March 2020 (online)

Also available at
 
 

    Introduction The aim of this study was to evaluate the performance of automated tumour delineation methods for 18F-FDG PET (PET) and compare the metabolic tumour volume with the CT volume, to investigate factors potentially affecting the segmentation result and to deduce recommendations for the use of certain delineation methods.

    Methods 19 different PET delineation methods were applied to assess the primary metabolic tumour volume (MTV) of the mediastinal tumour mass of 63 paediatric patients with untreated Hodgkin lymphoma. 13 segmentations based on region growing with varying SUV-thresholds (Hermes Hybrid 3D). Additionally, 6 based on two different iterative threshold algorithms (3 voxel- and 3 lesion-specific thresholds) (ABX-Rover).

    The results were compared with manual CT morphologic delineations (GTV) and visual optimal metabolic volume (vMTV) to determine the volume error (VE). Additionally, potential influence factors like the SUVmax, the coefficient of variance (COV and the size of the lesion were investigated. A detailed analysis of the origin of deviations was also accomplished.

    Results 18 of 19 delineation methods and the vMTV (median and range: 75,60ml; 4.10ml-389,70ml) were systematically smaller than the GTV (Med.: 141,80ml; 4,80ml-520,8ml) therefore, GTV is not a suitable reference standard for MTV in HL.

    Mean VE with vMTV was lowest for an iterative voxel-specific-threshold (0,74%; SD: 27,44%), the best region growing algorithms work less exact and satisfactory (3SUV: 6,68%, SD: 63,59%; -41%SUVpeak: -4,95%, SD: 41,42%).

    Conclusion GTV and MTV are different parameters in HL and GTV is not a suitable reference standard. Iterative segmentation methods, especially voxel-specific thresholds, measure the MTV more satisfactory, reproducible and less varying than region growing algorithms.


    #