FR3009109A1  Method for producing medical assisted medical images  Google Patents
Method for producing medical assisted medical images Download PDFInfo
 Publication number
 FR3009109A1 FR3009109A1 FR1357345A FR1357345A FR3009109A1 FR 3009109 A1 FR3009109 A1 FR 3009109A1 FR 1357345 A FR1357345 A FR 1357345A FR 1357345 A FR1357345 A FR 1357345A FR 3009109 A1 FR3009109 A1 FR 3009109A1
 Authority
 FR
 France
 Prior art keywords
 value
 voxel
 ki
 voxels
 corresponding
 Prior art date
 Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
 Pending
Links
Classifications

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
 G06T7/00—Image analysis
 G06T7/10—Segmentation; Edge detection
 G06T7/162—Segmentation; Edge detection involving graphbased methods

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
 G06T7/00—Image analysis
 G06T7/10—Segmentation; Edge detection
 G06T7/11—Regionbased segmentation

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
 G06T7/00—Image analysis
 G06T7/10—Segmentation; Edge detection
 G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
 G06T2200/00—Indexing scheme for image data processing or generation, in general
 G06T2200/04—Indexing scheme for image data processing or generation, in general involving 3D image data

 G—PHYSICS
 G06—COMPUTING; 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; 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
 G06T2207/10088—Magnetic resonance imaging [MRI]

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
 G06T2207/00—Indexing scheme for image analysis or image enhancement
 G06T2207/20—Special algorithmic details
 G06T2207/20072—Graphbased image processing

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
 G06T2207/00—Indexing scheme for image analysis or image enhancement
 G06T2207/20—Special algorithmic details
 G06T2207/20092—Interactive image processing based on input by user
 G06T2207/20104—Interactive definition of region of interest [ROI]

 G—PHYSICS
 G06—COMPUTING; 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
 G06T2207/30004—Biomedical image processing

 G—PHYSICS
 G06—COMPUTING; CALCULATING; COUNTING
 G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
 G06T2210/00—Indexing scheme for image generation or computer graphics
 G06T2210/41—Medical
Abstract
Description
The present invention relates to the technical field of obtaining and processing computerassisted medical images, in particular images of a scanner or an MRI (Magnetic Resonance Imaging) device. Computed tomography (CT), also called computed tomography (CT), computed tomography (CT), CATcomputer assisted tomography (CT), is a medical imaging technique that consists of to measure the Xray absorption generated by a scanner in different areas of a body, then, by computer processing, to digitize and finally to reconstruct 2D (twodimensional) or 3D images of the anatomical structures.
MRI is another medical imaging technique, which is based on the principle of nuclear magnetic resonance (NMR). Also known are complementary techniques of medical imaging aimed, from 2D and / or 3D medical images, to perform a segmentation of the image, so as to identify and segregate an area of interest, corresponding for example to a organ, an organ part, or a lesion, so as to be able to target this area, display one or more corresponding digital images, and facilitate observation by a medical operator. It is often necessary to be able to perform such segregation in a very short time, for example a few seconds at the most so as to minimize the time spent by the radiologist or the operator. These techniques therefore require significant computing power typically by means of a computer. To reduce this calculation time, or alternatively to improve the quality of the segmentation, one seeks fast and / or efficient calculation methods making it possible to reduce the calculation time for a given segmentation quality, or to improve the quality of segmentation for a given calculation time. Some of these complementary techniques use a particular iterative algorithm, called the Dijkstra algorithm, the name of its inventor, or simply dijkstra, to define an area of the image increasing with the number of iterations, from an origin point O chosen by the operator. The current techniques using a dijkstra to achieve, after a certain number of iterations, an image portion corresponding to the desired segmentation, however, are relatively inefficient for determining the appropriate iteration threshold for obtaining a complete image the area of interest, for example a lesion, without this image encompassing other image areas located outside the area of interest. The object of the invention is in particular to improve this determination of the number of iterations that is adequate to obtain a desired segmentation of an image, while requiring a limited calculation time. For this purpose, the subject of the invention is in particular a method for producing at least one piece of information relating to a particular area of at least one part of a body (2) from digital scanning or imaging data. by magnetic resonance, said RM RM of this at least part of the body, wherein:  this at least part of the body (2) being disposed within a real volume VR, is associated with this real volume VR a volume VN numeral formed of voxels, each voxel of VN corresponding to an elementary real volume of VR,  each voxel is associated with a numerical value correlated with a physical effect exerted by a scanner or MRI device (6) on the material contained in the elementary real volume corresponding to this voxel, characterized in that the following steps are performed by means of a computer:  a strictly positive parameter (for example an integer or a positive real number) is defined for each voxel ), said local score, correlated to the numerical value associated with this voxel, the numerical values associated with the voxels