Method and system for generating panoramic radiographic images

Abstract

Method and system for generating panoramic radiographic images by merging contributions from panoramic images obtained with respect to a plurality of layers, wherein the method provides for performing a so-called Multipan image acquisition and for identifying the image contributions of the various panoramic images that present the best focus of details reproduced therein and for combining said contributions of said images into a fused panoramic image by applying a wavelet decomposition transform to said images for identifying said contributions of said images that represent image details with the best focus and reconstructing from these the fused image by applying the inverse wavelet transform. According to an advantageous variant, the method provides for a step of aligning the horizontal magnification of the panoramic images prior to applying the wavelet transform for decomposition and reconstruction. The invention also relates to a system for implementing said method, comprising a radiographic device and a processing unit in which a program is loaded and from which said program is executable, said program comprising instructions for the processor of said processing unit such that, when said instructions are executed, the processing unit is capable of performing the steps of the method that is the subject of the present invention.

Description

[0001] Method and system for generating panoramic radiographic images

[0002] TEXT OF THE DESCRIPTION

[0003] The present invention relates to a method for generating panoramic dental radiographic images .

[0004] The acquisition of panoramic images is a known technique and widely used for a long time . It is performed by the simultaneous rotation of an X-ray generator and a detection system, thicklayer rotational curved tomography, both mounted on a rotating gantry and positioned on opposite sides of the patient .

[0005] The X-ray source and the sensor are arranged in opposite positions to each other and at a predetermined distance and perform a common rotary and / or roto-translatory movement around the patient' s head, which is positioned in an intermediate position between said source and said sensor , describing an arc of a predetermined angular amplitude , for example about 270 ° . The final image is the result of reconstructions of individual image sections of the maxillofacial area . The one-to-one correspondence between each point of the dental arches crossed by a collimated beam and its projection on the sensor or detector is the basis of the formation of the panoramic image .

[0006] Panoramic images are generated by the geometric projection of structures located in the maxillofacial area . The acquisition mode takes into account the geometric distortions that occur in anatomical structures depending on their position relative to an X-ray source . In particular , structures placed closer to the source appear enlarged in their dimensions ; conversely, structures located far from the source appear smaller .

[0007] The shape and width of a focal band are determined by the path and speed of the sensor and X-ray source pair , by the alignment of the X-ray beam, and by the setting of the collimator . Generally, in panoramic imaging , the focal band, also called focal plane or image layer , has limited dimensions , ranging from 6 mm to 8 mm and from 12 mm to 16 mm, respectively in the anterior and posterior areas . A distinct image of the structures located within the focal band is obtained, with the structures positioned at the center of said focal band being the sharpest , while objects located outside the focal band appear blurred and / or distorted and / or reduced / enlarged relative to their dimensions along different directions . Typically, teeth and other structures appear blurred, shortened, and narrowed, or blurred and enlarged in the horizontal direction when such structures are positioned respectively in front of or behind the focal band or a layer . The geometric distortions of anatomical structures are minimal in the intercanine area , since the anterior area is included in the focal band thanks to a mandatory positioning of the incisors by means of patient positioning devices , such as so-called bites and other positioning devices . In the molar areas , on the contrary, geometric distortions are more evident (5-10% compared to actual dimensions) due to interindividual anatomical variability in the conformations of the dental arches .

[0008] Recently, advances in panoramic imaging have had a significant impact on improving image sharpness without neglecting aspects related to patient health . The design of the most modern dental panoramic units provides for multi-layer recording , flexible collimation of the radiation beam, rapid scanning modes , correction of positioning errors and motion artifacts , and reduction of effective doses .

[0009] In relation to issues concerning focusing , multi-layer panoramic imaging generates a series of different focusing layers parallel to each other and thin within the focal band so that , through subsequent processing , manual or automatic , of the image , it becomes possible to select for each acquired layer the best segments of said layer with respect to focusing , and these segments are then combined with other segments from other layers and relating to other sections of the dental arch , generating a complete panoramic image that presents a high level of focusing for the entire dental arch reproduced in the panoramic image . This allows the physician to have a high-quality diagnostic image available , useful for analyzing and identifying any pathologies and / or anomalies .

[0010] During the rotation or roto-translation of the source and sensor pair positioned opposite each other and with the target head interposed between them, raw image data relating to a succession of frames are acquired . From these , different images can be reconstructed, each corresponding to a specific focusing layer , that is , to a predetermined trajectory of the focal point . Said frames are then combined according to the principle of tomosynthesis known as "shift and add, " and depending on the selected frames , it is possible to reconstruct a panoramic image of a predetermined layer . From the raw image data acquired during a scan , it is therefore possible to reconstruct panoramic images relating to a plurality of different layers placed side by side . In the panoramic images along said two or more trajectories of the focal point , i . e . said two or more layers , the reproduced details appear blurred in correspondence with the different zones of the dental arch under examination . Current techniques therefore provide for combining one or more frames relating to said two or more layers in which the reproduced image presents the best focusing for different zones or different sectors of the dental arch reproduced in said plurality of panoramic images . Various known techniques exist for combining frames relating to panoramic images reconstructed along different layers , which generally involve the use of image fusion algorithms . However , in addition to the focus condition , between two different layers the magnification ratio of the reproduced details also varies , especially in the horizontal direction , so that the combined image turns out to be affected by distortions due to the different magnifications of the frames taken from the different layers and combined with each other . While focusing is improved, on the other hand, the image obtained from the fusion of frames from the various layers presents alignment defects of said frames with respect to the same magnification , and therefore the fidelity of reproduction of the details reproduced in the fusion image is lost .

[0011] The document NGUYEN VAN-GIANG ET AL : "Autofocussing panoramic image by all-in-focus image fusion" , 2020 RIVF International Conference on Computing and Communication Technologies (RIVF) , October 14 , 2020 , pages 1 to 3 , describes a method and a system according to points a) to d) of main claim 1 .

