| Abstract [eng] |
Objectives: To investigate the efficacy of magnetic resonance imaging radiomics in the detection of breast cancer linked to BRCA1 and BRCA2 gene mutations and to identify precise radiomic features that could develop new non-invasive breast cancer diagnostic approaches, oriented at genetics. Methods: A prospective analysis of MRI radiomic features and genetic data from 42 patients diagnosed with invasive ductal carcinoma was performed. The cohort was evenly split between patients with BRCA gene mutations and those without. The process involved the segmentation of tumors on magnetic resonance images and the extraction of radiomic features using “Olea medical” software. This was followed by a statistical analysis – Mutual information analysis – using Python programming tools. Mutual information values greater than 0 were considered significant for further analysis. Several radiomic features were identified as having potential diagnostic value and further analysed using violet plots. Results: The analysis identified “First-order 10th Percentile”, “Gray-level Run Length Matrix”, and “Gray-level Co-occurrence Matrix” as important radiomic features having the highest Mutual information score out of 111 other radiomic features. All these features were associated with the grey scale and image texture. Conclusions: This study validated the hypothesis that particular radiomic features hold diagnostic value in breast cancer BRCA1/2 diagnostics. Notably, features such as “First-Order 10th Percentile”, “Gray-level Run Length Matrix”, and “Gray-level Co-occurrence Matrix” were identified as having significant correlations with the BRCA mutation status, they could potentially serve as future diagnostic biomarkers for breast cancer. Future studies should aim to expand the patient cohort to enhance the reliability of these findings and consider integrating automated segmentation techniques to increase the precision and reproducibility of radiomic analyses. |