corresponding to the particular body area being lower on average than the numerical values associated with the voxels corresponding to at least a portion of the adjacent body areas to the particular body area; a subset EO of VN is defined, said origin space, for example chosen by a medical operator, and comprising at least one voxel; a cost of a voxel Oi of the origin space EO is defined to a given voxel A along a determined path of neighbor voxels two by two joining Oi to A, as the sum of the local scores of the voxels of this path , without counting the score of Oi, each local score of a voxel of the path different from Oi being multiplied by a geometric coefficient, this coefficient being for example correlated to a relative position between this voxel of the path and the preceding neighbor voxel of this path ; a cost of each given voxel A is defined as the minimum cost among all the paths joining any voxel of the EO origin space and the voxel A; by successive iterations of a first iterative process, an arithmetic sequence of sets Ei of voxels of VN is defined, in which E0 is the origin space E0, and for any integer i included in an interval [1, M], where M is> 1, each set Ei is the set of voxels whose cost is less than or equal to a predetermined cost Ci, Ci being a predetermined arithmetic sequence of positive values (integers or positive real numbers) strictly increasing according to i; for each value of iz 1, we define a set called a shell of rank i, denoted by Ki, corresponding to the set of voxels belonging to Ei and not belonging to E (i 1), the shell K0 being defined as identical to the EO origin space; for each shell Ki, with i 0, we define an outer edge BKi of this shell, formed by all the voxels of Ei which are adjacent to a voxel not belonging to Ei; for each value of iz 1, a quantity Gi is determined which is correlated to a thickness of the shell Ki, for example to the maximum local thickness EPLMi or to an average thickness EPmi, the maximum local thickness EPLMi being the maximum value for the voxels of BKi, the smallest distance between a given voxel of BKi and any voxel of E (i1), and the average thickness EPmi being the average value of the local thickness, for the different voxels from BKi; a first segmentation, said target segmentation of the volume VN corresponding to the choice of a set Ec, is determined, automatically or at the request of a medical operator, with C being the integer É M, the computer determining the value of C in conjunction with the finding of a predetermined change in the value of Gi, for example obtaining a value of EPLMi, or EPmi, or EPLMi / EPLM (i1), or EPmi / EPm (i1) below a predetermined threshold; optionally, complementary sought medical information relating to the target segmentation, which is displayed and / or stored on a physical medium and / or broadcast on a telecommunication network, automatically or on request of a medical operator. The complementary soughtafter medical information relating to the target segmentation may notably belong to the group formed by a maximum dimension, a volume, a density, a surfacetovolume ratio, a volumetosurface ratio, and a statistical element, for example an average value. or a standard deviation of a medical quantity. It is already known in the state of the art of iterative processes making it possible to generate subsets of increasing dimension of a 2D digital space, composed of pixels, or of a 3D digital space, composed of voxels, each subset being included in the subsets corresponding to a higher number of iterations. Such processes generally use a dijkstra, i.e., an iterative process of defining, in a numerical space, a series of nested subsets into each other using the notion of increasing cost of progression in the sense of large. Each elementary iteration of the diskstra typically corresponds to the acquisition of an additional point in the digital space. Typically, the corresponding iterative process is stopped after a predetermined number of elementary iterations, or else an objective of progression of this dijkstra is identified, for example when the cost of a pixel or voxel reached at the last elementary iteration reaches a predetermined threshold. However, according to the invention, a very different approach is used, in which there is no interest in the level of the costs obtained by the voxels during the last elementary iteration. We determine, according to the first iterative process of the global iterations according to the index i, each global iteration between the indices i and i + 1, typically grouping together a large number of elementary iterations of the dijkstra: Typically, the costs increase by a constant value between two global iterations (delta cost [C (i + 1)  Ci] constant for the definition of spaces Ei and E (i + 1)). But according to the invention, one is interested in the speed of propagation of the dijkstra (and in that of the cost), which corresponds to the observation of a predetermined evolution of the value of a magnitude Gi which is correlated to a thickness of the shell Ki, for example the maximum local thickness EPLMi or the average thickness EPmi of the shell Ki corresponding to the new voxels obtained during the iteration i: The more the cost progresses weakly at each elemental iteration of the diskstra, and the greater the the thickness of the shell corresponding to the overall iteration considered is important. Indeed, a small elementary progression of the cost implies that a large number of elementary iterations will be necessary to increase the cost of a given value [C (i + 1)  Ci] (corresponding to a global iteration), which implies that the corresponding shell will include a high number of voxels). This last piece of information is used by the invention because it reflects the fact that one has left, or is moving out of the particular area of the desired body, with moderate local scores.