[0012] The document US2022 / 351347 describes the use of a decomposition algorithm by means of wavelet in an all-in-focus method . The present invention therefore aims to provide a method and a system for generating panoramic radiographic images that can improve current techniques for selecting and combining frames taken along two or more layers of a so-called multi-layer acquisition , overcoming the limitations of current fusion techniques both from the point of view of magnifications and from the point of view of better quality in terms of clarity and definition , that is , focusing of the image reconstructed from the fusion of the selected frames .

[0013] According to a first embodiment, the present invention relates to a method for generating panoramic radiographic images comprising the following steps : a) selection of an acquisition trajectory defined by the path of rotation and / or roto-translation of a pair consisting of an X- ray source and an X-ray detector which are positioned opposite each other and move together along said path of rotation and / or roto-translation around the patient' s head, which is positioned between said X-ray source and said X-ray detector during the entire course of rotation and / or roto-translation ; b) execution of said rotation and / or roto-translation with simultaneous emission of X-rays towards the patient' s head and towards the opposing sensor and storage of the raw image data generated by said sensor ; c) definition of at least two or more focusing layers and generation for each layer of a corresponding panoramic image from the raw image data acquired by tomosynthesis of the image frames of the corresponding layer , which layers correspond to focusing curves related to a predetermined focal length and / or within a predetermined depth of focus relative to said acquisition trajectory, also referred to as generating trajectory; d) identification in said images of zones thereof , or sections thereof , in which predetermined details and / or predetermined sections of the dental arch are reproduced, and where the image presents the best focus with respect to said details and / or said image section of the additional layers of said plurality; e) generation of the panoramic image of the dental arch by combining said zones and / or said sections of the panoramic images of the two or more layers selected in step d) ; and wherein : step d) is performed using a wavelet decomposition algorithm on said panoramic images relating to said two or more layers , which provides for decomposition into at least two or more levels , of which one main level at higher resolution and at least one further level at lower resolution , generating for each level of decomposition a predetermined number of detail coefficient matrices and a corresponding approximation coefficient matrix ; step e) is performed by recombining the approximation and / or detail coefficients of the wavelet decomposition calculated for each of said panoramic images of said two or more layers for said two or more levels and which present greater values for each of said images .

[0014] A more detailed description of the characteristics of wavelet transforms , and in particular of a DWT (Discrete Wavelet Transform) , is contained in the document "Trasformata di

[0015] Wavelet" chapter 2 published online at

[0016] / PhD. / cap wavelets . pdf , the content of which defines the basic technical knowledge of the person skilled in the art regarding the technical meaning of the terms used in the present description and in the claims .

[0017] The type of wavelet decomposition algorithm can be chosen from any of the known wavelet technologies in the state of the art .

[0018] However , many decomposition and reconstruction algorithms based on wavelets present shifts between input signals and reconstructed signals . For example , technologies known under the name DWT (Discrete Wavelet Transform) generate shifts in relation to the energy level between input signals and reconstructed signals following the transformation .

[0019] According to one embodiment , the present invention provides that the wavelet transformation algorithm used is an algorithm called DT-CWT , or Dual Tree Complex Wavelet Transform. In an embodiment , said algorithm includes filters consisting of quasi -orthonormal bidirectional filters for the real and imaginary parts of the coefficient matrices related to the first decomposition level and which consist of filters of the type referred to as Q-shift filters for the further one or more decomposition levels , which are different for the real and imaginary parts of the coefficient matrices related to said further decomposition levels . With respect to the Dual Tree Complex Wavelet Transform technology, the general detailed characteristics thereof are known from the document "The Dual-Tree Complex Wavelet Transform" by Ivan W . Selesnick , Richard G . Baraniuk , and Nick G . Kingsbury published online at https : / / eeweb . engineering . nyu . edu / iselesni / pubs / CWT_Tutorial . p df .

[0020] According to a further feature , step d) provides :

[0021] - for the coefficient matrices of the wavelet decomposition at the first level , namely the so-called approximation coefficient matrices , relating to each of said panoramic images respectively of said two or more layers , the generation of standard deviation maps of the coefficients of said matrices and the selection of one or more zones of the panoramic image relating to a corresponding layer in which said standard deviation is greater compared to the other zones ;

[0022] - for the coefficient matrices of the wavelet decomposition of said further level or said further levels , respectively for each level , the calculation of the energy of the coefficients of the corresponding detail coefficient matrix for each of the panoramic images relating to said two or more layers and the selection of the image zones of the panoramic images relating to said two or more layers where said decomposition coefficients exhibit the highest energy levels .

[0023] According to an alternative embodiment , the energy level is defined by the following equation : where :

[0024] Mi is the detail coefficient at decomposition level n and Enis said energy .

[0025] According to a further feature , in combination with the embodiment described above for step d) , step e) provides for performing a wavelet reconstruction step using as reconstruction coefficients of the wavelet functions for said two or more decomposition levels the coefficients selected in step d) for each of the panoramic images relating to said two or more layers .

[0026] Therefore , the zones of the panoramic images reconstructed by tomosynthesis along each of said two or more layers and whose wavelet decomposition coefficients meet the criteria of greater standard deviation for the first decomposition level and greater energy for the further decomposition levels are combined to obtain a final panoramic image that presents a depth of focus such as to reproduce the details of the dental arch with the best possible focus along said arch and for a predetermined depth , that is , for a band along a predetermined trajectory having a predetermined width .

[0027] According to a further additional feature , the present invention provides , prior to performing step d) , an additional step cl ) which consists of aligning the panoramic images along at least said two or more layers relative to one of said layers in terms of the magnification of the structures reproduced in the images of the different layers , while steps d) and e) are carried out with reference to the panoramic images of said two or more layers aligned with each other in step cl ) .

