Classifications

• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/70—Determining position or orientation of objects or cameras
  • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
  • G06T7/74—Determining position or orientation of objects or cameras using feature-based methods involving reference images or patches
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T7/00—Image analysis
  • G06T7/70—Determining position or orientation of objects or cameras
  • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
  • G06T7/75—Determining position or orientation of objects or cameras using feature-based methods involving models
• • A—HUMAN NECESSITIES
  • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
  • A61N—ELECTROTHERAPY; MAGNETOTHERAPY; RADIATION THERAPY; ULTRASOUND THERAPY
  • A61N5/00—Radiation therapy
  • A61N5/10—X-ray therapy; Gamma-ray therapy; Particle-irradiation therapy
  • A61N5/1048—Monitoring, verifying, controlling systems and methods
  • A61N5/1049—Monitoring, verifying, controlling systems and methods for verifying the position of the patient with respect to the radiation beam
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T17/00—Three dimensional [3D] modelling, e.g. data description of 3D objects
  • G06T17/20—Finite element generation, e.g. wire-frame surface description, tesselation
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
  • H04N13/10—Processing, recording or transmission of stereoscopic or multi-view image signals
  • H04N13/106—Processing image signals
  • H04N13/111—Transformation of image signals corresponding to virtual viewpoints, e.g. spatial image interpolation
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/10—Image acquisition modality
  • G06T2207/10028—Range image; Depth image; 3D point clouds
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/10—Image acquisition modality
  • G06T2207/10072—Tomographic images
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2207/00—Indexing scheme for image analysis or image enhancement
  • G06T2207/30—Subject of image; Context of image processing
  • G06T2207/30196—Human being; Person
• • G—PHYSICS
  • G06—COMPUTING OR CALCULATING; COUNTING
  • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
  • G06T2219/00—Indexing scheme for manipulating 3D models or images for computer graphics
  • G06T2219/20—Indexing scheme for editing of 3D models
  • G06T2219/2004—Aligning objects, relative positioning of parts
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N13/00—Stereoscopic video systems; Multi-view video systems; Details thereof
  • H04N2013/0074—Stereoscopic image analysis
  • H04N2013/0085—Motion estimation from stereoscopic image signals