This approach is based on anatomical findings, in connection with the definition of scores attributed to voxels and with that of the iterative process: When the segmentation sought corresponds, typically, to an area affected by a pathology, for example a lesion, or corresponds in one organ, this zone has a certain homogeneity and corresponds to voxels whose local score is chosen to be relatively low compared to the local score of the voxels corresponding to adjacent body areas, formed by other organs and / or by tissues that have not been affected by the pathology. Thus, the iterative process can progress more quickly (at a lower cost) in such a relatively homogeneous area with moderate local score.
Conversely, during structural changes in the material traversed, the cost of progression increases rapidly, because of the higher local scores, and the number of elementary iterations corresponding to a global iteration (given delta cost 5) becomes low. A rapid progression of the cost at each elemental iteration of the dijkstra is thus reflected in a decrease in the thickness, local or average, of the shell Ki, formed by the new voxels obtained during the iteration, which makes it possible to use this information reflecting the fact that one has gone out, or is moving out of the particular area of the body sought. Conventionally, one defines the local thickness of the shell Ki in a voxel X of its outer edge BKi as the smallest value, denoted d (X, E (i1)), of X to any voxel of E (i 1). The maximum thickness EPLMi of the shell Ki is the largest of the local thicknesses, for all the X voxels of BKi, and the average thickness EPmi of the shell Ki is the mean local thickness, for all the voxels X of BKi. According to a first embodiment of the method, the size Gi is correlated with the average thickness EPmi of the shell Ki, for example is a function of the volume / surface ratio of the outer edge of the shell Ki, in particular is equal to this ratio volume / surface of the outer edge, the volume of a shell Ki being defined in connection with the number of voxels of this shell and the surface of the outer edge being defined in connection with the number of voxels of this outer edge BKi. For example, if it is found that in a first range of values of i, ie INT1, the value of Gi evolves in an interval of values VG of determined amplitude DeltaG around a mean value Gm, then that for one or more values of i greater than the values of INT1, the value of Gi goes out of the VG value range, and deviates from the average value Gm by a difference value VE greater than a determined threshold, optionally correlated with DeltaG, a value of C is determined so that for i belonging to the interval [1, C], the value of Gi does not deviate from the mean value Gm by a value at least equal to the difference value at most two values of i, and preferably at most one value of i. This can be applied especially when Gi is the average thickness EPmi of the shell Ki.
According to a second embodiment of the method according to the invention, the magnitude Gi is correlated with the maximum local thickness EPLMi of the shell Ki, for example is equal to or proportional to EPLMi, and C is determined as the value i for obtaining a value of the maximum local thickness EPLMi or an EPLMi / EPLM ratio (i1) below a predetermined threshold.
The threshold can be predetermined in connection with the local thickness of the shell: If the cost of progression of the djkstra becomes too important, and correlatively the maximum local thickness of the hull becomes for the first time too weak for one  6 can still be in the particular body area, for example a lesion, this means that we have obtained all the particular area sought, and that we can determine that the last iteration makes it possible to obtain the desired target segmentation .
