The present invention relates to the processing of a magnetic resonance image of the rib cage of a subject in order to show the ribs flattened or straightened into a plane.
The rib cage in humans and most other mammals is a generally elliptical structure, when viewed in a cross-section orthogonal to the caudal-cranial axis, and surrounds a number of internal organs, including the lungs. In order to answer or provide insight into various types of medical questions, it is sometimes beneficial to view the ribs in a flattened or straightened form, so that the image of the ribs is basically unfolded with the ribs shown in a 2D representation, while still preserving relevant details of the individual ribs. For example, if a lesion is present on an inside surface of a rib, i.e., the surface thereof facing inwardly toward the internal organs, it may be difficult to see that lesion in a 3D image of the rib cage, even by computerized manipulation of that 3D image, because there may be other portions of the rib cage that present an obstruction. When the ribs of the rib cage are unfolded and seen flattened, however, all of their respective interior surfaces are clearly visible, and the presentation of the rib cage can be viewed with a single glance. The entire rib cage can be traversed with only a few scrolls, when all of the ribs are within a single plane.
Various image processing techniques are known that start with a volumetric (3D) data set that represents the rib cage and, by various image processing techniques, unfold the rib cage so that the ribs appear flattened or straightened. U.S. Pat. No. 9,189,847 is an example. The most widely used of these techniques make use of a computed tomography (CT) volumetric image as the starting point, as described in United States Patent Application Publication No. 2017/0032535. This is because, being an x-ray based technique, bone structures appear with excellent contrast in a CT image, with density variations being visible.
Such a technique is also known from United States Patent Application Publication No. 2013/0101197 that starts with a nuclear medicine emission image, such as a PET (Positron Emission Tomography) or SPECT (Single Photon Emission Computed Tomography) image.
Magnetic resonance (MR) imaging is another widely used imaging modality. MR imaging has the ability to show soft tissue structures with excellent contrast, but bone structures may be less precisely visible in a conventional MR image. A desirable feature of MR imaging is that it is very sensitive to lesions of the bone marrow, and therefore is a very useful tool for the assessment of multifocal bone diseases. Another desirable feature of MR imaging is that it does not make use of x-rays, and therefore the examination subject is not exposed to x-rays in an MR scan.
It is therefore desirable, despite the less precise visual appearance of bone structures in an MR image, to be able to use or start with an MR image in order to obtain an unfolded image of the rib cage. Being able to do so would have the advantage of avoiding the exposure of the subject to x-rays and, if it happens to be the case that an MR image of a subject is the only currently available image, and at some point in the subsequent medical diagnosis a need arises to view the rib cage in planar form, it would then not be necessary to schedule another imaging session for the patient, in order to obtain a CT image to use as the starting image.
An object of the present invention is to provide a method and an apparatus for rib unfolding, which allow the use of an MR image as the starting image that is processed in order to obtain a processed image wherein the rib cage is shown in an unfolded presentation.
This object is achieved in accordance with the present invention by a method and an apparatus wherein a computer is provided with an input data file formed of volumetric MR data that represent a 3D image of the rib cage and the lungs of a subject. In the computer, a view is selected wherein the ribs in the rib cage are represented as any smooth curved surface representing, such as a transverse slice through the 3D image, or an oblique view of the 3D image.
In the computer, the lungs in the selected view of the 3D image are used in order to define a first such smooth surface representation that is inside of the rib cage in the selected view. Also in the computer, further smooth surface representations are selectively defined starting from the first smooth curved surface representation and moving outwardly from the first smooth curved surface representation that respectively proceed through rib pairs in the rib cage in the selected image. These further smooth curved surface representations are then used in the computer to unfold the 3D image, by cutting and straightening these smooth curved surface representations, thereby obtaining an unfolded 3D image. The unfolded 3D image is then displayed at a display in communication with the computer.
As used herein, the term “smooth curved surface representation” means a curved surface that is defined by a smooth function. A “smooth function” is a well-understood term in the field of mathematics, and has many rigorous, known mathematical definitions. The most simplistic non-mathematical definition can be considered as a curve having no corners, or that is not jagged. A commonly employed mathematical definition of a smooth curve is a curve for which all derivatives exist and are continuous. Specific examples of a smooth curved surface representation are a surface defined by an ellipse, or a B-spline, or a Bezier curve.
