Automated pixel-by-pixel mapping of dynamic contrast-enhanced MR images for evaluation of osteosarcoma response to chemotherapy: preliminary results.

An automated technique for pixel-by-pixel computer mapping of tumor necrosis was developed to improve the accuracy and applicability of dynamic contrast agent-enhanced magnetic resonance (MR) imaging in assessing the response of osteosarcoma to preoperative chemotherapy. The technique was evaluated in nine osteosarcoma patients who underwent MR imaging at diagnosis (n = 7) and after preoperative chemotherapy (n = 9). Sequential FLASH (fast low-angle shot) images of the tumor were obtained in one plane every 15 seconds before and for 3 minutes after gadopentetate dimeglumine injection. A region of interest was selected that included the entire tumor area. Slopes representing percent increase in signal intensity over baseline values per minute were calculated automatically for each pixel and displayed as gray-scale maps of the tumor. Matching histologic maps of each tumor were obtained. Visual region-by-region comparison of the pixel and histologic maps showed a high degree of correlation and the ability of the MR imaging technique to help identify small foci of residual viable tumor.