It is also advantageous to carry out a step of calculating the maximum local thickness EPLMi of the shell Ki, by means of a second iterative process relating to a parameter D, of updatable value and not decreasing as a function of the number of iterations, in which: one chooses a starting voxel X0 of BKi, calculates the smallest distance d (X0, E (i1)) between X0 and any voxel Z of the outer edge BK (i1) of the shell K ( i1), that is to say the local thickness of Ki at voxel X0, that is to say Do, and one initializes the value of D to Do; at each iteration, the local thickness of Ki is computed or examined in a new voxel of BKi, different from a voxel for which the local thickness of Ki has already been calculated or examined, and a present value of D is determined. which is the identified maximum local thickness of the shell Ki during the iteration considered and previous iterations; one of said iterations is carried out at least one and preferably a plurality of particular iterations, socalled iterations with examination, such that for each of them:  one starts from a voxel X for which the local thickness d (X, E (i1)) of the shell Ki has already been calculated, and is less than the present value of D,  a new voxel Y of BKi, preferably contiguous with X0, is chosen, and the maximum possible local thickness of the hull Ki to the voxel Y is examined to determine if it can be greater than the present value of D, by calculating the sum S = d (X, E (i1)) + d ( X, Y), then:  If S is greater than the present value of D, the local thickness of Ki is computed at voxel Y, and if the value obtained DY is greater than D, D is updated to DY value. If, on the other hand, the value of S obtained is less than or equal to D, the present value of D is not changed, and the local thickness of Ki is not calculated at voxel Y. When, for each of the voxels of BKi, the local thickness has been calculated, and / or the maximum possible local thickness of Ki in one of the said iterations with examination is evaluated, the final updated value of the parameter D is assigned to the maximum local thickness EPLMi of Ki.
This first method makes it possible to reduce the calculation time, as explained below. It is also possible to carry out a step of calculation by iterations of the local thickness of the shell Ki in a given voxel To of BKi, in which: an overlap of BK (i1) is carried out by a plurality of subsets SEj , possibly having two by two of the common intersections, each of these subsets SEj being formed by voxels, and preferably all the voxels of BK (i1) located at a maximum distance Rj of a voxel of SEj, says anchor of SEj, and noted ANj; the distances of To are calculated for each of the anchors ANj, and the one whose distance to T0 is minimal is determined, ie AN * corresponding to a subset SE *; the minimum distance d (To, SE *) of To is calculated to any voxel of SE *, that is d *; the value of a distance parameter d which is updatable and not increasing is initialized as a function of the number of iterations of a third iterative process; the third iterative process is carried out in which each of the subsets SEj different from SE * is successively examined, by evaluating the value of the difference Diff = d (To, ANj) Rj. If Diff is smaller than the value of the parameter d, we compute the smallest distance from To to any voxel of SEj, that is d (To, SEj), and if the value obtained dj is less than d, we update d to the value dj. If, on the contrary, the value obtained dj is greater than or equal to d, d is not updated. If the difference Diff is greater than or equal to the value of the parameter d, we do not perform the computation of d (To, SEj). In all cases, we continue the third iterative process until the examination of all sets SEj; the final value of the parameter d is then assigned to the local thickness of the shell Ki at the determined voxel To of BKi, that is to say at the minimum distance d (To, E (i1)). This second method also makes it possible to reduce the calculation time, as explained below. It will be understood that the first, second, and third iterative processes are not successive processes, but can be nested. In particular, the second and / or the third iterative process can be embedded in steps of the first iterative process, and for example be executed one or more times at each step of this first iterative process. Advantageously, the first iterative process is stopped for the determination of the sets Ei when the number of iterations reaches the value N = C corresponding to the target segmentation EC, or else the value N = C + A corresponding to the set E (C + A), wherein A is a predetermined positive integer.
Thus, as soon as the target segmentation is reached, the iterative calculation is stopped, or a number A of additional iterations is continued so that, for example, a medical operator can decide to enlarge the zone of image covered by the target segmentation that covers the particular body area, for example, in the case where that particular body area is a lesion, to see if the neighboring areas do not include small separate lesions, or an extension of the main lesion, or simply to observe the peripheral areas of the lesion. The relatively fast shutdown of the iterative process provides a significant saving in computation time, especially when the particular body area is relatively limited.
