摘要
应用图像分析手段，对乳腺癌、乳腺腺病和正常乳腺组织进行了７项核形态参数测定。除形状因子外，其余６项均显示正常乳腺与Ⅰ级腺病间无差异（Ｐ＞０．０５）。Ⅰ、Ⅱ级腺病及乳癌间均存在显著性差异（Ｐ＜０．０１）。经逐步判别分析对７项核形态参数进行综合评定，选出三项较优指标，建立判别函数，对Ⅱ级腺病及乳癌的判定回代符合率分别为９３．５５％和８７．８％。结果表明：利用图像分析，结合有效的统计学方法，有可能对具有癌变可能的Ⅱ级腺病及癌的判定，提供客观可靠的手段。
Seven morphometric parameters of cell nuclei were measured by means of an automatic image analysis system (Quantimet 970) in 6 cases with normal breast tissue, 49 with adenoses and 41 with breast carcinoma. Apart from the shape factor, the other 6 parameters (area, mean diameter, perimeter, DNA ploid, mean optical density and integrated optical density) were significantlydifferent among adenosis Ⅰ,Ⅱand breast carcinoma. A preliminary data bank for objective and reproducible grading of the breast lesions by stepwise discriminant analysis of the 7 parameters was made.The discriminant analysis of adenosis Ⅱgroup and breast carcinoma group showed a discriminanting power of 93.55% and 87.8%, respectively, as compared to the original diagnoses. This result indicates that combining image analysis technique with valid statistical method(stepwise discriminant analysis)may provide an objective,reliable means for determining the diagnosis of carcinoma and adenosis Ⅱ which may transform into malignancy.
出处
《中国体视学与图像分析》
1998年第1期34-36,共3页
Chinese Journal of Stereology and Image Analysis