Definitions

• the present invention relates to patient monitoring. More particularly, embodiments of the present invention relate to monitoring the positioning of patients and also to enable the movement of patients to be detected.
• the invention is particularly suitable for use with radio therapy devices and computed tomography (CT) scanners and the like where accurate positioning and the detection of patient movement is important for successful treatment.
• CT computed tomography
• Radiotherapy consists of projecting onto a predetermined region of a patient's body, a radiation beam so as to destroy or eliminate tumors existing therein. Such treatment is usually carried out periodically and repeatedly. At each medical intervention, the radiation source must be positioned with respect to the patient in order to irradiate the selected region with the highest possible accuracy to avoid radiating adjacent tissue on which radiation beams would be harmful.
• the gating of treatment apparatus should be matched with the breathing cycle so that radiation is focused on the location of a tumor and collateral damage to other tissues is minimized. If movement of a patient is detected the treatment should be halted to avoid irradiating areas of a patient other than a tumor location.
• stereoscopic images of a patient are obtained and processed to generate data identifying 3D positions of a large number of points corresponding to points on the surface of an imaged patient.
• Such data can be compared with data generated on a previous occasion and used to position a patient in a consistent manner or provide a warning when a patient moves out of position.
• a comparison involves undertaking Procrustes analysis to determine a transformation which minimizes the differences in position between points on the surface of a patient identified by data generated based on live images and points on the surface of a patient identified by data generated on a previous occasion. Determining an accurate match between a current patient position and the position of a patient earlier in treatment is difficult.
• portions of a treatment apparatus may obscure the view of a stereoscopic camera so that data for only part of a surface of a patient may be generated. This can mean that there is only a limited area of overlap between an imaged surface and the surface used to generate target data.
• a computer implemented method of determining a rigid transformation for matching the position of an object with the position of a target object represented by a target model comprising data identifying 3D positions of a set of vertices of a triangulated 3D wire mesh and connectivity indicative of connections between vertices
• the method comprising: obtaining stereoscopic images of an object; utilizing a computer to process the stereoscopic images to identify 3D positions of a plurality of points on the surface of the imaged object; utilizing the computer to select a set of the identified 3D positions on the surface of an imaged object as points to be utilized to determine a rigid transformation for matching the position of an object with the position of a target object on the basis of the determined distances between the identified 3D positions and vertices of the target model identified as being closest to said positions; and utilizing the computer to calculate a rigid transformation which minimizes point to plane distances between the identified set of 3D positions and the planes containing the triangles of the target model
• selecting a set of the identified 3D positions on the surface of an imaged object and calculating a rigid transformation which minimizes point to plane distances between the identified set of 3D positions and the planes containing the triangles of the target model surface identified as being closest to those points may comprise for a number of iterations: determining threshold values to be utilized for a current iteration; selecting a set of the identified 3D positions on the surface of an imaged object as points to be utilized to determine a rigid transformation for matching the position of an object with the position of a target object on the basis of a comparison of the distances between the identified 3D positions and vertices of the target model identified as being closest to said positions with the threshold; calculating a rigid transformation which minimizes point to plane distances between the identified set of 3D positions and the planes containing the triangles of the target model surface identified as being closest to those points; wherein the rigid transformation for matching the position of an object with the position of a target object comprises the sum of the transformations determined at each iteration.
• an initial threshold value for a first iteration may be set and threshold values for subsequent iterations may be determined on the basis of the average distances between the identified 3D positions and vertices of the target model identified as being closest to said positions.
• the set of 3D positions on the surface of an imaged object utilized to determine a rigid transformation for matching the position of an object with the position of a target object may be selected on the basis of the relative orientation of the object at a point and the orientation of a triangle in the target model surface identified as being closest to that point.
• the selected set may be filtered to remove any points determined to project to the same location prior to calculating a rigid transformation which minimizes point to plane distances between the filtered identified set of 3D positions and the planes containing the triangles of the target model surface identified as being closest to those points.
• calculating a rigid transformation which minimizes point to plane distances between the identified points on the surface of an imaged object and the planes containing the triangles of the target model surface identified as being closest to those points may comprise determining for each of said points, the projections of said points to the planes containing the triangles of the target model surface identified as being closest to those points; determining a translation which aligns the centroids of the points identified as being locations on the surface of an object with the projections of said points to identified planes containing the triangles of the target model surface identified as being closest to those points; and determining a rotation which minimizes point to plane distances between the identified points on the surface of an imaged object and the planes containing the triangles of the target model surface identified as being closest to those points after the determined translation has been applied.
• Identification of triangles in the target model surface closest to 3D positions of the points on the surface of an imaged object may comprise generating an array, identifying for a regular grid of points in 3D space, the vertices of the target model closest to those points; identifying the portions of the regular 3D grid closest to points on the surface of an imaged object; and utilizing the vertices associated with the identified portions of the 3D grid by the array to identify the triangle in the target model surface closest to said points.
