Methods, apparatus and systems for deep learning based image reconstruction are disclosed herein. An example at least one computer-readable storage medium includes instructions that, when executed, cause at least one processor to at least: obtain a plurality of two-dimensional (2D) tomosynthesis projection images of an organ by rotating an x-ray emitter to a plurality of orientations relative to the organ and emitting a first level of x-ray energization from the emitter for each projection image of the plurality of 2D tomosynthesis projection images; reconstruct a three-dimensional (3D) volume of the organ from the plurality of 2D tomosynthesis projection images; obtain an x-ray image of the organ with a second level of x-ray energization; generate a synthetic 2D image generation algorithm from the reconstructed 3D volume based on a similarity metric between the synthetic 2D image and the x-ray image; and deploy a model instantiating the synthetic 2D image generation algorithm.

In the present embodiments, a statement related to an image point or an image region in a reconstructed x-ray image is made in relation to the reliability of the reconstructed grayscale value for the image points of a 2D/3D x-ray image. A confidence level is formed for the grayscale value from a first number of the available x-ray images in relation to a second number of required x-ray images for a complete reconstruction of the respective grayscale value of the 2D/3D x-ray image to be imaged.

Generally, there is provided a hybrid image reconstruction system. The hybrid image reconstruction system includes a deep learning stage and a compressed sensing stage. The deep learning stage is configured to receive an input data set that includes measured tomographic data and to produce a deep learning stage output. The deep learning stage includes a mapping circuitry, and at least one artificial neural network. The mapping circuitry is configured to map image domain data to a tomographic data domain. The compressed sensing stage is configured to receive the deep learning stage output and to provide refined image data as output.

A method of investigating a specimen using a tomographic imaging apparatus comprising: A specimen holder, for holding the specimen; A source, for producing a beam of radiation that can be directed at the specimen; A detector, for detecting a flux of radiation transmitted through the specimen from the source; A stage apparatus, for producing relative motion of the source with respect to the specimen, so as to allow the source and detector to image the specimen along a series of different viewing axes; A processing apparatus, for assembling output from the detector into a tomographic image of at least part of the specimen,
which method comprises the following steps: Considering a virtual reference surface that surrounds the specimen and is substantially centered thereon; Considering an incoming point of intersection of each of said viewing axes with this reference surface, thereby generating a set of such intersection points corresponding to said series of viewing axes; Choosing discrete viewing axes in said series so as to cause said set to comprise a two-dimensional lattice of points located areally on said reference surface in a substantially uniform distribution.

A first imaging unit obtains a first radiation image, which is imaged under first imaging conditions. A second imaging unit obtains a plurality of projection images by tomosynthesis imaging under second imaging conditions. A reconstructing unit reconstructs a plurality of projection images to generate a plurality of tomographic images of cross sectional planes of a subject. An image synthesizing unit generates a second radiation image employing the plurality of tomographic images. A subtraction processing unit administers a subtraction process on the first and second radiation images, to generate a subtraction image.

A method and a related apparatus for performing a tomographic examination of an object (2) which advances through an examination area (6), wherein the examination area (6) is irradiated with x-rays transversally to a motion trajectory of the object (2) and the residual intensity of the x-rays which have crossed the object (2) is repeatedly detected to obtain, for each detection, an electronic two-dimensional pixel map, the two-dimensional maps thus obtained being processed by a computer to obtain a three-dimensional tomographic reconstruction of the object (2); wherein, during the advancement, the object (2) is made or let rotate, at least partly uncontrolled, in such a way that the object (2) rotates around one or more rotation axes which are transversal both to the motion trajectory and to the propagation directions of the x-rays crossing it; and wherein a computer also determines the spatial position in which the object (2) is located relative to the one or more emitters (4) and/or the one or more detectors (5) at the instant when each two-dimensional map is detected, and factors this in the tomographic reconstruction.

Imaging methods and imaging systems are provided. Methods and systems of the subject invention can include linearly translating a source and a detector. The source and the detector can be moved in opposite or approximately opposite directions. Acquired data can be used to reconstruct a tomographic image by using, for example, a compressive sensing technique.

A method is for generating a first synthetic mammogram. In an embodiment, the method includes acquiring a tomosynthesis dataset including a plurality of projection images of a tissue region from different projection directions in a projection angle range; reconstructing a slice image dataset based on the tomosynthesis dataset; localizing tissue changes in the slice image dataset; determining a first projection direction for a first synthetic mammogram based on the spatial distribution of the tissue changes in the slice image dataset and generating the first synthetic mammogram in the first projection direction based on the tomosynthesis dataset.

In a mammography method and apparatus to generate a tomosynthetic x-ray image of a breast of a patient, two tomosynthesis scans are successively implemented with different x-ray energies. In one of the scans, a synthetic projection image is generated from at least two projection images acquired in this scan, this synthetic projection image corresponding to a projection at a projection angle at which a projection image has been acquired in the other scan. A difference image, used to reconstruct the tomosynthesis x-ray image, is generated from this synthetic projection image and the projection image acquired in the other scan. Alternatively, in each scan a synthetic projection image is generated from at least two projection images acquired in that scan. Each synthetic projection image represents a projection at the same projection angle. A difference image, used to reconstruct the tomosynthesis x-ray image, is generated from these two synthetic projection images.

Described here is a system and method for image reconstruction that can automatically and iteratively produce multiple images from one set of acquired data, in which each of these multiple images corresponds to a subset of the acquired data that is self-consistent, but inconsistent with other subsets of the acquired data. The image reconstruction includes iteratively minimizing the rank of an image matrix whose columns each correspond to a different image, and in which one column corresponds to a user-provided prior image of the subject. The rank minimization is constrained subject to a consistency condition that enforces consistency between the forward projection of each column in the image matrix and a respective subset of the acquired data that contains data that is consistent with data in the subset, but inconsistent with data not in the subset.