Blind Source Separation of Signals from Dynamic Contrast Enhanced Breast Magnetic Resonance Images
Zhang B, Wang Y, Roys S, Gullapalli R
University of Maryland School of Medicine
Dynamic contrast enhanced breast magnetic resonance imaging is known to differentiating benign from malignant breast lesions with high sensitivity and specificity. However the large amount of images generated makes the clinical diagnosis labor intensive. We explore here the use of blind source separation (BSS), a statistical technique to analyze the data automatically. BSS identified successfully all the lesions that were clinically diagnosed. The first two independent components from any given data satisfied the clinical diagnosis. No lesions that were clinically diagnosed were missed by BSS. The successful application of BSS to dynamic contrast-enhanced breast MRI demonstrates that BSS is capable of extracting lesions dynamic response automatically.