Fractal analysis of mammographic lesions: a prospective, blinded trial

Mammography has become the mainstay of breast cancer screening. However, widespread mammography has led to an increase of the number of breast biopsies done for benign disease. Therefore, a method to better discriminate benign from malignant lesions is needed. Fractal analysis is a mathematical method which can quantify complex shapes. It has been previously shown retrospectively that the composite fractal dimensions, D, of malignant mammographic masses is higher than for benign lesions. A prospective study of 75 patients who were recommended to undergo needle localized breast biopsy by independent radiologists had the composite D calculated. Fractal analysis was done without knowledge of the biopsy results. The mean composite D of malignant lesions was higher than benign lesions, 2.545+/-0.067 vs. 1.936+/-0.144 (p=0.00004). Calculation of a receiver-operating characteristic curve showed that a cutoff value of 2.067 had a 100% sensitivity and 63 % specificity (i.e., false positive rate of 37%). Mean D for fibroadenomas was 2.087+/-0.054, fibrocystic disease was 1.877+/-0.167, DCIS was 2.261+/-0.069, and invasive cancer was 2.634+/-0.039 (1-way ANOVA, p=0.00007). These data imply that fractal analysis may be beneficial in discriminating between benign and malignant lesions. However, further study in a larger number of patients with a variety of lesions is needed.