Texture analysis of high-resolution dedicated breast 18 F-FDG PET images correlates with immunohistochemical factors and subtype of breast cancer

• Published in: European journal of nuclear medicine and molecular imaging Vol. 45; no. 2; pp. 196 - 206
• Main Authors: Moscoso, Alexis, Ruibal, Álvaro, Domínguez-Prado, Inés, Fernández-Ferreiro, Anxo, Herranz, Míchel, Albaina, Luis, Argibay, Sonia, Silva-Rodríguez, Jesús, Pardo-Montero, Juan, Aguiar, Pablo
• Format: Journal Article
• Language: English
• Published: Germany Springer Nature B.V 01-02-2018
• Subjects: Breast cancer; Breast Neoplasms - diagnostic imaging; Breast Neoplasms - metabolism; Breast Neoplasms - pathology; Cancer; Classification; ErbB-2 protein; Feature extraction; Female; Fluorodeoxyglucose F18; Glycolysis; Heterogeneity; High resolution; Humans; Image Processing, Computer-Assisted; Image resolution; Immunohistochemistry; Male; Middle Aged; Positron emission; Positron emission tomography; Scanners; Signal-To-Noise Ratio; Tomography; Tumors; Dedicated breast; Heterogeneity; Texture analysis; Breast cancer; 18F-FDG; PET
• ISSN: 1619-7070; 1619-7089
• DOI: 10.1007/s00259-017-3830-1

This study aims to determine whether PET textural features measured with a new dedicated breast PET scanner reflect biological characteristics of breast tumors.
One hundred and thirty-nine breast tumors from 127 consecutive patients were included in this analysis. All of them underwent a
F-FDG PET scan before treatment. Well-known PET quantitative parameters such as SUV
, SUV
, metabolically active tumor volume (MATV) and total lesion glycolysis (TLG) were extracted. Together with these parameters, local, regional, and global heterogeneity descriptors, which included five textural features (TF), were computed. Immunohistochemical classification of breast cancer considered five subtypes: luminal A like (LA), luminal B like/HER2 - (LB -), luminal B like/HER2+ (LB+), HER2-positive-non-luminal (HER2pnl), and triple negative (TN). Associations between PET features and tumor characteristics were assessed using non-parametric hypothesis tests.
Along with well-established associations, new correlations were found. HER2-positive tumors had significantly higher uptake (p < 0.001, AUCs > 0.70) and presented different global and regional heterogeneity (p = 0.002, p = 0.016, respectively, AUCs < 0.70). Nine out of ten analyzed features were significantly associated with immunohistochemical subtype. Uptake was lower for LA tumors (p < 0.001) with AUCs ranging from 0.71 to 0.88 for each subgroup comparison. Heterogeneity metrics were significantly associated when comparing LA and LB - (p < 0.01), being regional heterogeneity metrics more discriminative than any other parameter (AUC = 0.80 compared to AUC = 0.71 for SUV). LB+ and HER2pnl tumors also showed more regional heterogeneity than LA tumors (AUCs = 0.79 and 0.84, respectively). After comparison with whole-body PET studies, we observed an overall improvement in the classification ability of both non-heterogeneity metrics and textural features.
PET parameters extracted from high-resolution dedicated breast PET images showed new and stronger correlations with immunohistochemical factors and immunohistochemical subtype of breast cancer compared to whole-body PET.