• utilizing the vertices associated with the identified portions of the 3D grid by the array to identify the triangle in the target model surface closest to a point may comprise: determining of the vertices associated with the identified portions of the 3D grid, the vertex of the model surface closest to the point on the surface of the object currently being processed; determining if any of the vertices directly connected to the determined closest vertex is closer to the 3D position of the point on the surface of an imaged object currently being processed; if any of the directly connected vertices is determined to be closer to the 3D position of the point on the surface of an imaged object currently being processed, identifying whether any vertices connected to that vertex is closer to the 3D position of the point on the surface of an imaged object currently being processed; and when it is determined that none of the directly connected vertices is determined to be closer to the 3D position of the point on the surface of an imaged object currently being processed, determining for each triangle in the target model containing the closest identified vertex, the distance between a
• a suitable array may be generated by: using, for each vertex in a target model, the 3D coordinates associated with a vertex of a target model to identify a portion of a regular 3D grid closest to the location of the vertex; determining the distance between the vertex and the positions corresponding to points in the identified portion of a regular 3D grid; and determining whether said identified points are associated with data indicative of the points being closer to another vertex in the target model and if not associating said points with data identifying the vertex and the distance between the vertex and said points.
• the portions of a regular grid not associated with data identifying vertices can then be associated with data by traversing the regular 3D grid in each of the directions corresponding to the axes of the regular 3D grid and determining when traversing the grid, whether a vertex associated with a neighbouring point on the grid in the direction of traversal is closer to the position associated with the current grid point than any vertex previously associated with that grid point and if so associating the grid point with data identifying the vertex associated with the neighbouring grid point.
• a patient monitoring system comprising: a 3D position determination module operable to process stereoscopic images of a patient to identify 3D positions of a plurality of points on the surface of an imaged patient; a target model store operable to store a target model comprising data identifying 3D positions of a set of vertices of a triangulated 3D wire mesh model and connectivity indicative of connections between vertices; and a matching module operable to identify the triangles in a target model surface stored in the target model store closest to points identified by the 3D position determination module and calculate a rigid transformation which minimizes point to plane distances between a selected set of the identified points and the planes containing the triangles of the target model surface identified as being closest to those points wherein the selected set of points utilized to determine the rigid transformation is selected on the basis of the determined distances between the identified 3D positions and vertices of the target model identified as being closest to said positions.
• Such a patient monitoring system may additionally comprise a stereoscopic camera operable to obtain stereoscopic images of a patient, wherein the 3D position determination module is arranged to process stereoscopic images obtained by the stereoscopic camera.
• the system may also comprise a mechanical couch, wherein the stereoscopic camera is operable to obtain stereoscopic images of a patient on the couch and the matching module is operable to generate instructions to cause the mechanical couch to align an imaged patient on the basis of a calculated rigid transformation which minimizes point to plane distances between the identified points on the surface of an imaged patient and the planes containing the triangles of the target model surface identified as being closest to those points.
• system may also comprise a treatment apparatus arranged to be inhibited from operating if the matching module determines that a calculated rigid transformation which minimizes point to plane distances between the identified points on the surface of an imaged patient and the planes containing the triangles of the target model surface identified as being closest to those points is indicative of a patient being out of position by more than a threshold amount.
• Figure 1 is a schematic diagram of a patient monitor in accordance with an embodiment of the present invention.
• Figure 2 is a flow diagram of the processing of the patient monitor of Figure 1 to determine a matching translation and rotation for matching two model surfaces;
• Figure 3 is a flow diagram for the processing of the monitor of Figure 1 to populate an array identifying nearest vertices
• Figures 4A-C are schematic illustrations of the processing of Figure 3 to a portion of an array; and Figure 5 is a flow diagram of the processing of the monitor of Figure 1 to identify the nearest triangle plane in a model surface to a point on another surface.
• FIG. 1 is a schematic diagram of a patient monitor in accordance with an embodiment of the present invention.
• set of stereoscopic cameras 10 that are connected by wiring 12 to a computer 14.
• the computer 14 is also connected to treatment apparatus 16 such as a linear accelerator for applying radiotherapy or an x-ray simulator for planning radiotherapy.
• treatment apparatus 16 such as a linear accelerator for applying radiotherapy or an x-ray simulator for planning radiotherapy.
• a mechanical couch 18 is provided as part of the treatment apparatus upon which a patient 20 lies during treatment.
• the treatment apparatus 16 and the mechanical couch 18 are arranged such that under the control of the computer 14 the relative positions of the mechanical couch 18 and the treatment apparatus 16 may be varied, laterally, vertically, longitudinally and rotationally.
• the stereoscopic cameras 10 obtain video images of a patient 20 lying on the mechanical couch 18.
• the computer 14 then processes the images of the patient 20 to generate a model of the surface of the patient. This model is compared with a model of the patient generated during earlier treatment sessions. When positioning a patient the difference between a current model surface and a target model surface obtained from an earlier session is identified and the positioning instructions necessary to align the surfaces determined and sent to the mechanical couch 18. Subsequently during treatment any deviation from an initial set up can be identified and if the deviation is greater than a threshold, the computer 14 sends instructions to the treatment apparatus 16 to cause treatment to be halted until a patient 20 can be repositioned.
• the computer 14 In order for the computer 14 to process images received from the stereoscopic cameras 10, the computer 14 is configured by software either provided on a disk 22 or by receiving an electrical signal 24 via a communications network into a number of functional modules 26- 36.