[0028] An embodiment of said alignment provides for performing an algorithm for modifying and standardizing at least the horizontal magnification , which algorithm provides the steps of associating each image layer with a corresponding focusing curve of the panoramic image relating to said layer for each panoramic image corresponding to a focusing curve , rescaling the value of the horizontal magnification so that it is identical to the vertical magnification intrinsic to the physics of the acquisition trajectory and / or the vertical magnification of a generating focusing curve and / or of the scanning trajectory, that is , the generating trajectory defined in step a) .

[0029] In one embodiment , the realignment between the panoramic images reconstructed by the tomosynthesis process referred to as shift & add is performed by defining a first focusing point shift curve during the common roto-translation of the source and the X-ray detector referred to as generating isocurve and determining the corresponding shift & add vector ; defining at least one or a predetermined number of additional so-called isocurves related to focusing point paths parallel and adjacent to the generating isocurve and at different distances from said generating isocurve and determining for each of said isocurves the corresponding shift & add reconstruction vector of the corresponding panoramic image ; generating a cumulative vector from the shift & add vectors based on the reference shift & add vector and the reconstruction shift & add vector and using a shrink algorithm to calculate the value of an interpolation function between said cumulative reference shift & add vector and said cumulative reconstruction shift & add vector for integer numbers of the argument of the function corresponding to the indices of the final image columns , while the values calculated by the algorithm define the indices of the columns of the initial image , which are redistributed in the final image according to the order defined by said indices of the final image , that is , the integer values of the interpolation function argument , with said steps being performed for each of the panoramic images reconstructed with reference to each of the isocurves .

[0030] The present invention also relates to a system for implementing the aforementioned method, which system comprises : a support structure for cantilevered support of an X-ray source and an X-ray detector which are carried by an arm or a combination of arms in a relative position in which the X-ray source is opposite the X-ray detector , with the emission side of the source oriented towards the reception side of the detector , while an examination position for the patient' s head is provided , optionally defined by positioning devices , which position is interposed between said X-ray source and said X-ray detector ; at least one processing unit for processing the signals acquired by the X-ray detector , in which processing unit is loaded or loadable an image processing program comprising instructions to enable said processing unit to perform the steps of the method according to one or more of the embodiments and / or variants and / or features described above .

[0031] According to one embodiment , said processing unit comprises : a unit for defining movement trajectories of the X-ray source and the X-ray detector and for controlling the movement of said arm that carries said X-ray source and said X-ray detector along said trajectory; a unit for reconstructing panoramic images from the image data relating to the signals acquired by said X-ray detector , which reconstruction unit comprises an input interface for the definition parameters of two or more layers corresponding to a respective focusing curve and performs the steps of determining the image frames relating to said two or more layers and tomosynthesis of said frames to generate a panoramic image corresponding to each of said two or more layers ; a unit for defining the focusing quality of the structures reproduced in the various zones and / or the various sections of said panoramic images and for selecting said zones and / or sections having the best focus ; a unit for merging said zones and / or sections of said panoramic images along said two or more layers , generating a final panoramic image having the best focus along its entire extension , and wherein : the unit for defining the focusing quality of the structures reproduced in the various zones and / or sections of said panoramic images and for selecting said zones and / or sections having the best focus comprises a wavelet decomposition section of said panoramic images relating to said two or more layers , in which wavelet decomposition section a wavelet decomposition program is loaded or loadable , in which the instructions for the execution of said wavelet decomposition and the settings of the criteria for the identification and selection of the coefficients of one or more decomposition levels corresponding to the zones and / or sections of the panoramic images having the best focus are coded, and which wavelet decomposition section executes said program, whereby it is capable of performing said wavelet decomposition step and said step of identifying and selecting the coefficients of one or more decomposition levels corresponding to the zones and / or sections of the panoramic images having the best focus ; said unit for merging said zones and / or sections of said panoramic images along said two or more layers comprising a merging section and a wavelet recombination section , in which section a merging and / or wavelet reconstruction program is loaded or loadable , in which the instructions for reconstructing a panoramic image comprising said zones or sections of the panoramic images relating to said two or more layers corresponding to the wavelet decomposition coefficients that meet said criteria for the identification and selection of said wavelet decomposition coefficients are coded .

[0032] According to a further feature , the processing unit further comprises a unit for aligning the panoramic images relating to said two or more layers , in which alignment unit an alignment program is loaded or loadable , which comprises the instructions that render said alignment unit capable of executing a rescaling and standardization algorithm for at least the horizontal magnification , and in particular an algorithm that performs a modification of the images relating to said two or more layers according to the method steps described above .

[0033] These and other features and advantages of the present invention will become more clearly apparent from the following description of some exemplary embodiments illustrated in the accompanying drawings , in which :

[0034] Figure 1 shows an example of a device for acquiring panoramic images .

[0035] Figure 2 schematically shows the geometry of the acquisition system.

[0036] Figure 3 shows the set of isocurves , i . e . , the different trajectories of the focal point along which the various panoramic images are reconstructed, from the combination of which an image having improved focus is generated .

[0037] Figure 4 shows the horizontal magnification calculated for each of the isocurves of Figure 3.

[0038] Figure 5 shows the vertical magnification calculated for each of the isocurves of Figure 3.

[0039] Figure 6 shows the graph relating to the shifts calculated according to the tomosynthesis algorithms "shift and add" for each isocurve , for which the horizontal magnification is equal to the vertical magnification .

[0040] Figure 7 shows an example of two isocurves for panoramic images of different layers , with the aid of which the method for calculating shrink vectors from the two "shift & add" vectors of the panoramic images relating to said two isocurves is illustrated, to align the horizontal magnification of said two panoramic images .

[0041] Figure 8 shows the horizontal magnification of the frames for the reconstruction trajectories according to the two isocurves of Figure 7 . Figure 9 shows the values of the shift vector as a function of the frame index number referring to the two isocurves of Figure 7 .

