The purpose of our study was to determine the frequency of BI-RADS 3 lesions in breast MR imaging in a clinical patient population and their frequency of malignancy in follow-up breast MR imaging. In 44/698 (6.3%) patients with breast MR imaging, 56 lesions were categorized to BI-RADS 3. These lesions were all not palpable and not detectable at conventional mammography or ultrasound. In follow-up, lesions were score in complete resolved (CRL), partial resolved (PRL), stable lesions (SL), and progressive lesions (PL). Initial signal enhancement of lesions was coded by color intensity (bright for high, medium for medium, dark for low), the postinitial signal enhancement by color hue (blue for increase, green for plateau, red for wash-out). In first follow-up breast MR imaging 23/56 (41%) lesions were PRL, 14/56 (25%) lesions were CRL, 14/56 (25%) lesions remained SL. In one of five PL lesions, histopathology revealed a malignant tumor. In initial breast MR imaging, CRL showed significant fewer high pixels (p = 0.002), medium pixels (p = 0.006) significant more low pixels (p = 0.005) and significant more increase pixels (p = 0.037) than PRL. In a clinical patient population the frequency of malignancy of BI-RADS 3 lesions in breast MR imaging and their frequency of malignancy are similar to that in conventional mammography. In initial breast MR imaging, complete resolved lesions showed less suspicious contrast kinetics than other lesions. In follow-up, the increase of lesion size should warrant histopathological diagnosis.