In the unfolded image in accordance with the invention, the unfolded ribs will not appear straight, as in a conventional unfolded image starting from a CT image, but will exhibit curves in the plane in which the ribs are unfolded. Nevertheless, all or substantially all of the interior surfaces of the ribs will be visible in the unfolded image in accordance with the invention.
In a preferred embodiment, the first smooth curved surface representation is defined by segmenting the lungs in the selected view, and defining the first smooth curved surface representation using the segmented lungs. It is also possible to define the first smooth curved surface representation manually or semi-automatically, by displaying the selected view, with a user, by user interaction with an interface such as a touchscreen, inscribing a smooth curved surface representation on the interface that proceeds through the lungs.
The identification of a smooth curved surface representation that results in the unfolding, can also be done automatically, such as by a known pattern recognition algorithm that identifies when a smooth curved surface representation, in the smooth curved surface representations proceeding outwardly from the first smooth curved surface representation, proceeds through the ribs in the selected view. It is possible in the displayed image of the selected view for a user to see when one of the successively defined smooth curved surface representations proceeds through the ribs in the displayed view to a good approximation.
Using the lungs as a reference, a first smooth curved surface representation, in this embodiment indicated by the innermost ellipse with the open circles, is defined in a computer using the image shown in
Starting from this innermost ellipse, further ellipses are defined that are concentric with the innermost ellipse, but that progressively proceed outwardly from the lungs, toward the rib cage.
Since the individual ribs exhibit different curvatures, the successively defined concentric ellipses will approximate the curvature of different ones, or different groups, of the ribs. In the example shown in
Although this embodiment has been described using ellipses, any smoothed curved surface can be used, such as surface defined by a B-spline or a 3D Bezier curve.
The result of the inventive procedure is shown in
A flowchart that shows the basic steps of the method in accordance with the invention is shown in
In a second step S2, the computer selects a view through the 3D image, in which the rib cage and the lungs are visible. If it is desired to show only the lungs (and not other internal organs) in this view, these lungs can be segmented using known techniques, such as active contours, graph cuts, etc. One known procedure for lung segmentation is described in the article by Ray et al., “Merging Parametric Active Contours Within Homogeneous Image Regions for MRI-Based Lung Segmentation,” IEEE Trans. on Medical Imaging, Vol. 22, No. 2, pages 189-198 (2003).
This can be followed by generating a mesh of the lung surface from the segmentation mask, such as by using the known techniques of marching cubes or marching tetrahedra. The basic marching cube or marching square technique is described in U.S. Pat. No. 4,710,876.
The segmentation and generation of a suitable representation of the lung in the image that is used in accordance with the present invention can also be implemented manually or semi-automatically, by displaying the selected view and with a physician or technician interacting with an interface, such as a touch screen, to outline the lungs in the displayed image.
In step S3, once the lungs have been suitably designated in the image, the shape of the rib cage is predicted using the lung surfaces by fitting a smooth curved surface representation (which in turn defines a corresponding curved cylinder in the 3D data set) that best approximates the outline or shape of the lungs in the selected image. Using the lung-based smooth curved surface representation as a first smooth curved surface representation, further smooth curved surface representations (ellipse cylinders) are fitted to the rib cage, by proceeding outwardly from the lung-based smooth curved surface representation. This can be done automatically by pattern recognition that identifies when an ellipse proceeds through a number of image points that represent a pair of ribs in the rib cage that exceeds a predetermined threshold, for example. Since the pairs of ribs in the rib cage will exhibit respectively different curvatures, more than one such smooth curved surface representation may be identified, as illustrated in the example shown in
In step S4, once the one or more rib-based ellipses have been identified, the MR volume can be re-sampled based on cylinders associated therewith, by cutting and straightening the smooth curved surface representations, so that an image of the type shown in
The computer 4 is in communication with a source 6 of an MR torso image. The source 6 may be an MR apparatus that includes an MR data acquisition scanner for acquiring such an image, or the source 6 may be a database in which a previously-acquired image has been entered and stored.
The computer 4 is provided with a non-transitory, computer readable data storage medium that is included with programming instructions that configure the computer 4 in order to perform the method according to the invention.
Although modifications and changes may be suggested by those skilled in the art, it is the intention of the Applicant to embody within the patent warranted hereon all changes and modifications as reasonably and properly come within the scope of the Applicant's contribution to the art.