According to an optional variant of the method according to the invention (or also of another method for producing at least one information relating to a particular zone of at least a part of a body): digital volume VN by eight portions of digital volume PVj corresponding to portions of the actual volume delimited by three mutually perpendicular planes, of which an axial (or transverse) plane XY, a sagittal plane YZ, and a coronal plane XZ; first iterative process in parallel in each of the digital volume portions PVj to determine sets Ei, j, Ki, j, BKi, j and the evaluation of a magnitude Gi, j by calculating in parallel these different elements in each of the portions of volume PEj as respectively Ei, Ki, BKi and Gi in each of these volume portions PVj; When a predetermined evolution of one or a plurality of the magnitudes Gi, j is observed, a set of frontier voxels of order 1, which are the frontier voxels belonging to a set Ei, j of a portion of digital volume PVj and which do not belong to a set Ei, k of another portion of digital volume PVk, with k different from j. ; at least one additional iteration is carried out according to the first iterative process in the digital volume VN starting from the set of first order frontier voxels as the origin space, each voxel of this first set retaining its local cost to obtain at least one minus a complementary portion of ECO digital space portion. This ECO space can thus be integrated with the target segmentation, or, where appropriate, with the segmentation of interest, by merging with the spaces Ei, j. Typically, the iterative process is continued until the same maximum cost is obtained as for the sets. To determine the three perpendicular planes two to two above (axial, sagittal, and coronal) we can choose their common point at a point corresponding preferably to a voxel of the original space EO, or to a voxel obtained by calculating the center of gravity. voxels from the EO origin space (assigned an identical weight). This characteristic variant of the invention also makes it possible to obtain a significant gain in the calculation time: in fact, it is possible to carry out in parallel a large part of the iterative process by means of preferably several microprocessors or microprocessors having several cores, or multiple computers. After merging the sets Ei, j obtained, the additional step (s) make it possible to obtain a significant additional progression of the iterative process, to determine complementary zones of the particular zone sought, with a number of iterations. moderate in the global space, and a limited calculation time, as will be explained later. This variant can be implemented with other methods of producing at least one piece of information relating to a particular area of at least a part of a body. In this case, the digital volume VN is recovered by eight portions of digital volume PVj corresponding to portions of the real volume delimited by three mutually perpendicular planes, of which an axial (or transverse) plane XY, a sagittal plane YZ, and a coronal plane XZ, we define a first iterative process of construction of an arithmetic sequence of sets Ei, nested in each other, to cover increasing zones corresponding to the particular zone of the body, from the selection from an origin space E0, the first iterative process is carried out in parallel in each of the digital volume portions PVj to determine the sets Ei, j, by calculating in parallel these different sets in each of the volume portions PEj as Ei in each of these PVj volume portions; From a number of iterations that one chooses, one determines a set of frontier voxels of order 1, which are the border voxels belonging to a set Ei, j of a portion of digital volume PVj and which n do not belong to a set Ei, k of another portion of digital volume PVk, with k different from j. ; at least one additional iteration is carried out according to the first iterative process in the digital volume VN starting from the set of first order frontier voxels as the origin space, to obtain at least one complementary digital space portion ECO . This ECO space can thus be merged with the spaces Ei, j to obtain a target segmentation, or, where appropriate, a segmentation of interest. Advantageously, it is determined and displayed on a screen, automatically or at the request of a medical operator, and optionally produced on a film support, at least one socalled primary digital image obtained from Ec, for example a 2D image corresponding to a axial or sagittal or coronal section of a body part corresponding to Ec, or a 3D view of a body part corresponding to Ec.
Preferably, at least one primary digital image 2D corresponding to an axial, sagittal or coronal section of a part of the body is automatically determined and displayed, the image corresponding to the segmentation Ec. If, for example, a medical operator wishes to visualize a segmentation different from that of the target segmentation, it is then possible, after displaying this primary digital image, to display another digital image corresponding to the same part of the body as that of the primary image, this other image, called secondary image, corresponding to a set E (c + p) said segmentation of interest, P being a positive or negative integer, the computer realizing if necessary one or more complementary iterations, according to the first iterative process, to reach the rank of iteration C + P. Typically, at least one memory of the computer and / or broadcast on a telecommunications network is stored at least one medical information, for example, data corresponding to the sets Ei, and / or data corresponding to the target segmentation and / or or data corresponding to the segmentation of interest. It will be understood that the term "display" means a display of a digital image on a screen. It is irrelevant whether there is a single screen, or several different screens, to display such (these) or such (s) digital image (s) in the implementation of the invention.