[0042] Figures 10 and 11 show the panoramic images reconstructed with reference to the two isocurves of Figure 7 .

[0043] Figure 12 shows the reference shift vector calculated as a function of a non-limiting example of a generating isocurve , the alignment algorithm being applicable to align the images of any possible isocurve for the acquisition system and the related shift & add vector .

[0044] Figures 13 and 14 respectively show the redistribution of the image columns in the case where the shrink vector has not been applied to the panoramic image of Figures 10 and 11 and in the case where said shrink vector has been applied to align said images with reference to the horizontal magnification .

[0045] Figure 15 shows a graphical example of an interpolation function of a shrink algorithm used for alignment between individual panoramic images relating to the different isocurves .

[0046] Figure 16 shows a flowchart of the method for combining panoramic images obtained by means of a so-called multipan acquisition for different layers in order to generate a panoramic image having improved focus across the entire image .

[0047] Figure 17 shows a block diagram of the process for merging two images using a decomposition and reconstruction algorithm by means of a wavelet transform.

[0048] Figure 18 shows a block diagram of the operating principle of a wavelet transform called DT-CWT , i . e . , Dual Tree Complex Wavelet Transform.

[0049] Figures 19. 1 to 19.3 show in greater detail the steps of the method according to the present invention relating to the decomposition of multipan images by means of the wavelet transform, of the merging of said images , and of the reconstruction by means of the wavelet transform of a final image having improved focus along the entire extension of the image .

[0050] Figure 20 shows in grayscale the color maps relating to the decomposition coefficients of various levels of the wavelet transformation algorithm.

[0051] Figure 21 shows in grayscale the map of the relative difference , normalized by division by the maximum wavelet coefficient , between the modulus of the maximum wavelet coefficient and the modulus of the minimum wavelet coefficient in the same image pixel for each pixel across the different layers of the planned isocurves ; thus , if the difference is very high , it means that the focus variation on that pixel across the various layers is very high , whereas if the difference is close to 0 , it means that the focus variation on that pixel across the various layers is low .

[0052] Figure 22 shows a panoramic fusion image reconstructed by merging panoramic images of a multipan system according to the present invention .

[0053] Images 23. 1 to 23.3 show various comparisons between panoramic images generated without applying the method of the present invention and with the application of the method of the present invention , in which the areas of improved focus are highlighted by circled lines .

[0054] In Figure 1 , a radiographic machine for acquiring panoramic images is shown . The machine comprises a column 2 with a stationary part resting on the floor by means of a foot 10 and optionally an anchoring 12 in the upper area to a wall 8 .

[0055] The column 2 slidably supports along its vertical length an image acquisition and patient positioning unit globally indicated as 3. Said unit comprises , at its upper end, a horizontal arm projecting in cantilever for a predetermined measure and containing a mechanism for ro to- translation of a small shaft not illustrated in detail as it is not the subject of the present invention , to which a second arm 4 is suspended, carrying at its opposite ends respectively an X-ray source 5 and an X-ray detector 6 .

[0056] A further arm located below the upper arm carries at its free end, which is vertically coincident or aligned with the area between the X-ray source 5 and the X-ray detector 6 , a patient positioning device globally indicated as 7 . This is a known device and can be implemented in various configurations present in the state of the art , such as in the form of a so- called bite or in the form of cooperating elements with anatomical parts of the head intended to hold the head in a predetermined position relative to the X-ray source and the X- ray detector .

[0057] As regards the radiographic image acquisition machine , it is understood that the one illustrated and described here is only an example among the many possible variants also present in the state of the art . For the purpose of acquiring radiographic image data and generating panoramic images using the well-known tomosynthesis principle called shift & add, the machine must be constructed so as to rotate together around the patient' s head the X-ray source 5 and the X-ray detector 6 along a predetermined angular path and according to a trajectory defined by a rotation of said shaft and said combined source-detector pair , combined with a translation of said shaft in the horizontal plane , i . e . , in a plane transverse to its longitudinal axis .

[0058] The raw image data acquired during scanning , i . e . , the simultaneous rotation of the source and detector and the emission of X-rays toward the patient' s head as well as their reception after passing through are supplied to a processing unit indicated as 9 , which may consist of generic processing hardware such as a computer , a personal computer , or similar , comprising the usual interfaces for data and / or command input , for storage and retrieval of data and / or images , for image display, and other optional peripherals globally indicated as 20 , such as a memory stick reader / writer or other rewritable and non-rewri table storage media .

[0059] In addition to the program for reconstructing panoramic images from raw image data acquired during scanning , the processing unit may also be provided in combination with a control unit for machine operation . This control unit may also be implemented only in software and include control programs executed by the same processing unit that performs image reconstruction .

[0060] During scanning , the ro to- translation movement of the source-detector pair 5 , 6 is imposed so that the trajectory of the focal point substantially follows a path coinciding with the dental arch curve . During roto- translation , image data are acquired at a predetermined frequency . The acquired image data allow reconstruction of panoramic images along a plurality of different focal point trajectories , parallel and adjacent to each other , at different distances from a generating or reference trajectory coinciding with the curve formed by the dental arch . The panoramic image is obtained by combining a sequence of image frames , and the focusing curve— called a layer to which the reconstructed panoramic image refers is defined by selecting individual image frames generated from said raw data according to their relative position in said sequence .

[0061] Figure 2 shows the path of the source 20 , that of the detector 21 , and the focusing curve 22 , i . e . , the trajectory of the focal point along the dental arch , as well as the intersection lines 23 between source and detector referring to the so-called frame step , i . e . , the step between one frame and the next in a sequence of frames reconstructed from raw image data .

[0062] The graph of Figure 3 shows the generated isocurves . By way of example and not limitation , a total of five isocurves have been selected and shown , in addition to the generating one . In the example shown , and without this constituting a limitation , the isocurves have a focal depth of 13.5 mm, the first being placed at -2 . 75 mm relative to the selected generating isocurve and the last at 10 . 75 mm from said generating isocurve .

