(Sunday, 7/29/2018) 4:30 PM - 5:00 PM
Room: Exhibit Hall | Forum 9
Purpose: Preclinical imaging is commonly used in longitudinal studies to test the efficacy of novel treatments. Contouring of MR images is a time-consuming process when performed manually. The purpose of this study is to investigate the accuracy of a semi-automated graphical user interface (GUI) that allows the user to measure tumor volume using a combination of a threshold-based region-growing algorithm and edge analysis.
Methods: Images of mice (n=16) with flank tumors (n = 30 tumors) with a large range in size (68 – 2600 mm³) were analyzed using a GUI that implements two methods to contour. The first is a threshold-based region-growing algorithm for which the user chooses a seed point within the tumor and a background point outside the tumor; the threshold is set as a value intermediate to the average voxel values in these two regions by a preset fraction. Canny edge detection is used to define the edges of the tumor, with an adjustable smoothing parameter (sigma in this study was ~0.2 pixels); manual contouring may also be used to additionally adjust the contour. The performance of this tool was tested by comparing measured volumes to those measured by manually contouring using a separate, independently validated in-house program.
Results: On average, the volumes measured with the GUI differed from those measured manually by 99±107 mm³ (1 σ), indicating that delineation of the tumor boundaries may be aided using this GUI. At less sharp boundaries, determination of the edge is sensitive to the parameters of the GUI (e.g., choice of the smoothing parameter).
Conclusion: The study indicates that these contouring methods allow for an initial, objective delineation of the tumor, which may be then fine-tuned by the user. Further studies are warranted to compare the accuracy of other semi-automated contouring methods applied to these studies.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