The invention also relates to a computer program characterized in that it comprises program code instructions for the execution of the aforementioned process steps which are performed by means of a computer. The invention also relates to a CT scan or computerassisted MRI method of at least a part of a body, making it possible to visualize a particular zone of this at least part of the body, characterized in that it determines and displays at least one primary digital image of the target segmentation obtained by the aforementioned method, and / or a secondary image of the segmentation of interest obtained by the aforementioned method. The scanning or MRI method may comprise at least the following steps:  one or more digital image (s) of CT scan or MRI of at least part of the body, said to be determined and displayed; initial image (s), preferably one or more 2D image (s);  by means of digital or physical pointing means, one or more pixels of this or these initial image (s) are selected, to which (s) are associated one or more voxels of VN, which together define the origin space E0; the steps of the aforementioned method are carried out in order to determine and display the primary image; a variation P is determined of the number of iterations of the basic iterative process, corresponding to a set E (cFp) said segmentation of interest, P being a positive or negative integer, less than C in absolute value if P is negative with adjustable means for changing the number of iterations; if necessary, one defines by the computer one or more sets Ei by continuing the basic iterative process so as to be able to define the set E (Crp)  one displays a secondary image corresponding to this segmentation of interest E ( c + p)  It is particularly important for a medical operator to be able to acquire quickly and substantially in real time the image of a particular body area sought, typically a CT or MRI image, with a process assisted by computer. It is also important to be able to benefit from the personal expertise of the medical operator to define an area of interest, which may be different from the particular body area, precisely because of the expertise of the operator . However, it is difficult to combine human expertise and IT resources in real time. However, the Applicant has found a CT or MRI method, as well as a CT or MRI device to combine in a particularly effective and ergonomic way human expertise on the one hand and other computer means on the one hand, for determining, typically in real time, a particular body area searched for a body, for example a lesion or an organ, and producing targeted images of that particular body area searched for. This method of scanning or of ergonomic MRI can be implemented with any method of producing an information and / or a digital image relating to a particular area of the body from digital scanner data or from MRI, especially to visualize this particular area of the body, and therefore not only with the aforementioned method. In such a method, the at least part of the body being disposed within a real volume VR, this real volume VR is associated with a digital volume VN formed of voxels, each voxel of VN corresponding to an elementary real volume of VR, and each voxel is associated with a numerical value correlated with a physical effect exerted by a CT or MRI device on the material contained in the elementary real volume corresponding to this voxel. Then, using a computer, at least the following steps are performed: § one or more digital scanographic or MRI image (s) are produced and displayed from an axial section of the body to at least one level longitudinal of this body, said initial 2D image (s); § one or more pixels of this or these initial 2D image (s) are defined by means of digital or physical pointing means, to which (s) are associated one or more voxels of VN which together define an origin space E0; § from this source space E0, a basic iterative process implemented by means of a computer, calculates a target segmentation Ec corresponding substantially to a zone covering a particular body area searched for the body, corresponding to a number of C iterations; § at least one image of the target segmentation Ec is displayed; § determining a variation in the number of iterations of the basic iterative process, corresponding to a set E (c + p) said segmentation of interest, P being a positive or negative integer, less than C in absolute value if P is negative with adjustable means for changing the number of iterations; § if necessary, using the computer one or more sets Ei is defined by continuing the basic iterative process so as to be able to define a set E (c + p) § at least one image corresponding to this segmentation of interest is displayed E (c + p) The invention also relates to a set of CT or MRI comprising a CT or MRI device and a computer, separate or integrated in this device, characterized in that this facility is able to implement the abovedescribed CT or MRI procedure. In particular, the computer typically comprises a physical medium on which is recorded a computer program as defined above. This device may notably comprise means for displaying on a screen a 2D digital image corresponding to an axial section of a part of the body at a longitudinal level of the body, the image corresponding to the target segmentation EC, or segmentation of interest. It