• the functional modules 26-36 illustrated in Figure 1 are purely notional in order to assist with the understanding of the working of the claimed invention and may not in certain embodiments directly correspond with blocks of code in the source code for the software. In other embodiments the functions performed by the illustrated functional modules 26-36 may be divided between different modules or may be performed by the re- use of the same modules for different functions.
• the functional modules 26-36 comprise: a 3D position determination module 26 for processing images received from the stereoscopic cameras 10, a model generation module 28 for processing data generated by the 3D position determination module 26 and converting the data into a 3D wire mesh model of an imaged computer surface; a generated model store 30 for storing a 3D wire mesh model of an imaged surface; a target model store 32 for storing a previously generated 3D wire mesh model; a matching module 34 for determining rotations and translations required to match a generated model with a target model; and an nearest vertex array 36.
• the 3D position determination module 26 In use, as images are obtained by the stereoscopic cameras 10, these images are processed by the 3D position determination module 26. This processing enables the 3D position determination module to identify 3D positions of corresponding points in pairs of images.
• the position data generated by the 3D position determination module 26 is then passed to the model generation module 28 which processes the position data to generate a 3D wire mesh model of the surface of a patient 20 imaged by the stereoscopic cameras 10.
• the 3D model comprises a triangulated wire mesh model where the vertices of the model correspond to the 3D positions determined by the 3D position determination module 26. When such a model has been determined it is stored in the generated model store 30.
• the matching module 34 is then invoked to determine a matching translation and rotation between the generated model based on the current images being obtained by the stereoscopic cameras 10 and a previously generated model surface of the patient stored in the target model store 32.
• the matching module 34 determines such a matching utilising the nearest vertex array 36 in a manner which enables a match to be rapidly determined and which enables a highly accurate match to be determined.
• the determined translation and rotation can then be sent as instructions to the mechanical couch 18 to cause the couch to position the patient 20 in the same position relative to the treatment apparatus 16 as they were when they were previously treated.
• the stereoscopic cameras 10 can continue to monitor the patient 20 and any variation in position can be identified by generating further model surfaces and comparing those generated surfaces with the target model stored in the target model store 32. If it is determined that a patient has moved out of position, the treatment apparatus 16 can be halted and the patient 20 repositioned, thereby avoiding irradiating the wrong parts of the patient 20.
• the processing undertaken by the matching module 34 will now be described in detail with reference to Figures 2-5.
• Figure 2 is a flow diagram of the processing undertaken by the matching module 34, when the matching module 34 is first invoked, initially (s2-1) the matching module 34 utilizes the data stored in the target model store 32 to populate the nearest vertex array 36 to enable the matching module 34 to identify the closest vertices in 3D space of a stored model to other points in space in an efficient manner.
• processing is undertaken to identify the closest vertex in a target model to a grid of positions and this data is stored in the nearest vertex array 36.
• the nearest vertex array is utilized to identify a small number of vertices in a target model which may be the closest vertices to a particular point in space. This then reduces the volume of processing models with N points from the order of N 2 to the order of N.
• FIG. 3 is a flow diagram of the processing undertaken by the matching module 34, initially (s3-1) the matching module 34 selects a first vertex from the target model stored in the target model store 32.
• the matching module 34 then (s3-2) proceeds to utilize the geometry data identifying the position in 3D space of the currently selected vertex to identify the portion of the nearest vertex array corresponding to that area and to determine distance values identifying the distances between the position of the selected vertex and positions corresponding to the corners of a cube in a 3 dimensional grid containing that position.
• the nearest vertex array 36 comprises a data store for storing data associated with a 3 dimensional array where each of the entries in the array is associated with a point on a regular 3D grid in space.
• the geometry data of the vertex is used to identify the eight entries in the vertex array 36 corresponding to the closest point in the array to that vertex.
• the matching module 34 would identify the closest points on the grid represented by the nearest vertex array 36 as being the points associated with the positions (x,y,z), (x+1 ,y,z), (x,y+1 ,z), (x+1 ,y+1 ,z) (x,y,z+1), (x+1 ,y,z+1), (x,y+1 ,z+1), and (x+1 ,y+1 ,z+1).
• the matching module 34 then proceeds to determine the Euclidian distance between the position identified by the geometry data of the vertex being processed and each of these positions and then (s3-3) determines whether those identified positions in the nearest vertex array 36 are associated with data identifying a vertex distance greater than those determined distances.
• the matching module 34 determines that the current distance data associated with an entry in the vertex array is greater than the Euclidian distance between the position represented by the entry in the array and the vertex being processed or if the matching module 34 determines that no data is currently associated with an identified portion of the vertex array 36, the matching module then proceeds to store in that portion of the nearest vertex array 36 data identifying the current vertex and the distance between that vertex and the position associated with that entry in the array.
• vertex A was associated with co-ordinates (x + ⁇ , y + 5y, z + ⁇ )
• the Euclidian distance between the vertex and position (x, y, z) would be determined which in this example would be ( ⁇ 2 + 5y 2 + ⁇ 2 ) 1 2 . If this distance was less than the current distance associated with point (x, y, z) or no data was associated with this point, data identifying vertex A and the calculated distance would be stored.
• the matching module 34 When the matching module 34 has checked all eight of the entries in the nearest vertex array 36 associated with the 8 closest points in the array to the vertex being processed and updated the data associated with those points as necessary, the matching module 34 then determines (s3-5) whether all of the vertices of the target model have been processed. If this is not the case, the next vertex in the target model is selected (s3-6) and data in the nearest vertex array 36 is then updated (s3-2-s3-4) before the matching module 34 checks once again whether all the vertices in the target model have been processed.
• Figure 4 is a schematic illustration of the processing undertaken by the matching module 34 for part of a nearest vertex array 36.
