Search

Objective To investigate the MRI feature of continuous enhancement from index lesion to nipple-areola complex (NAC) for preoperative prediction of occult NAC involvement in patients with breast cancer. Methods Between January 2017 and September 2018. 242 patients with breast cancer were enrolled in this prospective study. The patients underwent breast MRI before radical mastectomy. They were divided into a NAC-positive and NAC-negative groups based on the pathologic results regarding NAC involvement. Preoperative MRI features such as lesion type, location, tumor size, enhancement pattern, multicentricity, tumor—to—NAC continuous enhancement, tumor—to—NAC distance and nipple retraction were compared between the two groups using Pearson X2 test. The diagnostic performance of tumor-to-NAC continuous enhancement for occult NAC involvement was evaluated. Results Of the 242 breast cancer patients, 47 were NAC-positive and 195 were NAC-negative. The tumor-to-NAC continuous enhancement was significantly different between NAC-positive and NAC-negative groups (P<0.001). It was present in 34 (72.3%) NAC-positive patients and in 11 (5.6%) NAC-negative patients. The other 7 MRI features were not significantly different between the two groups (P=0.110-0.696). Tumor-to-NAC continuous enhancement was 72.3% sensitive, 94.4% specific, 90.1% accurate with 75.6% positive predictive value and 93.4% negative predictive value for the diagnosis of occult NAC involvement. Conclusion Tumor-to-NAC continuous enhancement is an effective MRI feature for the preoperative prediction of occult NAC involvement in breast cancer. [ABSTRACT FROM AUTHOR]

Ｔｏ ｐｒｅｄｉｃｔ ｐｒｅｏｐｅｒａｔｉｖｅ ｌｙｍｐｈｏｖａｓｃｕｌａｒ ｉｎｖａｓｉｏｎ(ＬＶＩ)ｓｔａｔｕｓ ｏｆ ｂｒｅａｓｔ ｃａｎｃｅｒ ｐａｔｉｅｎｔｓꎬ ａ ｃｏｍｐｕｔｅｒ￣ａｉｄｅｄ ａｎａｌｙｓｉｓ ｍｅｔｈｏｄ ｗｈｉｃｈ ｃｏｍｂｉｎｅｓ ａｎ ａｃｔｉｖｅ ｃｏｎｔｏｕｒ ｍｏｄｅｌ ａｎｄ ｒａｄｉｏｍｉｃｓ ｉｓ ｐｒｏｐｏｓｅｄ. Ｆｉｒｓｔꎬ ｔｈｅ ｉｍａｇｅ ｆｒｏｍ ｄｙｎａｍｉｃ ｃｏｎｔｒａｓｔ ｅｎｈａｎｃｅｄ ｍａｇｎｅｔｉｃ ｒｅｓｏｎａｎｃｅ ｉｍａｇｉｎｇ(ＤＣＥ￣ ＭＲＩ)ｏｆ ｂｒｅａｓｔ ｃａｎｃｅｒ ｉｓ ｓｅｇｍｅｎｔｅｄ ｂｙ ａｎ ａｃｔｉｖｅ ｃｏｎｔｏｕｒ ｍｏｄｅｌ(ＡＣＭ)ｂａｓｅｄ ｏｎ ｐｏｓｔ ｐｒｏｂａｂｉｌｉｔｙ ａｎｄ ｆｕｚｚｙ ｖｅｌｏｃｉｔｙ ｆｕｎｃｔｉｏｎ. Ｂｙ ｃｏｎｓｔｒｕｃｔｉｎｇ ｔｈｅ ｒｅｇｉｏｎ ｔｅｒｍ ｏｆ ｔｈｅ ａｃｔｉｖｅ ｃｏｎｔｏｕｒ ｍｏｄｅｌ ｂａｓｅｄ ｏｎ ｐｏｓｔ ｐｒｏｂａｂｉｌｉｔｙ ｉｎ ｔｈｅ ｗａｖｅｌｅｔ ｄｏｍａｉｎ ａｎｄ ｔｈｅ ｅｄｇｅ ｓｔｏｐ ｔｅｒｍ ｏｆ ｔｈｅ ａｃｔｉｖｅ ｃｏｎｔｏｕｒ ｍｏｄｅｌ ｂｙ ｕｓｉｎｇ ｔｈｅ ｆｕｚｚｙ ｖｅｌｏｃｉｔｙ ｆｕｎｃｔｉｏｎꎬ ｔｈｅ ａｃｃｕｒａｃｙ ｏｆ ｂｒｅａｓｔ ｃａｎｃｅｒ ｌｅｓｉｏｎ ｓｅｇｍｅｎｔａｔｉｏｎ ｃａｎ ｂｅ ｉｍｐｒｏｖｅｄ. Ｓｅｃｏｎｄꎬ ｔｈｅ ｉｍａｇｅ ｆｅａｔｕｒｅｓ ｓｕｃｈ ａｓ ｍｏｒｐｈｏｌｏｇｙꎬ ｇｒａｙｓｃａｌｅ ａｎｄ ｔｅｘｔｕｒｅ ａｒｅ ｅｘｔｒａｃｔｅｄ. Ｆｉｎａｌｌｙꎬ ａ ｍｏｄｅｌ ｆｏｒ ｐｒｅｄｉｃｔｉｎｇ ＬＶＩ ｓｔａｔｕｓ ｉｓ ｄｅｖｅｌｏｐｅｄ ｂｙ ｔｈｅ ｒａｎｄｏｍ ｆｏｒｅｓｔ ｃｌａｓｓｉｆｉｅｒ ａｎｄ ｉｔｓ ｐｒｅｄｉｃｔｉｖｅ ａｂｉｌｉｔｙ ｉｓ ｖｅｒｉｆｉｅｄ ｂｙ ｅｘｐｅｒｉｍｅｎｔａｌ ｒｅｓｕｌｔｓ. [ABSTRACT FROM AUTHOR]