Combining radiomic analysis of breast tumors with that of parenchyma may improve the accuracy of digital mammography for diagnosing breast cancer, according to a new study published online in Radiology.
Researchers from the University of Chicago’s Department of Radiology, led by Hui Li, PhD, found that using radiomics to add contralateral breast parenchyma analysis in the characterization of breast lesions at full-field digital mammography improved breast lesion classification compared to using radiomic tumor features alone.
For their study, Li and colleagues recruited 182 patients (average age of 56 years) with no personal history of cancer who underwent screening mammography exams at the University of Chicago Medical Center between June 2002 and July 2009. Among the patients were a total of 106 malignant and 76 benign breast lesions.
The researchers then performed automatic lesion segmentation and radiomic analysis for each breast tumor. Radiomic texture analysis also applied to designated normal regions in the contralateral breast parenchyma to assess mammographic parenchymal patterns.
Lastly, the classification performance of both features and results from a Bayesian artificial neural network were evaluated with the leave-one-patient-out method. Specifically, this method used the area under the receiver operating characteristic curve (AUC) to differentiate between malignant and benign lesions, according to the researchers.
The combined lesion and parenchyma classifier, when differentiating between malignant and benign mammographic lesions, performed better than using the lesion features alone. The results showed an AUC of .84 versus .79, respectively, according to Li et al.
On the combined feature set, six radiomic features (spiculation, margin sharpness, size, circularity from the tumor feature set and skewness and power law beta from the parenchymal feature set) were selected more than half the time during the feature selection process.
“Because breast parenchyma may reflect the biologic risk factors associated with breast cancer development, yielding the stromal parenchyma as an indicator of precancer, the combination of parenchyma and tumor characteristics may provide a stronger predictive model of malignancy,” the researchers concluded.
In the future, the team hopes to evaluate breast parenchyma symmetric to the tumor location on the contralateral breast, the entire contralateral breast parenchyma and parenchyma around the breast lesion to better understand the role of parenchyma stroma in assessing whether breast tumors are malignant.
They also plan to perform quantitative imaging analysis on both raw and processed images from various digital mammography unit manufacturers in order to “assess the robustness of the methods for parenchyma analysis and breast tumor characterization and to allow for more generalization and flexibility in different clinical settings,” the authors wrote.