• points A-D of an exemplary vertex are shown in positions relative to a grid where the intersections of grid lines correspond to positions associated with data in the nearest vertex array 36.
• only a two dimensional array is illustrated whereas in an actual embodiment data would be stored in a three dimensional array.
• the co-ordinate data for vertex A would enable the square containing A to be identified and the corners of that square would then be associated with data identifying vector A and the distance between the corner and the actual position of vertex A.
• vertex B the square containing vertex B would be identified and the portions of the array associated with the corners of that square would be updated to identify vertex B.
• vertex C the same points in the array would then be identified as both B and C lie within the same square.
• the corners of the array corresponding to the square which were closer to vertex C than vertex B would then be updated to become associated with vertex C rather than vertex B.
• the matching module 34 proceeds to propagate vertex and distance data throughout the grid. More specifically, starting at the bottom left hand corner of the array, each entry in the array is considered in turn. In the initial pass the entries are processed from left to right. The array is scanned until the first entry associated with vertex and distance data is identified. For each subsequent position in the array, if no vertex data is associated with a position in the array data identifying the most recently identified closest vertex and distance data identifying the distance between that vertex and the currently identified point is stored.
• the vertex associated with the portion of the array being processed is closest to that point or whether the most recently identified vertex is closest and that point in the array is associated with data identifying the closer vertex and the distance to that vertex.
• Figure 4C is a schematic illustration of the portion of the array of Figure 4B after nearest vertex data has been propagated throughout the entire array. As can be seen in Figure 4C after processing every point in the nearest vertex array 36 is associated with data identifying a nearest vertex. The way in which the vertex array is populated is such to ensure that the vertex data identifies the vertex of the target model closest to the point in space associated with the vertex. As the points identified in the vertex array are points on a regular grid the points in an array closest to any position in space can be rapidly determined from coordinate data.
• the array can then be utilised to assist with identify portions of a generated model surface corresponding to portions of the target model in the target model store 32.
• initially all vertices in a generated model stored in the generated model store 30 are considered as being suitable for being matched to part of the target model.
• FIG. 5 is a flow diagram of the processing of the matching module 34 to identify the closest triangle in a target model to a vertex in a generated model, initially (s5-1) a seed for identifying the portion of the target model closest to the position of the current vertex is identified by using the nearest vertex array 36 as a look up table based on the coordinates of the vertex being processed.
• the matching module 34 would use the vertex data in the nearest vertex array 36 to identify the vertices associated with the positions (x,y,z), (x+1 ,y,z), (x,y+1 ,z), (x+1 ,y+1 ,z) (x,y,z+1), (x+1 ,y,z+1), (x,y+1 ,z+1), and (x+1 ,y+1 ,z+1) in the nearest vertex array 34 .
• the matching module 34 would then (s5-2) identify a single seed point for identifying the closest triangle in the target model by determining which of the vertices identified by vertex data in the portion of the nearest vertex array 36 accesses was closest to the position of the vertex currently being processed.
• the distance between the position of the vertex currently being processed and the candidate vertex is determined and the vertex in the target model determined to be closest in space to the position of the vertex being process is identified.
• the matching module 34 then (s5-3) utilizes the connectivity data of the model stored in the target model store 32 to identify vertices in the target model which are connected to the identified closest seed vertex and calculates for each of the vertices the Euclidian distance between the positions of the identified vertices in the target model and the position associated with the vertex of the currently generated model currently being processed.
• the matching module 34 determines (s5-4) that any of the vertices connected to the current seed vertex is closer in space to the position of the vertex in the currently generated model being processed, the vertex identified as being closest to the position of the vertex in the generated model currently being processed is set to be the seed vertex and vertices connected to that new seed vertex are then identified (s5-5) using the connectivity data for the target model before distance values for those connected vertices are then determined (s5-3). The matching module 34 then again determines (s5-4) whether any of the connected vertices are closer in space to the vertex in the generated model currently being processed than the current seed vertex.
• the matching module 34 undertakes a walk over the surface of target model starting at a position selected using the nearest vertex array 36 and moving at each iteration towards vertices in the target model which are even closer to the position of vertex in the generated surface currently being processed.
• the matching module 36 will determine (s5- 4) that none of the vertices connected to the current seed vertex is closer in space to the position of the vertex in the generated model being processed than the seed vertex itself.
• the matching module 34 identifies which of the triangle planes containing the seed vertex is closest to the vertex in the generated model being processed.
• the matching module 34 uses the connectivity data for the target model surface to identify all triangles in the target model which include that vertex.
• a point to plane distance is then calculated for the distance between the position of the vertex in the generated model currently being processed and each of the planes containing each of the identified triangles.
• Data is then stored identifying the closest point to plane distance and the triangle in the target model contained in the plane associated with that distance.
• the matching module 34 will have identified the plane containing the triangle in the target model which is closest to the vertex being processed. If the current model and the target model correspond to the same surface it is likely that the vertex in question will correspond to a point somewhere within that triangle or somewhere nearby on the surface.
• the matching module then (s2-3) proceeds to determine a set of thresholds for selecting vertices to be utilized for determining a matching transformation.
• the matching module 34 utilizes three criteria for filtering.
• the vertices are filtered to remove any matches where the point to plane distance exceeds a threshold.
• this threshold is set to a high value such as a distance corresponding to a meter so as only to eliminate outliers and erroneous data points.