may also include means for modifying the number of iterations to obtain a segmentation of interest, corresponding to a set E (c + p), where P is a positive or negative integer, the computer realizing, if necessary, complementary iterations according to the first iterative process making it possible to reach the iteration rank C + P, and / or the display means of another 2D digital image corresponding to an axial section of the same part of the body, at the same determined longitudinal level of this body, this other image corresponding to the segmentation of interest. The adjustable means for modifying the number of iterations are, for example, means belonging to the group formed by a physical pointer with a click (s) and / or a knurling key, a device with a rotary or rotary ball, a tactile device, by example with a touch screen or touch pad, an optical pointer or an optical pen, a tilting pointing device, and a graphics tablet. The adjustable means can also be nonphysical computing means, sometimes called "virtual", such as for example a digital pointer, etc. The invention finally relates to a CT or MRI image on a physical film medium, characterized in that it was produced from the target segmentation determined during the execution of the aforementioned method, or from the aforementioned segmentation of interest. The invention will be better understood on reading the appended figures, which are provided by way of example and are in no way limiting, in which the figures show schematically or diagrammatically steps, optionally optional, of the method according to the invention applied to digital computed tomography data of a part of a body: Figure 1 illustrates a method of obtaining a target segmentation showing the progression of a first iterative process. Figure 2 shows the evolution of a magnitude Gi corresponding to the average thickness of a shell in the first iterative process. FIG. 3 schematically represents a shell Ki, and a method for determining the maximum local thickness of this shell. Figure 4 schematically illustrates a method of iterative calculation of the local thickness of a shell Ki.
Figure 5 schematically represents a method for obtaining a target segmentation using a partition of the digital space. Figure 6 schematically shows a set of CT or MRI 14 according to the invention. In what follows, and for the sake of clarity, we will represent various 3D elements (in three dimensions) of the digital space by a twodimensional image in the figures. Thus, a voxel will be represented by a point, or an elementary square in Figure 5, a shell by a surface, the edge of a shell by a curve, and so on. Referring now to FIG. 1. FIG. 1 very schematically shows a modified CT image, for example an image of a portion of a body in an axial plane, showing a lesion whose contour has been represented. 2 in bold, the other parts of the CT image are not shown. The part of the body being disposed inside a real volume VR, this real volume VR is associated with a digital volume VN formed of voxels, each voxel of VN corresponding to an elementary real volume of VR, and associated with each voxel a numerical value correlates a physical effect exerted by the scanning device on the material contained in the elementary real volume corresponding to this voxel. We then implement a first iterative process, diskstra type, as follows:  for each voxel is defined a strictly positive parameter correlated to the numerical value associated with this voxel, the numerical values associated with the voxels corresponding to the corresponding particular body area the lesion being smaller on average than the numerical values associated with the voxels corresponding to at least a portion of body areas adjacent to the particular body area (lesion). The medical operator, in view of a nontransformed CT image, then identifies the existence of a lesion, and chooses, by numerical score, a voxel O of this lesion, forming the EO origin space, which is also the first, E0, numerical spaces defined by the first iterative process. EO could also include several voxels, the operator choosing several voxels, at different points of the lesion. a cost of a voxel Oi of the origin space O is defined to a given voxel A along a determined path of neighbor voxels two by two joining Oi to A, as the sum of the local scores of the voxels of this path , without counting the score of Oi, each local score of a voxel of the path different from Oi being multiplied by a geometric coefficient, this coefficient being for example correlated to a relative position between this voxel of the path and the preceding neighbor voxel of this path . a cost of each given voxel A is defined as the minimum cost among all the paths joining any voxel of the origin space O and the voxel A. by successive iterations of a first iterative process, an arithmetic sequence of sets Ei of voxels of VN is defined, in which EO is the origin space O, and for any integer i included in an interval [1, M], with M being> 1, each set Ei is the set of voxels whose cost is less than or equal to a predetermined cost Ci, Ci being a predetermined arithmetic sequence of strictly increasing positive values as a function of i. for each value of iz 1, a set called hull of rank i, denoted by Ki, corresponding to the set of voxels belonging to Ei and not belonging to E (i 1), is defined, the shell KO being