摘要:
For three-dimensional rendering (32), a machine-learnt model is trained (14) to generate representation vectors for rendered images formed with different rendering parameter settings. The distances between representation vectors of the images to a reference are used to select (36) the rendered image and corresponding rendering parameters that provides a consistency with the reference. In an additional or different embodiment, optimized pseudorandom sequences are used for physically-based rendering (52). The random number generator seed is selected (46) to improve the convergence speed of the renderer and to provide higher quality images, such as providing images more rapidly for training compared to using non-optimized seed selection.
摘要:
A method of generating a virtual collision map for use in optimizing planning of a radiation treatment by a radiation beam to be delivered from an associated treatment device on a gantry to a patient target volume (PTV) within a treatment volume of an associated patient disposed on a couch operatively coupled with the associated treatment device, the method comprising: acquiring segmented data, the segmented data being representative of a characteristic of the PTV, and representative of a non-treatment volume comprising an organ at risk (OAR) of the associated patient; modeling the PTV of the segmented data as a source of simulated rays projecting outwardly relative to the PTV and having a predetermined initial value; modeling the OAR of the segmented data as having an assigned ray attenuation feature to reduce an intensity of a selected one or more of the simulated rays passing through the OAR; defining a virtual map surface surrounding the PTV and the OAR; calculating an accumulated intensity value for points on the virtual map surface having the simulated rays passing through the points, the accumulated intensity value of each point being i) the predetermined initial value for areas of the virtual map surface having simulated rays passing directly from the PTV to the virtual map surface, and ii) the predetermined initial value attenuated by the assigned ray attenuation feature of the OAR for areas of the virtual map surface having the OAR disposed between the PTV and the virtual map surface; generating an intensity distribution on the virtual map surface in accordance with the calculated accumulated intensity values; and determining, in accordance with the intensity distribution, the virtual collision map defining a relationship between the associated patient disposed on the couch and plural positions of the treatment device on the gantry relative to the associated patient for delivering the radiation beam to the PTV.
摘要:
In some embodiments, an imaging system displays a multidimensional visualization of a surgical site. In some embodiments, the imaging system receives a selection corresponding to a portion of the displayed multidimensional visualization of the surgical site. At the selected portion of the multidimensional visualization, the imaging system displays a portion of a multidimensional reconstructed image which corresponds to the selected multidimensional visualization such that the displayed portion of the multidimensional reconstructed image is fused with the displayed multidimensional visualization.
摘要:
A computer-implemented medical visualization method includes identifying a three-dimensional model of an anatomical item of a particular mammal; automatically identifying an open path in three-dimensional space through the anatomical item; fitting a smooth curve to the open path; and displaying the anatomical item and a visual representation of the smooth curve to a user on a three-dimensional imaging system.
摘要:
The present invention relates to a method for generating a combined projection image (SM) from a medical inspection object, comprising the steps of capturing (S101) a set of initial projection images (PI); reconstructing (S102) a first and a second three-dimensional volume (RV1, RV2) from the set of initial projection images (PI); generating (S103) a first re-projection image (RP1) from the first three-dimensional volume (RV1) and a second re-projection image (RP2) from the second three-dimensional volume (RV2); weighting (S104) the first re-projection image (RP1) and the second re-projection image (RP2); and combining (S105) the weighted first re-projection image (RP1) and second re-projection image (RP2) for generating the combined projection image (SM).
摘要:
A method and apparatus are provided to improve large field of view CT image acquisition by using at least two scanning procedures: (i) one with the radiation source and detector centered and (ii) one in an offset configuration. The imaging data obtained from both of the scanning procedures is used in the reconstruction of the image. In addition, a method and apparatus are provided for detecting motion in a reconstructed image by generating a motion map that is indicative of the regions of the reconstructed image that are affected by motion artifacts. Optionally, the motion map may be used for motion estimation and/or motion compensation to prevent or diminish motion artifacts in the resulting reconstructed image. An optional method for generating a refined motion map is also provided.
摘要:
Steps S101 and S102 of displaying images which input two-dimensional tomographic images of the lower limb including the knee joint and displays three-dimensional images of the femur and the tibia including the knee joint from the image input, steps S103 to S105 and S106 to S108 of determining artificial joint which determine the artificial joint to be replaced from the three-dimensional images of each knee joint of the femur and tibia, and steps S113 and S114 of parameter determination which determine various parameters used in an artificial knee joint replacement operation using an alignment rod in the marrow to be inserted into the femur based on the artificial joint and reference points of the knee joint determined in steps S103 to S105 of determining femur side artificial joint are performed by a computer.
摘要:
A shape data generation method includes: generating a target shape of transformation from plural tomographic images of an object; specifying, from among plural vertices of a first shape that is a reference shape of the object, plural first vertices, each first vertex of which satisfies a condition that a normal line of the first vertex passes through a point that is located on the target shape and is located on a boundary of the object in any one of the plural tomographic images; identifying, for each of the plural first vertices, a second vertex that internally divides a segment between the first vertex and the point; transforming the first shape so as to put each of the plural first vertices on a corresponding second vertex; setting a shape after the transforming to the first shape; and executing the first specifying and the subsequent processings a predetermined number of times.