• a threshold is set based on the average point to plane distance from the previous iteration. In order to balance flexibility with a need to reduce a data set to a manageable size, it has been found that setting the distance threshold at one and a half times the average determined point to plane distance is acceptable. Such filtering enables outlier data points to be eliminated and therefore prevents such outliers from affecting a match between a current and a model surface.
• distance vertices may also be filtered using an orientation measure.
• An orientation vector for a vertex in the currently generated model can be determined by calculating the average normal vector of triangles in the currently generated model containing that vertex.
• the relative orientation of the two models can then be determined by calculating the cross product of the normal to the plane containing the identified closest triangle in the target model and the average normal vector calculated for the vertex being processed.
• Such a dot product will range from -1 to 1 where -1 indicates that the two surfaces at their closest points are oriented in opposite directions and a value of 1 indicates complete correspondence of orientation.
• the determined dot product can be compared with a threshold which initially may be set not to eliminate any vertices and which is progressively increased to require increased similarity in orientation. This progressive approach maximizes the likelihood that matched portions of a model correspond with one another as it would be expected that if points match the orientation at the surface will also match.
• the matching module 34 also checks to see if for any two vertices the projection of a vertex on to an identified triangle plane identifies the same location and if this is the case, such vertices are removed from further consideration. Elimination of such duplicates ensures that a unique transformation to minimize point to plane distances can be determined. Having determined the distance and orientation thresholds to be utilized for the current iteration, the matching module 34 then proceeds (s2-4) to identify the set of vertices which fulfill the current matching criteria. That is to say the matching module 34 identifies the set of vertices which are associated with point to plane distances less than the current distance threshold and which are associated with orientation measures no greater than the current orientation threshold and identify unique closest vertices in the stored model.
• the matching module 34 Having filtered the vertices based on distance, orientation and ensuring that each vertex is associated with a unique projection to a triangle plane, the matching module 34 then proceeds (s2-5) to determine a translation and rotation of the current model which minimizes the point to plane distances between vertices in the current model and identified closest triangle planes in the target model.
• a suitable translation can be determined by considering the translation required to align the centroid of the vertices in the current model currently being used to match the surfaces with the centroid of a set of points corresponding to the projection of each of those points to the respective closest triangle planes determined for those points. Having determined a suitable translation, a rotation for matching the two surfaces can then be determined by applying Procrustes analysis using as a measure of goodness of match a measure which seeks to minimize the sum of the squares of the point to plane distance measures associated with the vertices being matched.
• the matching module 34 determines (s2-6) whether the required number of iterations have been performed. If this is not the case the matching module 34 applies (s2-7) the determined translation and rotation to the vertex data for the currently generated model being matched.
• the nearest triangle planes in the target model to each of the vertices in the current model being utilised to match the current model to the target model are then determined (s2-2) before being filtered (s2-3-s2-4) and utilized to determine (s2-5) a translation and rotation for an improved match.
• the matching module 34 determines (s2-6) that the final iteration has been reached, then the matching module 34 then outputs (s2-8) as a calculated final transformation the sum of the translations and rotations determined for matching the generated surface to the model surface.
• the determined transformation can then either be used to cause the mechanical couch 18 to re-orientate the patient or alternative to trigger a warning or halt treatment where a patient is detected as being out of position by more than an acceptable amount.
• the determination of a matching transformation in the manner described has been determined to be able to generate a more accurate match than simply attempting to undertake Procrustes analysis to match a current and a target model based on attempting minimize the least squares distances for vertices in a pair of models.
• the vertices in models only correspond to points on the surface and not necessarily the same points on the surface of the patient. Rather than trying to determine a match between vertices in a model, the described system determines a match which facilitates the matching of vertices to any position identified as being on a target model surface including all of the surface represented by triangles. The described approach also enables such matches to be determined rapidly. Pre- populating a nearest vertex array 36 enables initial seed points for determining matches to be determined very quickly.
• the speed of matching can be further increased by only considering a subset of vertices for matching at each iteration. As described a suitable subset which retains vertices which are most likely to correspond to the area of overlap between two models can be determined by considering the relative orientation and plane to point distance measures at points being matched. By gradually increasing the threshold for continued use of a vertex only those points determined as being most suitable for matching are retained.
• the embodiments of the invention described with reference to the drawings comprise computer apparatus and processes performed in computer apparatus, the invention also extends to computer programs, particularly computer programs on or in a carrier, adapted for putting the invention into practice.
• the program may be in the form of source or object code or in any other form suitable for use in the implementation of the processes according to the invention.
• the carrier could be any entity or device capable of carrying the program.
• the carrier may comprise a storage medium, such as a ROM, for example a CD ROM or a semiconductor ROM, or a magnetic recording medium, for example a floppy disc or hard disk.
• a storage medium such as a ROM, for example a CD ROM or a semiconductor ROM, or a magnetic recording medium, for example a floppy disc or hard disk.
• the carrier may be a transmissible carrier such as an electrical or optical signal which may be conveyed via electrical or optical cable or by radio or other means.
• the carrier When a program is embodied in a signal which may be conveyed directly by a cable or other device or means, the carrier may be constituted by such cable or other device or means.
• the carrier may be an integrated circuit in which the program is embedded, the integrated circuit being adapted for performing, or for use in the performance of, the relevant processes.