[0063] The generating isocurve can be selected at will according to considerations dictated by the characteristics of the system. Furthermore , the value of the focal depth is a parameterized variable of the system whose setting can be varied .

[0064] The aforementioned isocurves represent the path of the focus along the dental arch at different depths of its thickness in the plane subtended by said arch , and by combining the frames acquired according to different values of the shift & add vectors used for reconstructing the panoramic images , it is possible to reconstruct a panoramic image relating to the layer of the dental arch along each of said isocurves and along the so-called generating isocurve .

[0065] Graph of figure 4 shows , as known and caused by the physics of the radiographic system, the different magnifications in the horizontal direction of the panoramic images reconstructed with reference to the different isocurves and before applying a step of aligning the magnification of the panoramic images according to the different isocurves .

[0066] Similarly, Figure 5 shows the different magnifications of said panoramic images in the vertical direction .

[0067] Figure 6 shows the shift values between frames relating to the different isocurves that have been calculated to obtain a horizontal magnification identical to the vertical magnification for the panoramic images relating to each of the isocurves .

[0068] With reference to Figure 16 , this figure shows the flowchart of the steps of the method for generating panoramic radiographic images having improved focus across the entire extension of the panoramic image , obtained by merging panoramic radiographic images acquired along different layers , i . e . , along different isocurves .

[0069] Step 160 : A predetermined trajectory of the focal point path corresponding to a generating isocurve and to a layer of the dental arch is selected .

[0070] Step 161 : Calibration of the raw data relating to the acquired frames is performed .

[0071] Step 163 : The alternative of automatic or manual patient positioning is provided in relation to the source-detector pair (5 , 6) of Figure 1 .

[0072] If automatic positioning is not performed, proceed to Step 164 , where a so-called PAN Rec Algorithm is executed , reconstructing a panoramic image for each of the planned isocurves . These panoramic images are then processed to identify the pixels of the panoramic images already reconstructed by tomosynthesis and relating to the selected isocurves that present the best focus conditions , and to merge the images of said plurality of panoramic images to generate a final panoramic image having improved focus along its entire extension .

[0073] The reconstruction process of a plurality of panoramic images relating to two or more isocurves , such as the five in the present example , is known in the state of the art as a multipan process . As evident from what has been explained above with reference to the graphs of Figures 3 to 7 , these panoramic images relating to the different isocurves present different magnifications , particularly with reference to horizontal magnification , since focal distances vary between isocurves and therefore determine different magnifications of the illustrated details .

[0074] The method according to the present invention provides a step indicated as 165 , in which the different panoramic images are modified to align their horizontal magnifications , i . e . , to modify said panoramic images relating to the different planned isocurves to a common horizontal and vertical magnification .

[0075] Once the panoramic images have been aligned at step 166 performs a decomposition of each of said images by means of a wavelet transform.

[0076] At Step 167 , based on the decomposition coefficients of the approximation matrix and the detail matrices respectively for decomposition level 1 and for the subsequent additional decomposition levels , a fusion algorithm is executed that provides selection criteria for the decomposition coefficients of said matrices to generate matrices whose coefficients consist of said selected coefficients .

[0077] Step 168 provides for reconstructing a fused panoramic image by means of the inverse wavelet transform using the decomposition coefficient matrices comprising said coefficients selected at Step 167 .

[0078] If at Step 163 the automatic alignment function is selected, a Step 169 is provided for wavelet decomposition of a plurality of panoramic images relating to the selected isocurves , each relating to at least one or to a plurality of specific ROIs (Regions of Interest) defined in said panoramic images .

[0079] The values of the wavelet decomposition coefficients are used to evaluate the focus conditions relating to the various selected isocurves and to generate , based on these coefficients , selection parameters for a new set of isocurves with reference to which Steps 170 of reconstruction of panoramics along the isocurves defined in the new set and 171 of alignment of magnification are executed . Following this last step , the process continues with Steps 166 , 167 , and 168 .

[0080] Regarding Step 165 for aligning multipan panoramic images , i . e . , reconstructed with reference to the different planned isocurves and / or Step 171 provided in combination with automatic patient positioning , Figures 7 to 15 show various features of an example of this step , referred for simplicity to only two isocurves ISO1 and ISO2 and to a generating isocurve (not illustrated) .

[0081] The alignment step therefore provides for defining a generating isocurve and additional parallel and adjacent isocurves at different distances from said generating isocurve , as shown in Figure 7 .

[0082] As shown in Figure 8 , from the isocurves it is possible to calculate the magnification for the individual frames forming the panoramic images relating to each of them, as indicated with MISO1 and MISO2 in Figure 8 .

[0083] Shift & add vectors of the frames of each panoramic image relating to each of the isocurves are calculated so that the vertical magnification equals the horizontal magnification . Figure 9 shows , for each frame identified by its position index in the sequence of frames composing a panoramic image for the corresponding isocurve , the shift & add values for the tomosynthesis of said image , indicated in Figure 9 as SHF-ISO1 and SHF-ISO2 respectively for isocurve ISO1 and ISO2 of Figure 7 .

[0084] This step generates panoramic images relating to the two isocurves that present a horizontal magnification equal to the vertical magnification . However , the magnifications of the various images are still different from each other , since the starting isocurves are different . Figures 10 and 11 show the panoramic images relating to the two isocurves ISO1 and ISO2 rescaled so as to present a horizontal magnification identical to the vertical magnification of the corresponding image . It is possible to verify visually that there are dimensional differences in the horizontal direction of the image , even if slight . However , these differences are significant when the image is used, for example , as a reference for reconstruction or other procedures . According to one aspect of the present invention , to achieve alignment or compensation of the different magnification between the various images relating to the different isocurves , the use of a so-called shrink algorithm is provided .

