Program Information
Time Evolution of Texture Parameters of Subtracted Images Obtained by Contrast-Enhanced Digital Mammography (CEDM)
M-J Mateos1*, A Gastelum2 , J Marquez2 , M-E Brandan1 , (1) Instituto de Fisica, Universidad Nacional Autonom de Mexico, Mexico, Distrito Federal, (2) Centro de Ciencias Aplicadas y Desarrollo Tecnologico Universidad Nacional Autonoma de Mexico, Mexico, Distrito Federal
Presentations
TU-CD-207-3 (Tuesday, July 14, 2015) 10:15 AM - 12:15 PM Room: 207
Purpose:To evaluate the time evolution of texture parameters, based on the gray level co-occurrence matrix (GLCM), in subtracted images of 17 patients (10 malignant and 7 benign) subjected to contrast-enhanced digital mammography (CEDM). The goal is to determine the sensitivity of texture to iodine uptake at the lesion, and its correlation (or lack of) with mean-pixel-value (MPV).
Methods:Acquisition of clinical images followed a single-energy CEDM protocol using Rh/Rh/48 kV plus external 0.5 cm Al from a Senographe DS unit. Prior to the iodine-based contrast medium (CM) administration a mask image was acquired; four CM images were obtained 1, 2, 3, and 5 minutes after CM injection. Temporal series were obtained by logarithmic subtraction of registered CM minus mask images.
Regions of interest (ROI) for the lesion were drawn by a radiologist and the texture was analyzed. GLCM was evaluated at a 3 pixel distance, 0° angle, and 64 gray-levels. Pixels identified as registration errors were excluded from the computation. 17 texture parameters were chosen, classified according to similarity into 7 groups, and analyzed.
Results:In all cases the texture parameters within a group have similar dynamic behavior. Two texture groups (associated to cluster and sum mean) show a strong correlation with MPV; their average correlation coefficient (ACC) is r^2=0.90. Other two groups (contrast, homogeneity) remain constant with time, that is, a low-sensitivity to CM uptake. Three groups (regularity, lacunarity and diagonal moment) are sensitive to CM uptake but less correlated with MPV; their ACC is r^2=0.78.
Conclusion:This analysis has shown that, at least groups associated to regularity, lacunarity and diagonal moment offer dynamical information additional to the mean pixel value due to the presence of CM at the lesion. The next step will be the analysis in terms of the lesion pathology. Authors thank PAPIIT-IN105813 for support.
Funding Support, Disclosures, and Conflict of Interest: Consejo Nacional de Ciencia Y Tecnologia, PAPIIT-IN105813
Contact Email: