Contribution of tissue analysis to radiological findings in breast malignant tumors

Purpose The aim of this study was to examine the magnetic resonance (MR) images of patients with a diagnosis of breast tumor by texture analysis and to investigate its place in differential diagnosis. In our study, it is aimed to investigate the correlation between the parameters to be obtained by tissue analysis from MR images in breast tumors and immunohistochemistry. It is predicted that quantitative data obtained by tissue analysis from MR images may correlate with immunohistochemistry, and enable non-invasive evaluation of tumor features and behaviors. Material and Method Between January 2016 and December 2019, 115 patients who had undergone MR imaging with a pre-diagnosis of gastric tumor were included in the study. Patient records and images were evaluated retrospectively. Patients with missing file or pathology records, who had undergone surgical or medical treatment prior to imaging, and those whose MR images were not suitable for analysis were excluded. Images of 98 patients included in the study were transferred to the open source LIFEx software and pixel size adjustment, gray level discretization and gray level normalization procedures were applied. Then, ROI was determined from the axial slice where the lesion was the widest and a total of 38 texture features were calculated for each patient. Firstly, patients are divided into four groups. Since the distribution of the patients was not normal, first Kruskal Wallis test is performed. After this process, complementary comparison techniques are used in order to determine which groups are significantly different. Subsequently, we compared each breast tumor subtype with others using the Mann-Whitney U-test. We also compared breast tumors with low or high Ki67 proliferative index among themselves. Results Although; in our study we could not find a parameter that successfully differentiated all tumor subtypes, pairwise comparisons yielded better results. In the statistical analysis, GLRLM_HGRE, GLRLM_GLNU, GLRLM_RLNU, GLZLM_HGZE, GLZLM_GLNU, GLZLM_ZLNU values revealed significant differences among Luminal A and HER2+ groups (P < .05). In our study, in the comparative analysis of Luminal A and HER2 + groups with other breast tumors,15 and 8 out of 38 feature values are significantly different, respectively. Lastly, 10 out of 38 tissue features are significantly different between tumors with low and high Ki67 proliferative index. (P <.05) Conclusion Our study showed that quantitative radiomics imaging features of breast tumour extracted from MRI are associated with breast cancer subtypes and Ki67 expression. Future larger studies are needed in order to further evaluate the findings.