[0085] In one embodiment , for the shrink algorithm it is necessary to define a reference shift & add vector . In the present example , the reference shift & add vector is defined as the vector relating to the trajectory associated with the generating isocurve , whose values , as a function of the frame indices relating to their relative position in the frame sequence , are indicated in the graph of Figure 13.

[0086] The shrink algorithm has the effect of redistributing the content of the panoramic images according to the different isocurves (Figures 11 and 12 ) following the reference shift & add vector calculated based on the generating isocurve and illustrated in Figure 13. This redistribution effect is obtained by means of a linear interpolation operation using the reference shift & add vector as defined above and the shift & add vector used during reconstruction of the individual panoramic images relating to the different isocurves . In this way, the shrink vector is calculated, allowing local resizing of the image .

[0087] According to one embodiment , for calculating the shrink vector , the reference shift & add vectors and those used during reconstruction of the MULTIPAN panoramic images , i . e . , for each of the defined isocurves , are taken . The cumulative vectors of these shift & add vectors are calculated . The cumulative vector indicates where the i-th frame will be added in the tomosynthesis algorithm. For example , if frame number 200 corresponds to a cumulative value of 300 , the shift & add algorithm will start from column 300 of the image for frame number 200 . The two cumulative vectors— one obtained from the reference shift & add vector and the other obtained from the shift & add vector used for reconstructing the panoramic images relating to the different isocurves— are used to generate an interpolation function . On the x-axis , the reference cumulative vector is placed, while on the y-axis , the cumulative vector for reconstructing the panoramic images relating to the planned isocurves is placed .

[0088] Once the interpolation function is defined, the value of this function is calculated for integer arguments "x" ( 1 , 2 , 3 , 4 , 5 . . . ) . These integers represent the indices of the columns of the final image , i . e . , the image processed with the shrink algorithm, while the value calculated by the function (the "y") is the number of the column of the initial image that must be redistributed in the new image . Thus , the content of the column identified by index y in the initial image will become the content found in the column identified by index x of the final output image of the shrink algorithm.

[0089] The steps described above are carried out for each reconstructed image relating to one of the different planned isocurves , so that all these images are aligned to a "standard" image , i . e . , the one obtained using the reference shift & add vector for its reconstruction .

[0090] Figures 14 and 15 show the transformation with the shrink vectors and the condition in the absence of said transformation , respectively, for the panoramic image relating to isocurve IS01 and to isocurve IS02 .

[0091] With reference to the decomposition steps of the panoramic images relating to the defined isocurves— typically five to seven , but not limited to these values— the general principle of the process for decomposing and generating a fused image from two images by means of a wavelet transform is illustrated in Figure 18 . In this figure , an example is shown of merging two images II and 12 to which a wavelet transform is applied, having at least two decomposition levels , preferably at least four decomposition levels , as illustrated and provided for the method according to the present invention and as will be described in greater detail in the following description . The wavelet transform provides for each level respectively an approximation coefficient matrix for the first level and a plurality of detail coefficient matrices for the subsequent levels . These matrices are schematically illustrated by the icon identified as "wavelet coefficients" in Figure 8 .

[0092] For generating a fused image , the process by means of wavelet transform provides that combination rules for the decomposition coefficients of the various levels are defined, as indicated by the arrow "Fusion Rules . " The selection and combination criteria for the coefficients of the various decomposition matrices may differ , particularly for the initial decomposition level 1 and for the subsequent levels , as indicated by the example in Figure 18 with criterion 1 ) Pixel -wise and 2 ) Region-based . Based on these criteria , matrices of coefficients obtained from the combination of decomposition coefficients for the various levels are generated, while applying the inverse wavelet transform using said matrix of combined coefficients generates the fused image I from the two initial images II and 12 .

[0093] Various types of wavelet transforms are known . According to one feature of the present invention , the wavelet transform known as DUAL TREE COMPLEX WAVELET TRANSFORM (DT-CWT) has proven to offer the best performance for generating a fused panoramic image having the best image quality, i . e . , the best focus across the entire area of the image itself .

[0094] Further details on this type of transform are described in various publications , including "The Dual-Tree Complex Wavelet Transform" by Ivan W . Selesnick , Richard G . Baraniuk , and Nick G . Kingsbury, published online at : https : / / eeweb . engineering . nyu . edu / iselesni / pubs / CWT_Tutori al . pdf .

[0095] The following list indicates the advantages of this type of transform compared to other wavelet transforms known in the state of the art : a) Approximate shift invariance ; b) Good directional selectivity in two dimensions ; c) Phase information ; d) Perfect reconstruction using short linear-phase filters ; e) Limited redundancy, independent of the number of scales , 2 : 1 for ID (2m: 1 for mD) ; f) Efficient computation of order N : only twice that of simple DWT for ID (a factor of 2m for mD) .

[0096] It has the ability to differentiate positive and negative frequencies and produces six sub-bands oriented at ±15 ° , ±45 ° , ±75 ° .

[0097] Figure 19 shows a diagram of the functional principle of a DT-CWT wavelet transform with four decomposition levels , in which two branches a and b are provided . Furthermore , Figure 19 shows the two-band reconstruction block by applying the inverse transform.

[0098] With reference to Figures 20 . 1 to 20 .3 , these schematically show in detail Steps 166 to 168 of the flowchart according to Figure 17 .

[0099] The decomposition step provides for decomposition into four levels by means of a DT-CWT wavelet transform of each of the panoramic images relating to each of the defined isocurves and / or focusing layers .

[0100] Each decomposition level ( 1 to 4 ) produces an approximation coefficient matrix indicated as MAI to MA4 in Figure 20 . 1 and six detail coefficient matrices indicated as MD1 to MD4 .

[0101] The decomposition matrix of level 1 is filtered by means of a quasi -orthogonal filter Fl identical for both branches of the transform.

