Evaluation of lymphovascular invasion in invasive breast carcinoma using MRI texture analysis

The objective of this study was to investigate the preoperative predictability of the presence of lymphovascular invasion in patients with invasive breast carcinoma by examining the association between lymphovascular invasion and features obtained through texture analysis from both the tumoral and peritumoral regions on pre-biopsy magnetic resonance imaging. Between January 2017 and August 2024, a total of 117 adult female patients with histopathologically confirmed invasive breast carcinoma, who had undergone diagnostic dynamic contrast-enhanced breast magnetic resonance imaging prior to biopsy and had not received any neoadjuvant therapy, were included in the study. Based on the postoperative histopathological evaluation, the patients were divided into two groups: those with lymphovascular invasion (n=65) and those without lymphovascular invasion (n=52). For each lesion, data regarding tumor size, shape, the presence of intratumoral necrosis, histological subtype, nuclear grade, molecular subtype, and Ki-67 index were recorded. Additionally, for each patient, fibroglandular tissue density, background parenchymal enhancement, the presence of pathological axillary lymph nodes on magnetic resonance imaging, peritumoral and subcutaneous edema, and the presence of adjacent vessel sign in the peritumoral region were assessed. Three-dimensional segmentation was performed for each lesion on T2-weighted and contrast-enhanced T1-weighted series. Following all preprocessing steps, a total of 1,129 features were extracted from each sequence using the 3D Slicer software. Statistical analyses were conducted using the Student's t-test, the Mann–Whitney U test, and the Chi-square test. For features found to be statistically significant, receiver operating characteristic curve analysis was used to calculate the area under the curve, cutoff value, sensitivity, and specificity. In addition, multivariate logistic regression analysis was applied to evaluate clinical and radiological variables in relation to the presence of lymphovascular invasion. All statistical analyses were performed using the Statistical Package for the Social Sciences software, version 25.0. A p-value less than 0.05 was considered statistically significant. In the group with lymphovascular invasion, tumor size was significantly larger compared to the group without lymphovascular invasion (p=0.014); however, it was not identified as an independent predictive factor in the multivariate model. The presence of pathological axillary lymph nodes was found to be strongly associated with lymphovascular invasion (p<0.001) and was determined to be an independent predictor in the multivariate analysis (odds ratio = 27.92; p=0.001). Regarding the features extracted through texture analysis from the intratumoral and peritumoral regions on T2-weighted and contrast-enhanced T1-weighted magnetic resonance sequences, 182 features derived from the tumoral region and 56 features from the peritumoral region showed statistically significant differences between the two groups. Of the features showing significant differences between the two groups, 61 were obtained from contrast-enhanced T1-weighted sequences, while 177 were obtained from T2-weighted sequences. All features demonstrating statistically significant differences between the groups were derived from processed images using image filters such as Laplacian of Gaussian and wavelet transforms. In conclusion, texture analysis has gained importance as a non-invasive, effective, and relatively cost-efficient method for the preoperative assessment of the presence of lymphovascular invasion in patients with invasive breast carcinoma.