Back to results
US 12,718,449 B2Granted
Metal artifact reduction algorithm for CT-guided interventional procedures
AssigneeMayo Foundation for Medical Education and Research
InventorsChristopher P. Favazza, Andrea Ferrero, Liqiang Ren
Life of this patent
Estimating the term…
Abstract
Metal artifacts are reduced in x-ray computed tomography (“CT”) images using a suitably trained neural network, such as a convolutional neural network (“CNN”). Virtual metal DATA objects are inserted to either the raw projection data or CT image data (e.g., from pre-procedural CT scans) to generate sets of matching artifact-corrupted and artifact-uncorrupted images, and a CNN, or other neural network, is trained to separate the contribution to each image pixel due to patient anatomy, metal object, or metal object-induced artifact. The contributions from metal object-induced artifacts can then be removed to generate a final, artifact-reduced image.
Claims22
A method for generating a metal artifact reduced image, the method comprising:
- (a) accessing computed tomography (CT) image data with a computer system, the CT image data being acquired from a subject using a CT imaging system and depicting anatomy of the subject, a metal object present in the subject, and artifacts caused by the metal object;
- (b) accessing a trained neural network with the computer system, the trained neural network being trained on training data in order to separate anatomy and metal objects from artifacts caused by metal objects, wherein the neural network is trained on training data comprising artifact-corrupted image data comprising CT images to which virtual metal object data have been inserted in a projection domain, and artifact-uncorrupted image data comprising the CT images to which the virtual metal object data have been inserted in an image domain;
- and
- (c) inputting the CT image data to the trained neural network, generating output as artifact-reduced image data depicting the anatomy of the subject and the metal object, wherein the artifacts caused by the metal object are reduced.
1 more element
A method for generating an image that depicts a virtual metal object and corresponding artifacts, the method comprising:
- (a) accessing computed tomography (CT) projection data with a computer system, the CT projection data being acquired from a subject using a CT imaging system;
- (b) accessing virtual metal object projection data with the computer system, the virtual metal object projection data comprising a forward projection of a virtual model of a metal object;
- (c) generating combined projection data by combining the CT projection data and the virtual metal object projection data while using a beam hardening model to account for beam hardening effects attributable to the virtual model of the metal object in the combined projection data and a noise model to account for noise effects attributable to the virtual model of the metal object in the combined projection data, wherein the noise model inserts both quantum noise and electronic noise;
- and
- (d) reconstructing an image from the combined projection data, wherein the image depicts anatomy of the subject, the virtual metal object, and artifacts corresponding to the virtual metal object.
2 more elements
A method for generating a metal artifact reduced image, the method comprising:
- (a) accessing computed tomography (CT) image data with a computer system, the CT image data being acquired from a subject using a CT imaging system and depicting anatomy of the subject, a metal object present in the subject, and artifacts caused by the metal object;
- (b) accessing a first trained neural network with the computer system, the first trained neural network being trained on training data in order to separate anatomy from metal objects and artifacts caused by metal objects;
- (c) inputting the CT image data to the first trained neural network, generating output as anatomy image data depicting the anatomy of the subject and artifact containing metal object image data depicting the metal object and artifacts caused by the metal object;
- (d) accessing a second trained neural network with the computer system, the second trained neural network being trained on training data in order to separate metal objects from artifacts caused by metal objects;
- (e) inputting the artifact containing metal object image data to the second trained neural network, generating output as metal object image data depicting the metal object and artifact image data depicting the artifacts caused by the metal object;
- and
- (f) generating a combined image by combining the anatomy image data and the metal object image data with the computer system, wherein the combined image depicts the anatomy of the subject and the metal object without the artifacts caused by the metal object.
4 more elements
A method for generating a metal artifact reduced image, the method comprising:
- (a) accessing computed tomography (CT) image data with a computer system, the CT image data being acquired from a subject using a CT imaging system and depicting anatomy of the subject, a metal object present in the subject, and artifacts caused by the metal object;
- (b) accessing a trained neural network with the computer system, the trained neural network being trained on training data in order to separate anatomy and metal objects from artifacts caused by metal objects, wherein the neural network is trained on training data comprising artifact-corrupted image data comprising CT images to which metal object mask data have been inserted in an image domain, and artifact-uncorrupted image data comprising the CT images to which the metal object mask data have been inserted in the image domain;
- and
- (c) inputting the CT image data to the trained neural network, generating output as artifact-reduced image data depicting the anatomy of the subject and the metal object, wherein the artifacts caused by the metal object are reduced.
1 more element
Dependencies are inferred from the claim text.