[0102] For the matrices relating to levels 2 to 4 , filtering is performed by means of so-called Q-filters F2 , which in one embodiment are chosen differently for the two branches of the transform.

[0103] Furthermore , for each level from 1 to 4 , the dimensions are halved .

[0104] The decomposition process is performed for each of the panoramic images relating to the different planned isocurves defined at Step 160 .

[0105] Figure 20 .2 shows in greater detail the image fusion step by means of fusion algorithm 167 , in which the selection criteria for the decomposition coefficients of the various matrices for the various decomposition levels ( 1 to 4 ) are set , and in which the coefficients of the approximation matrix and the detail coefficient matrices corresponding to said criteria are selected .

[0106] For the approximation coefficient matrices for decomposition level 1 of the panoramic images relating to the various planned isocurves , standard deviation maps DV of said coefficients are generated for each of said images II , IN (with N=5 in the case of the present example of the method, which provides for defining five isocurves) , summarized in a color map indicated as MAPI .

[0107] For the subsequent steps of the process , based on said map , the coefficients of the detail coefficient matrices of all the images associated with the highest standard deviation values are selected .

[0108] For levels two to four , the energy level for the coefficients of the matrices of the corresponding levels is calculated, as indicated with EL , also in this case for each of the panoramic images II to IN relating to the number of isocurves defined at Step 160 .

[0109] Various functions for determining the energy level are possible . In one exemplary embodiment , the energy level is calculated as follows : where : o)i is the detail coefficient at decomposition level n , and Enis said energy .

[0110] Color maps indicated as MAP>1 are generated also in this case .

[0111] Examples of these color maps and a map of the focus levels are shown in grayscale in Figures 21 and 22 .

[0112] As shown in Figure 20 .3 , using the new approximation coefficient matrices and detail decomposition coefficient matrices generated according to the fusion algorithm and applying the inverse wavelet transform, a panoramic image IP is reconstructed in which the contributions of the various panoramic images relating to the different planned isocurves are merged, and which relate to zones of the image and / or details represented in said images that present the best focus . Thus , said reconstructed panoramic image IP globally presents improved focus of the reproduced details across its entire extension .

[0113] Figure 23 shows an example of a panoramic image obtained from the fusion process according to the method of the present invention and according to the embodiments described above .

[0114] With reference to Figures 23. 1 to 23.3 , these show comparative examples between panoramic images obtained without applying the method of the present invention and the corresponding panoramic image obtained using the method of the present invention , called Best Focus in the figures .

[0115] The figures relate to parts of a panoramic image in which the differences are most evident to the naked eye . The areas where such differences appear most evident are delimited by closed lines surrounding the details that appear more detailed and in focus compared to known techniques when the method of the present invention has been used . As regards a system for implementing the method described above , considering the example of Figure 1 , the processing unit provided may comprise a memory in which an image processing program is stored, in which the instructions for a processor of said processing unit are coded, which , when executed, render said processing unit capable of performing the steps of the method according to the embodiments of the method described above .

[0116] Said processing unit may be integrated into the radiographic machine or may be a separate unit that communicates with a control unit of said machine .

Claims

CLAIMS1 . Method for generating panoramic radiographic images , comprising the following steps : a) selecting an acquisition trajectory defined by the path of rotation and / or roto-translation of a pair composed of an X-ray source and an X-ray detector which are positioned opposite to each other and move together along the said path of rotation and / or roto-translation around the patient ’ s head, which is positioned between the said X-ray source and the said X-ray detector during the entire course of rotation and / or roto-translation ; b) executing the said rotation and / or roto-translation with the simultaneous emission of X-rays towards the patient ’ s head and towards the opposing sensor and storing the raw image data generated by the said sensor ; c) defining at least two or more focusing layers and generating for each layer a corresponding panoramic image from the raw image data acquired through tomosynthesis of the image frames of the corresponding layer , which layers correspond to focusing curves related to a predetermined focal length and / or within a predetermined depth of focus relative to the said acquisition trajectory, also referred to as generating trajectory; d) identifying in the said images zones thereof , or sections thereof , in which predetermined details and / or predetermined sections of the dental arch are reproduced, and where the image presents the best focus with respect to the said details and / or the said image section of the additional layers of the said plurality; e) generating the panoramic image of the dental arch by combining the zones and / or the sections of the panoramic images of the two or more layers selected in step d) ;and wherein : step d) is performed using a wavelet decomposition algorithm on the said panoramic images relating to the said two or more layers , which provides for decomposition into at least two or more levels , of which one main level at higher resolution and at least one further level at lower resolution , generating for each level of decomposition a predetermined number of detail coefficient matrices and approximation coefficient matrices ;Step e) is performed by recombining the detail coefficients of the wavelet decomposition calculated for each of the said panoramic images of the said two or more layers for the said two or more levels and which present greater values for each of the said images .2 Method according to claim 1 , wherein for the wavelet decomposition and the wavelet reconstruction of the images it is provided a wavelet transformation algorithm called DT-CWT , or Dual Tree Complex Wavelet Transform.

3. Method according to claim 2 , wherein the said algorithm includes filters consisting of quasi -orthonormal bidirectional filters for the real and imaginary parts of the coefficient matrices related to the first decomposition level and which consist of filters of the type referred to as Q-shift filters for the further one or more decomposition levels , which are different for the real and imaginary parts of the coefficient of the coefficient matrices related to the said further decomposition levels .4 . Method according to one or more of the preceding claims , wherein step d) provides :- for the coefficient matrices of the wavelet decompositionat the first level relating to each of the said panoramic images respectively of the said two or more layers , the generation of standard deviation maps of the coefficients of the approximation coefficient matrices of the decomposition and the selection of one or more zones of the panoramic image relating to a corresponding layer in which the said standard deviation is greater compared to the other zones ;- for the coefficient matrices of the wavelet decomposition of the said further level or the said further levels , respectively for each level , calculating the energy of the coefficients of the corresponding coefficient matrix for each of the panoramic images relating to the said two or more layers and selecting the image zones of the panoramic images relating to the said two or more layers where the said decomposition coefficients exhibit the highest energy levels .5 . Method according to claim 4 , wherein the energy level is defined by the following equation :where :Wi is the detail coefficient at the decomposition level n and Enis the said energy .

