Critical analysis of the effect of various methodologies to compute breast cancer tumour blood flow-based texture features using first-pass 18F-FDG PET. (November 2022)

• Record Type: Journal Article
• Authors: Payan, Neree; Presles, Benoit; Truntzer, Caroline; Courcet, Emilie; Coutant, Charles; Desmoulins, Isabelle; Brunotte, François; Vrigneaud, Jean-Marc; Cochet, Alexandre
• Is Part Of: Physica medica. Volume 103(2022)
• Page Start: 98
• Page End: 107
• Publication Date: 2022-11
• Subjects: Breast cancer -- 18F-FDG PET/CT -- Blood flow -- Texture features; Medical physics -- Periodicals; Biophysics -- Periodicals; Biophysics -- Periodicals; Imagerie médicale -- Périodiques; Radiothérapie -- Périodiques; Rayons X -- Sécurité -- Mesures -- Périodiques; Physique -- Périodiques; Médecine -- Périodiques; 610.153
• DOI: 10.1016/j.ejmp.2022.09.015
• Languages: English
• ISSNs: 1120-1797

Abstract

Abstract: Purpose: Assessment of tumour blood flow (BF) heterogeneity using first-pass FDG PET/CT and textural feature (TF) analysis is an innovative concept. We aim to explore the relationship between BF heterogeneity measured with different TFs calculation methods and the response to neoadjuvant chemotherapy (NAC) in patients with newly diagnosed breast cancer (BC). Methods: One hundred and twenty-five patients were enrolled. Dynamic first-pass and delayed FDG PET/CT scans were performed before NAC. Nine TFs were calculated from perfusion and metabolic PET images using relative (RR) or absolute (AR) rescaling strategies with two textural matrix calculation methods. Patients were classified according to presence or absence of a pathologic complete response (pCR) after NAC. The relationship between BF texture features and conventional features were analysed using spearman correlations. The TFs' differences between pCR and non-pCR groups were evaluated using Mann–Whitney tests and descriptive factorial discriminant analysis (FDA). Results: Relation between tumour BF-based TFs and global BF parameters were globally similar to those observed for tumour metabolism. None of the TFs was significantly different between pCR and non-pCR groups in the Mann–Whitney analysis, after Benjamini–Hochberg correction. Using a RR led to better discriminations between responders and non-responders in the FDA analysis. The best results were obtained by combining all the PET features, including BF ones. Conclusion: A better differentiation of patients reaching a pCR was observed using a RR. Moreover, BF heterogeneity might bring a useful information when combined with metabolic PET parameters to predict the pCR after neoadjuvant chemotherapy. Highlights: BF-based texture feature values are impacted by the feature computation methodology. A relative rescaling led to better differentiation between pCR and non-pCR. No global BF or metabolic parameter was significantly different between patients. BF and metabolic parameters combined led to the best patient differentiation.