CT synthesis from MR images for orthopedic applications in the lower arm\n using a conditional generative adversarial network

Abstract

Purpose: To assess the feasibility of deep learning-based high resolution\nsynthetic CT generation from MRI scans of the lower arm for orthopedic\napplications.\n Methods: A conditional Generative Adversarial Network was trained to\nsynthesize CT images from multi-echo MR images. A training set of MRI and CT\nscans of 9 ex vivo lower arms was acquired and the CT images were registered to\nthe MRI images. Three-fold cross-validation was applied to generate independent\nresults for the entire dataset. The synthetic CT images were quantitatively\nevaluated with the mean absolute error metric, and Dice similarity and surface\nto surface distance on cortical bone segmentations.\n Results: The mean absolute error was 63.5 HU on the overall tissue volume and\n144.2 HU on the cortical bone. The mean Dice similarity of the cortical bone\nsegmentations was 0.86. The average surface to surface distance between bone on\nreal and synthetic CT was 0.48 mm. Qualitatively, the synthetic CT images\ncorresponded well with the real CT scans and partially maintained high\nresolution structures in the trabecular bone. The bone segmentations on\nsynthetic CT images showed some false positives on tendons, but the general\nshape of the bone was accurately reconstructed.\n Conclusions: This study demonstrates that high quality synthetic CT can be\ngenerated from MRI scans of the lower arm. The good correspondence of the bone\nsegmentations demonstrates that synthetic CT could be competitive with real CT\nin applications that depend on such segmentations, such as planning of\northopedic surgery and 3D printing.\n

Ähnliche Arbeiten