6. Method according to one or more of the preceding claims , wherein step e) provides for performing a wavelet reconstruction step using as reconstruction coefficients of the wavelet functions for the said two or more decomposition levels the coefficients selected in step d) for each of the panoramic images relating to the said two or more layers .7 . Method according to one or more of the preceding claims , wherein prior to performing step d) , an additional step cl ) is performed, which consists of aligning the panoramic images along at least the said two or more layers relative to one of the said layers in terms of the magnification of the structures reproduced in the images of the different layers , while steps d) and e) are carried out with reference to the panoramic images of the said two or more layers aligned with each other in step cl ) .8 . Method according to claim 7 , wherein the said alignment involves performing an algorithm for modifying and standardizing at least the horizontal magnification , which algorithm provides for associating each image layer with a corresponding focusing curve of the panoramic image relating to the said layer for each panoramic image corresponding to a focusing curve , rescaling the value of the horizontal magnification so that it is identical to the vertical magnification intrinsic to the physics of the acquisition trajectory and / or the vertical magnification of a generating focusing curve and / or of the scanning trajectory, that is , the generating trajectory defined in step a) .

9. Method according to claims 7 or 8 , wherein the alignment between the panoramic images reconstructed through the tomosynthesis process referred to as shift & add is performed by defining a first focusing point shift curve during the common roto-translation of the source and the X-ray detector referred to as generating isocurve and determining the corresponding shift & add vector ; defining at least one or a predetermined number of additional so-called isocurves related to focusing point paths parallel and adjacent to the generating isocurve and at differentdistances from the said generating isocurve and determining for each of the said isocurves the corresponding shift & add reconstruction vector of the corresponding panoramic image ; generating a cumulative vector from the shift & add vectors based on the reference shift & add vector and the reconstruction shift & add vector and using a shrink algorithm to calculate the value of an interpolation function between the said cumulative reference shift & add vector and the said cumulative reconstruction shift & add vector for integer numbers of the argument of the function corresponding to the indices of the final image columns , while the values calculated by the algorithm define the indices of the columns of the initial image , which are redistributed in the final image according to the order defined by the said indices of the final image that is the integer values of the interpolation function argument , with the said steps being performed for each of the panoramic images reconstructed with reference to each of the isocurves .10 . Radiographic system for implementing the method according to one or more of the preceding claims 1 to 9 , the said system comprising : a support structure for cantilevered support of an X-ray source and an X-ray detector , which are carried by an arm or a combination of arms in a relative position where the X-ray source is opposite the X-ray detector , with the emission side of the source oriented towards the reception side of the detector , while an examination position for the patient' s head is provided, optionally defined by positioning devices , which position is interposed between the said X-ray source and the said X-ray detector ; at least one processing unit for processing the signals acquired by the X-ray detector , in which processing unit isloaded or loadable an image processing program comprising instructions to enable the said processing unit to perform the steps of the method according to one or more of the preceding claims 1 to 9 .11 . System according to claim 10 , wherein the said processing unit comprises : a unit for defining the movement trajectories of the X-ray source and the X-ray detector and for controlling the movement of the said arm that carries the said X-ray source and the said X-ray detector along the said trajectory; a unit for reconstructing panoramic images from the image data relating to the signals acquired by the said X-ray detector , and which reconstruction unit comprises an input interface for the definition parameters of two or more layers corresponding to a respective focusing curve and performs the steps of determining the image frames relating to the said two or more layers and tomosynthesis of the said frames to generate a panoramic image corresponding to each of the said two or more layers ; a unit for defining the focusing quality of the structures reproduced in the various zones and / or the various sections of the said panoramic images and for selecting the said zones and / or sections having the best focus ; a unit for merging the said zones and / or sections of the said panoramic images along the said two or more layers , generating a final panoramic image having the best focus along its entire extension , and wherein : the unit for defining the focusing quality of the structures reproduced in the various zones and / or sections of the said panoramic images and for selecting the said zones and / or sections having the best focus comprises a waveletdecomposition section of the said panoramic images relating to the said two or more layers , in which wavelet decomposition section a wavelet decomposition program is loaded or loadable , in which the instructions for the execution of the said wavelet decomposition and the settings of the criteria for the identification and selection of the coefficients of one or more decomposition levels corresponding to the zones and / or sections of the panoramic images having the best focus are coded, and which wavelet decomposition section executes the said program, whereby it is capable of performing the said wavelet decomposition step and the said step of identifying and selecting the coefficients of one or more decomposition levels corresponding to the zones and / or sections of the panoramic images having the best focus ; the said unit for merging the said zones and / or sections of the said panoramic images along the said two or more layers comprising a merging section and a wavelet recombination section , in which section a merging and / or wavelet reconstruction program is loaded or loadable , in which the instructions for reconstructing a panoramic image comprising the said zones or sections of the panoramic images relating to the said two or more layers corresponding to the wavelet decomposition coefficients that meet the said criteria for the identification and selection of the said wavelet decomposition coefficients are coded .12 . System according to one of claims 10 or 11 , wherein the processing unit further comprises a unit for aligning the panoramic images relating to the said two or more layers , in which alignment unit an alignment program is loaded or loadable , which comprises the instructions that render the said alignment unit capable of executing a rescaling and standardization algorithm for at least the horizontalmagnification , and in particular an algorithm that performs a modification of the images relating to the said two or more layers according to the method steps described above .

Images