6533b873fe1ef96bd12d48ba

RESEARCH PRODUCT

18F-FDG PET for breast cancer : combined analysis of tumour perfusion and metabolism for tumour characterisation and neoadjuvant chemotherapy response prediction

subject

description

Neoadjuvant chemotherapy (NAC) is a common treatment in patients with locally advanced or large breast cancer at diagnosis. A pathological complete response (pCR) at the end of NAC is recognized as a good surrogate marker of relapse-free survival. An early identification of the pathological response has then become a key parameter to monitor new therapeutic strategies. Studies, focusing on predictive biomarkers identification, have shown that early changes in tumour metabolism, assessed by Positron Emission Tomography (PET) using 2-desoxy-2-18F-fluoro-D-glucose (18F-FDG), allow the early assessment of the pathological response at the end of treatment. However, given the diversity of breast cancers and the complexity of the mechanisms underlying the tumour response, a multi-parametric approach is necessary. 18F-FDG PET, the gold standard for in vivo evaluation of tumour metabolism, also provides estimation of tumour blood flow using a dynamic acquisition performed immediately after injection. Furthermore, the evaluation of tumour heterogeneity, for both metabolism and perfusion, brings new insights into breast lesions characterization. The main objective of this thesis was then to evaluate the combined tumour metabolism and perfusion information, assessed by 18F-FDG PET perform before treatment, in order to predict pathological response to NAC. These biomarkers were also associated with clinical and biological parameters as well as texture parameters to quantify tumour heterogeneity. All the studies were conducted using a cohort of 246 patients. In a first analysis, we evaluated the impact of several texture features calculation methods on their relationships with the pathological response to NAC. The results showed that a relative rescaling for the PET images appears more appropriate to identify good responders. In a second study, conventional metabolism and perfusion parameters with their heterogeneity features were analysed according to tumour biological characteristics and molecular subtypes. Results showed that T3 and T4 stage tumours and tumours with lymph node involvement presented a higher perfusion, whereas no significant differences in SUV max or SUV mean were reported. In addition, tumour blood flow heterogeneity was different among breast cancer phenotypes and this was not previously observed with a global analysis of the perfusion. Finally, a treatment response prediction analysis was performed in a last study, with prognostic models using logistic regression and univariate feature selection. The best prediction performances were noted for models combining clinical and metabolic parameters (conventional and textural), whereas the tumour blood flow did not apparently improve the prediction.