Publications Repository - Helmholtz-Zentrum Dresden-Rossendorf

2 Publications
Identification of tumour sub-volumes for improved radiomic risk modelling in locally advanced HNSCC
Leger, S.; Zwanenburg, A.; Pilz, K.; Lohaus, F.; Linge, A.; Zöphel, K.; Kotzerke, J.; Schreiber, A.; Tinhofer, I.; Budach, V.; Sak, A.; Stuschke, M.; Balermpas, P.; Rödel, C.; Ganswindt, U.; Belka, C.; Pigorsch, S.; Combs, S.; Mönnich, D.; Zips, D.; Krause, M.; Baumann, M.; Richter, C.; Troost, E.; Löck, S.;
Purpose/Objective: Radiomics aims to characterise the tumour phenotype using advanced image features to predict patient-specific outcome. Commonly, image features are calculated from the entire gross tumour volume (GTVe). However, tumours are biologically complex, e.g., expressing necrosis merely in the core and tumour cell proliferation at the periphery. The identification of sub-volumes to incorporate regional tumour variation into the risk models may lead to an improved outcome prediction. Therefore, we investigated different sub-volumes of the GTVe using CT imaging, developed radiomic signatures, and compared prognostic power and stratification performance of the signatures.
Material/Methods: A multicentre cohort consisting of 302 patients with advanced stage head and neck squamous cell carcinoma (HNSCC) was collected and divided into an exploratory and a validation cohort (208 and 94 patients, respectively). All patients received primary radio-chemotherapy at one of the six DKTK partner sites and underwent a non-contrast-enhanced CT scan for treatment-planning purposes. The analysis was divided into two subsequent steps (Fig. 1): (a) two distinct sub-regions were extracted from GTVe: the tumour boundary of different widths (3,5,10 mm) and the corresponding remaining core volumes. (b) extension of the highest prognostic tumour-boundary sub-volume by different widths (1,2,3,5 mm) beyond the GTVe. 1555 image features were extracted from each sub-volume. Different machine-learning algorithms were used to build radiomic models for the prediction of loco-regional tumour control (LRC). The prognostic performance was measured by the concordance index (C-Index). Finally, patients were stratified into groups of low and high risk of recurrence using the median risk value. Differences in LRC were evaluated by log-rank tests.
Results: The validation C-Index averaged over all learning algorithms and feature selection methods using the GTVe revealed a high prognostic performance for LRC (C-Index: 0.63±0.03 (mean±std)). The boundary sub-volumes GTV5mm and GTV10mm showed a slightly improved accuracy (C-Index: 0.64±0.03 and 0.64±0.02, respectively), while models based on the corresponding core volumes had a lower accuracy (C-Index: 0.59±0.03 and 0.60±0.03, respectively, (Fig. 2A)). Also the risk groups could be better separated using the GTV5mm (p<0.001), compared to the GTVe (p=0.005) and the corresponding core volume (p=0.16, (Fig. 2B)). The extension of the GTV5mm sub-region by 2mm led to a similar prognostic performance (C-Index: 0.65±0.03).
Conclusions: In our investigation, radiomic models based on the boundary of the tumour showed a higher prognostic performance for LRC compared to models based on the tumour core. This indicates that the tumour boundary may contain more prognostic information than other parts of the tumour. The identification of tumour sub-volumes associated with treatment outcome may further improve the performance of radiomic risk models.
  • Lecture (Conference)
    ESTRO 37, 20.-24.04.2018, Barcelona, España
  • Open Access LogoAbstract in refereed journal
    Radiotherapy and Oncology 127(2018), S263-S264
    DOI: 10.1016/S0167-8140(18)30818-1

Publ.-Id: 26230 - Permalink