Automatic Placement of Regions of Interest using Distance transform to Measure Spatial Resolution on the Clinical Computed Tomography Images : A Pilot Study

Abstract

We propose a new algorithm called distance transform region of interest (DT-ROI) to deal with the irregular patient's surface. The ROIs can be placed orthogonally along the patient’s surface to get spatial resolution. The algorithm was developed using several image processing techniques. The original image was first segmented to obtain a segmented image. The segmented image was eroded and dilated to obtain an eroded and dilated image. Both the eroded and dilated images were edge detected to obtain the edge images of the eroded and dilated image. The edge images were distance transformed to obtain the closest pixel coordinate. Finally, ROIs were placed based on the coordinates obtained before. The DT-ROI was then assessed qualitatively by comparison with the ROI placement from the standard radial ROI (SR-ROI) on a Polymethyl methacrylate (PMMA) phantom, an anthropomorphic phantom, and the patient’s computed tomography images. The algorithm resulted in orthogonalized ROIs, both along the irregular object and the circular object. The ROI comparison between DT-ROI and SR-ROI shows a little difference in terms of orthogonality on PMMA phantom. Meanwhile, on the anthropomorphic phantom and the patient’s CT image, the DT-ROI produced a lot more orthogonal ROIs than the SR-ROI. Several ROIs of the DT-ROI have decreased orthogonality at certain sections, which can be observed in both phantom and patient images. However, theoretically, a slight decrease in orthogonality will not affect the modulation transfer function (MTF) measurement significantly. The DT-ROI algorithm has been successfully developed based on distance transformation and performed as the design. The algorithm can automatically place ROIs along the patient’s irregular surface better than the SR-ROI algorithm. However, not all ROIs placed from DT-ROI are well-orthogonalized. DT-ROI still needs to be improved before it is used to measure MTF to obtain a more optimal measurement.