Canon Medical Systems Corporation

University of Chicago

Yu-Bing Chang

Joseph Manak

John Baumgart

Xiaochuan Pan

Dan Xia

Zheng Zhang

Yi Hu

Haruki Iwai

James Begelman

Jingwu Yao

Rui Hua

A system and method is provided for controlling an acquisition system to obtain, based on a currently acquired set of projection data, a next set of projection data that is to be acquired using a set of learned imaging conditions. The set of learned imaging conditions are, at least in part, based on training data including projection data and dose information. In one embodiment, the initial dose and angle for obtaining the initial set of projection data are chosen at random. In another embodiment, the initial dose and angle for obtaining the initial set of projection data are chosen based on a test scan and a trained neural network for outputting the initial dose and angle using a loss function.

Apparatus, method, and non-transitory computer-readable storage medium for generating a task-aware dynamic sparse-view scanning procedure

A general workflow for deep learning based image restoration in X-ray and fluoroscopy/fluorography is disclosed. Higher quality images and lower quality images are generated as training data. This training data can further be categorized by anatomical structure. This training data can be used to train a learned model, such as a neural network or deep-learning neural network. Once trained, the learned model can be used for real-time inferencing. The inferencing can be more further improved by employing a variety of techniques, including pruning the learned model, reducing the precision of the learned mode, utilizing multiple image restoration processors, or dividing a full size image into snippets.

X-ray imaging restoration using deep learning algorithms

An apparatus and method are provided for computed tomography (CT) imaging to reduce artifacts due to objects outside the field of view (FOV) of a reconstructed image. The artifacts are suppressed by using an iterative reconstruction method to minimize a cost function that includes a de-emphasis operator. The de-emphasis operator operates in the data domain, and minimizes the contributions of data inconsistencies arising from attenuation due to objects outside the FOV. This can be achieved by penalizing images that manifest indicia of artifacts due to outside objects especially those outside objects have high-attenuation densities and minimizing components of the data inconsistency likely attributable to the outside object.

Method and apparatus to reduce artifacts in a computed-tomography (CT) image by iterative reconstruction (IR) using a cost function with a de-emphasis operator

A method and apparatus is provided to generate a multiresolution image having at least two regions with different pixel pitches. The multiresolution image is reconstructed using projection data having various pixel pitches corresponding to the pixel pitches of the multiresolution image. By using a higher resolution inside regions of interest (ROIs) in both the image and projection domains and lower resolution outside the ROIs, fast image reconstruction can be performed while avoiding truncation artifacts, which result imaging is limited to an ROI excluding attenuation regions. Further, those regions of greater clinical relevance and greater structural variance within the reconstructed images can be selected to be within the ROIs to improve the clinical benefit of the multiresolution image. The multiresolution image can be reconstructed using an iterative reconstruction method in which the high- and low-resolution regions are uniquely evaluated.

Multiresolution iterative reconstruction for region of interest imaging in X-ray cone-beam computed tomography

Growing community of inventors

Albany, NY, United States of America

Yu-Bing Chang