This study aimed to investigate the discriminative performance of pharmacokinetic quantitative parameters with histogram and texture features on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in predicting axillary lymph nodes metastasis in breast cancer. Results showed that 6 pharmacokinetic quantitative and their histogram parameters,53 texture features of DCE-MRI were statistically difference between the positive and negative group. In cross-validation, the accuracy of the classifier obtained 90% for identifying the metastatic axillary lymph node. The radiomic features based on quantitatively pharmacokinetic DCE-MRI demonstrated promising application in discriminating between metastatic positive and negative axillary nodes of breast cancer.
Background and Objective: Radiomics is a field of medical study that aims to extract high-throughput quantitative features from medical images by data-characterisation algorithms[1]. These features, termed radiomic features, have the potential to decode the invisible disease characteristics which are useful for individualized treatment. Breast cancer is a known heterogeneity disease caused by variations in local microenvironmental conditions mainly governed by spatial and temporal changes in blood flow, which could be represented throuth different contrast enhancement patterns on DCE-MRI. Further the heterogeneity could be quantitatively assessed by radiomic methods based on pharmacokinetic modeling DCE-MRI[2-5].Now that breast tumors are heterogenieity, the metastatic axillary lymph node of breast tumor might be heterogenieity too. The aim of this study was to investigate the discriminative performance of pharmacokinetic quantitative parameters with histogram and texture features on dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) in predicting axillary lymph nodes metastasis in breast cancer.
Methods: 40 patients with histopathologically proven axillary lymph nodes metastasis of breast cancer were retrospectively reviewed from the database of our cancer center. DCE-MRI were performed using a 3.0T MRI scanner (Magnetom Skyra, Siemens Healthcare, Erlangen, Germany) using a 16-channel breast coil. A prototype CAIPIRINHA-Dixon-TWIST-VIBE sequence was used to obtain DCE date. Parameters were: TR, 5.64 ms; TE, 2.46/3.69 ms; FOV, 360 × 360 mm; slice thickness, 2.5 mm; no gap; matrix, 269 × 384; flip angle, 10°; temporal resolution, 11.8 s/phase, all 26 phases, and total acquisition time(TA), 5 min 12 s. Each metastatic axillary lymph node and selected paired contralateral negative axillary lymph node were segmented in 3D and voxel-wise analyzed in a dedicated post-processing software (Omni-Kinetics; GE Healthcare, Milwaukee, WI). The whole lesion’s pharmacokinetic quantitative parameters (Ktrans, Kep, and Ve) with their corresponding histogram features and 75 texture features of DCE-MRI were obtained. The parameters and features were compared using the wilcoxon signed-rank test between the positive group of metastatic axillary lymph nodes and the negative group of the selected contralateral negative axillary lymph nodes. The Bayes discriminant analysis in a leave–one-case-out-cross-validation was used to assessed the discriminative performance of these parameters and features between positive and negative group.
Results: 6 pharmacokinetic quantitative and their histogram parameters and 53 texture features of DCE-MRI and pharmacokinetic parameters data were statistically difference between the positive and negative group, the features list was provided in Table1, 2. In cross-validation, the accuracy of the classifier based on some features obtained 90% for the task of identifying the metastatic axillary lymph node in the study, yielding a considerably excellent ROC and AUC=0.987(Fig1) . The texture features of DCE-MRI had more prominent discriminative performance than the pharmacokinetic quantitative and their corresponding histogram parameters.
Conclusions: The radiomic features based on quantitatively pharmacokinetic DCE-MRI demonstrated promising application in discriminating between metastatic positive and negative axillary nodes.
1. Gillies RJ, PE Kinahan, and H Hricak (2016) Radiomics: Images Are More than Pictures, They Are Data. Radiology 278(2):563-577.
2. Gerlinger M, AJ Rowan, S Horswell, et al (2012) Intratumor heterogeneity and branched evolution revealed by multiregion sequencing. N Engl J Med 366(10):883-892.
3. Kim JH, ES Ko, Y Lim, et al (2017) Breast Cancer Heterogeneity: MR Imaging Texture Analysis and Survival Outcomes. Radiology 282(3):665-675.
4. Koren S and M Bentires-Alj (2015) Breast Tumor Heterogeneity: Source of Fitness, Hurdle for Therapy. Mol Cell 60(4):537-546.
5. Martelotto LG, CK Ng, S Piscuoglio, B Weigelt, and JS Reis-Filho (2014) Breast cancer intra-tumor heterogeneity. Breast Cancer Res 16(3).