Image processing using a machine learning model is enabled, thereby accurately reducing noise to improve image quality. A medical image is acquired; and it is evaluated whether noise in the medical image exceeds a predetermined reference value. A noise reducer reduces the noise of the medical image that has been determined to include noise that exceeds the reference value. The noise of the medical image is reduced using a machine learning model constructed by collecting a plurality of learning data sets that include an image with noise as input data and an image without noise as output data. The machine learning model includes a plurality of layers that perform convolution on an image that is input, one layer of which includes a filter layer in which a plurality of linear or nonlinear filters are incorporated, and convolution coefficients of the plurality of linear or nonlinear filters are predetermined.

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.

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.

According to one embodiment, a medical information processing apparatus includes processing circuitry. The processing circuitry is configured to receive data acquired by scan for an object, and output a reconstructed image data based on the data and a trained model that accepts the data as input data and outputs the reconstructed image data corresponding to the data. The trained model is trained by learning using raw data generated based on a numerical phantom and the numerical phantom.

A medical apparatus of embodiments includes processing circuitry. The processing circuitry is configured to input third projection data to a first trained model to generate fourth projection data, the first trained model being generated through learning using first projection data collected by a first X-ray detector included in a first scanner and relatively greatly affected by scattered rays as learning data of an input side and using second projection data relatively less affected by scattered rays as learning data of an output side, the first trained model being configured to generate, on the basis of the third projection data collected by a second X-ray detector included in a second scanner, the fourth projection data in which the influence of scattered rays in the third projection data has been reduced. The first projection data is collected by the first X-ray detector in a case where a collimator provided in a first X-ray source included in the first scanner has a first opening width. The second projection data is collected by the first X-ray detector in a case where the collimator has an opening width smaller than the first opening width.

A method may include obtaining a first set of projection data with respect to a first dose level; reconstructing, based on the first set of projection data, a first image; determining a second set of projection data based on the first set of projection data, the second set of projection data relating to a second dose level that is lower than the first dose level; reconstructing a second image based on the second set of projection data; and training a first neural network model based on the first image and the second image. In some embodiments, the trained first neural network model may be configured to convert a third image to a fourth image, the fourth image exhibiting a lower noise level and corresponding to a higher dose level than the third image.

A medical image processing apparatus according to an embodiment is a medical image processing apparatus that performs processing using a trained model to generate a first output medical image by subjecting a first input medical image to predetermined processing, and includes a processing circuit. The processing circuit generates a plurality of second output medical images for a second input medical image by randomly switching ON/OFF a connection of a plurality of neurons included in the trained model.

Nondestructive evaluation (NDE) of objects can elucidate impacts of various process parameters and qualification of the object. Computed tomography (CT) enables rapid NDE and characterization of objects. However, CT presents challenges because of artifacts produced by standard reconstruction algorithms. Beam-hardening artifacts especially complicate and adversely impact the process of detecting defects. By leveraging computer-aided design (CAD) models, CT simulations, and a deep-neutral network high-quality CT reconstructions that are affected by noise and beam-hardening can be simulated and used to improve reconstructions. The systems and methods of the present disclosure can significantly improve the reconstruction quality, thereby enabling better detection of defects compared with the state of the art.

Systems and methods for reconstructing medical images are disclosed. Measurement data, such as sinogram data, is received from an image scanning system. A plurality of masks are applied to corresponding portions of the measurement data to generate a plurality of masked measurement data portions. In some examples, the measurement data is encoded before the plurality of masks are applied. A neural network including a plurality of fully connected layers is applied to the plurality of masked measurement data portions to generate a plurality of image patches. The plurality of image patches are then combined to generate an initial image. In some examples, refinement and scaling operations are applied to the initial image and corresponding attenuation maps to generate a final image. In some examples, the final image is stored in a database. In some examples, the final image is displayed for diagnosis.

Methods, devices, systems and apparatus for medical imaging, e.g., Magnetic Resonance (MR) imaging or Computed Tomography (CT) imaging, using neural networks are provided. In one aspect, an imaging method includes: determining a first neural network and a second neural network corresponding to a target imaging task, the first neural network including a first neural network parameter and a first neural network model, the second neural network including a second neural network parameter and a second neural network model, obtaining a reconstructed image by performing reconstruction for down-sampling data of a tissue under test using the first neural network, the target imaging task corresponding to the tissue under test, and obtaining an image output by a second neural network as a target image of the tissue under test by performing an image processing operation corresponding to the target imaging task for the reconstructed image using the second neural network.