Method and system for material decomposition in dual- or multiple-energy x-ray based imaging

Inventors

Peng, Yu

Assignees

Curvebeam AI Ltd

Abstract

A method and system for generating material decomposition images from plural-energy x-ray based imaging, the method comprising: modelling spatial relationships and spectral relationships among the plurality of images by learning features from the plurality of images in combination and one or more of the plurality of images individually with a deep learning neural network; generating one or more basis material images employing the spatial relationships and the spectral relationships; and generating one or more material specific or material decomposition images from the basis material images. The neural network has an encoder-decoder structure and includes a plurality of encoder branches; each of one or more of the plurality of encoder branches encodes two or more images of the plurality of images in combination; and each of one or more of the plurality of encoder branches encodes a respective individual image of the plurality of images.

Core Innovation

The invention generates one or more basis material images and one or more material specific or material decomposition images from a plurality of images obtained with plural-energy x-ray based imaging, where the plurality of images correspond to respective energies. The invention models spatial relationships and spectral relationships among the plurality of images by learning features from the plurality of images in combination and from one or more of the plurality of images individually using a deep learning neural network.

The neural network uses an encoder-decoder structure and includes a plurality of encoder branches. Each encoder branch encodes two or more images of the plurality of images in combination, while one or more encoder branches encode a respective individual image of the plurality of images. This branch structure supports learning from both combined image information and individual energy image information within the same model.

From the generated basis material images, the invention derives one or more material specific or material decomposition images. The described material decomposition includes calcium/water/fat/iodine and further examples including bone marrow and knee cartilage, as well as tumor and iodine contrast, and muscle and fat decomposition. The invention further includes generating metal artefact and beam hardening reduced images to improve image quality.

Claims Coverage

The independent claims cover a deep-learning method and a corresponding system, both using an encoder-decoder neural network with multiple encoder branches that learn spatial and spectral relationships across plural-energy x-ray images to produce basis material images and material decomposition images. The core claim set centers on three inventive features: multi-branch encoder learning from combinations and individual images, generation of basis material images using spatial/spectral relationships, and derivation of material specific/material decomposition images from the basis material images.

Across the independent claims, the coverage is directed to generating material decomposition outputs from plural-energy x-ray based image inputs by learning spatial and spectral relationships using an encoder-decoder deep learning neural network with multiple encoder branches. The architecture is defined such that encoder branches encode both combinations of two or more energy images and respective individual energy images, enabling generation of basis material images and subsequent derivation of material-specific/material decomposition images.

Stated Advantages

Documented Applications