Landscapes

• Engineering & Computer Science (AREA)
• Physics & Mathematics (AREA)
• General Physics & Mathematics (AREA)
• Theoretical Computer Science (AREA)
• Computer Vision & Pattern Recognition (AREA)
• Health & Medical Sciences (AREA)
• Biomedical Technology (AREA)
• Signal Processing (AREA)
• Multimedia (AREA)
• Pathology (AREA)
• Public Health (AREA)
• Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
• Radiology & Medical Imaging (AREA)
• Life Sciences & Earth Sciences (AREA)
• Animal Behavior & Ethology (AREA)
• General Health & Medical Sciences (AREA)
• Software Systems (AREA)
• Veterinary Medicine (AREA)
• Geometry (AREA)
• Computer Graphics (AREA)
• Apparatus For Radiation Diagnosis (AREA)
• Image Analysis (AREA)
• Image Processing (AREA)

Priority Applications (4)

Applications Claiming Priority (2)

Publications (1)

Family

ID=46261590

Family Applications (1)

Country Status (6)

Cited By (1)

Families Citing this family (11)

Citations (4)

Family Cites Families (5)

• 2012
  • 2012-04-19 GB GB1206915.9A patent/GB2501308A/en not_active Withdrawn
• 2013
  • 2013-04-17 CN CN201380020931.1A patent/CN104246827B/zh active Active
  • 2013-04-17 US US14/394,704 patent/US9420254B2/en active Active
  • 2013-04-17 WO PCT/GB2013/050976 patent/WO2013156775A1/en not_active Ceased
  • 2013-04-17 EP EP13718213.5A patent/EP2839432B1/en active Active
  • 2013-04-17 JP JP2015506304A patent/JP5996092B2/ja active Active

Patent Citations (4)

Non-Patent Citations (7)

Also Published As

Similar Documents

Legal Events