Kurtosis Diffusion-Weighted Imaging May Reduce False Positive Mammographies

Radiomics feature model reduced false positives, with benefit seen for BI-RADS 4a, 4b lesions
Radiomics feature model reduced false positives, with benefit seen for BI-RADS 4a, 4b lesions

HealthDay News — A radiomics model based on kurtosis diffusion-weighted imaging reduces false positives in women with suspicious findings on mammography, according to a study published online February 20 in Radiology.

Sebastian Bickelhaupt, MD, from the University Hospital Erlangen in Germany, and colleagues examined 222 women at two independent study sites (site 1: training set of 95 patients with 61 malignant and 34 benign lesions; site 2: independent test set of 127 patients with 61 malignant and 66 benign lesions). All participants presented with a finding suspicious for cancer at X-ray mammography (Breast Imaging Reporting and Data System [BI-RADS]) and an indication for biopsy. Diffusion-weighted magnetic resonance imaging was performed before biopsy. Lesions were segmented and voxel-based kurtosis fitting was performed, which was adapted to account for fat signal contamination. 

Related Articles

At the predefined sensitivity of greater than 98.0%, the researchers found that the radiomics feature model reduced false-positive results from 66 to 20 in the independent test set (specificity, 70.0%); BI-RADS 4a and 4b lesions benefited from the analysis (74.0 and 60.0%, respectively), while no added benefit was seen for BI-RADS 5 lesions. Compared with the median apparent diffusion coefficient and the apparent kurtosis coefficient alone, the model significantly improved specificity.

"A radiomics model based on kurtosis diffusion-weighted imaging performed by using magnetic resonance imaging machines from different vendors allowed for reliable differentiation between malignant and benign breast lesions in both a training and an independent test data set," the authors write.

Several authors disclosed financial ties to the medical device industry.

Abstract/Full Text