Comparing the linear structure features of normal breast and architectural distortion tissue based on Linear Structure Identification (LSI)

Abstract

     Architectural distortion (AD) indicates the occurrence of early breast cancer. In mammographic image, the line correspond to the linear structure of normal breast tissue differ from the distorted tissue. For this reason, this feature may benefit for seeking the abnormality. In a previous study, we found the computerized Linear Structure Identification method (LSI) has been successfully detected the linear structure in mammographic image. The purpose of this study is to compare the linear structure feature of normal breast and AD tissue based on linear structure identification (LSI). The method includes developing computer code based on the LSI, testing the robustness of the computer code on the simulation images, detecting linear structures and statistical analysis. The areas for examination consist of AD, fibroglandular and fatty tissue. The linear structure features extracted from the LSI are minimum standard deviation, variation of standard deviation and distribution of angle of minimum standard deviation. The results show that the set of pixels belonging to linear structures are more in AD than fibroglandular and fatty tissue. The mean of minimum standard deviation of AD, fibroglandular and fatty tissue are 0.4507, 0.4327 and 0.2583, respectively. Comparing results between AD and fibroglandular tissue show no significant difference between two areas (p =0.219) but the comparative results between AD and fatty tissue show significant difference (p<0.0001). The mean of variances are 0.2973, 0.2535 and 0.1216. Comparing results between AD and fibroglandular tissue show significant difference (p=0.022) similarly to the comparative results between AD and fatty tissue (p <0.0001). The distribution of angle with minimum standard deviation of all three areas show significant difference (df =18, p<0.05). In conclusion, LSI is successful in detecting linear structure in AD tissue. The linear structure features from the LSI are difference between AD and normal breast tissue. In the comparison between AD and fibroglandular tissue, linear structure features show no difference in the mean of minimum standard deviation but difference in the mean of variance and distribution of angle with minimum standard deviation. Likewise, linear structure features between AD and fatty tissue are difference all three values.