defined as identical to the origin space O.  for each shell Ki, with i 1, is defined an outer edge surface BKi of this shell, formed by all the voxels of Ei which are adjacent to a nonvoxel voxel not to Ei and the outer edge BK0 is defined as the voxel O. In FIG. 1, each curve represents the limit reached by a set Ei, during one of the successive iterations. Each curve also represents the outer or inner edge of a shell, a shell being defined by the surface between two successive curves. Thus, the curves 4, 6, 8, 10, 12, 14 respectively correspond to the limits of the sets E1, E2, E3, E4, E5, E6, and represent the outer edges BK1, BK2, BK3, BK4, BK5, BK6 of shells K1, K2, K3, K4, K5, and K6, represented by references 16, 18, 20, 22, 24, and 26. Typically, each iteration of the first iterative process comprises a plurality of elementary iterations. It can be seen that when one is in a zone corresponding to the interior of the lesion, the shells are relatively thick, with medium and maximum thicknesses important. On the contrary, when one is outside the limits of the lesion, the thicknesses of the shells are weaker. Thus, in FIG. 1, the maximum thicknesses 28, 30, 32 of the shells K2, K3, K5 located at least partly in the lesion are greater than the maximum thickness 34 of the shell K8, referenced 36, between curves 38 and 40, this shell being located outside the lesion. This results from the fact that the scores of the voxels outside the lesion are relatively high, which increases the cost of progression at each iteration. For a given cost difference corresponding to two sets Ei and Ei + 1, fewer elementary iterations are made, and the thickness of the shell is smaller. This change in the thickness of the shells is exploited by the method according to the invention to determine the target segmentation. It can also be seen in Figure 1 that some shells 22, 24, 26, and the seventh shell 38 are partly inside and partly outside the lesion. It is desirable that the target segmentation comprises all the voxels of the hulls covering the whole of the lesion, ie the set E7 comprising the origin point O and the set of shells K1 to K7. Typically, for each value of i z 1, a quantity Gi is determined which is correlated with a thickness of the shell Ki, for example a maximum local thickness EPLMi or an average thickness EPmi.
Conventionally, for i z 1, the local thickness of the shell Ki at a point X0 of the outer edge BKi of Ki is the minimum distance from X0 to a point of E (i1). The maximum local thickness EPLMi and the average thickness EPmi are respectively the maximum and average values of the local thickness. The distances are considered conventionally with respect to the central point of each voxel considered.  Then we realize, automatically or on request of a medical operator, a first segmentation, said target segmentation of the volume VN corresponding to the choice of a set Ec, with C integer 5 M, the computer determining autonomously the value of C in connection with the finding of a predetermined change in the value of Gi, for example obtaining a value of EPLMi, or EPmi, or EPLMi / EPLM (i1), or Epmi / EPm ( i1) below a predetermined threshold; optionally, additional searched information relating to the target segmentation, which is displayed and / or stored on a physical medium and / or broadcast on a telecommunication network, automatically or on request of a medical operator, is determined. Referring now to Figure 2 which shows the evolution of the value Gi according to the number of iterations, for example the change in the average thickness of the shell Ki. During the first iterations, the value of Gi evolves little around a mean value Gm. Then, the value of Gi decreases, typically when a portion of the shell is outside the lesion, which reduces the local thickness, the local progression of the dijkstra being slowed down by increased costs (increase in local scores ). The value of Gi ends up leaving the VG interval. The diskstra is then stopped, or the value C of the number of iterations retained for the target segmentation is determined when the value of Gi deviates from Gm by a value greater than a predetermined value of deviation VE. , in Figure 2, corresponds to the value of C = 7. One can also choose to keep one or two additional iterations, in order to more likely include any end of the lesion. Referring now to Figure 3, which shows a shell Ki comprising an outer edge BKi, and bounded by the outer edge BKi1 of the hull Ki1, which also delimits E (i1). The local thickness of Ki at a point X0 of its outer edge BKi is referenced 40, while the local thickness of Ki at point X is referenced 42. To perform a step of calculating the maximum local thickness EPLMi of the shell Ki is advantageously carried out as follows: by means of a second iterative process relating to a parameter D, of updatable value and not decreasing as a function of the number of iterations, in which: a starting Voxel X0 of BKi is chosen, calculates the smallest distance d (X0, E (i1)) between X0 and any voxel Z of the outer edge BK (i1) of the shell K (i1), i.e. the local thickness of Ki at voxel X0, say Do, and we initialize the value of D at Do; at each iteration, the local thickness of Ki is computed or examined in a new voxel of BKi, different from a voxel for which the local thickness of Ki has already been calculated or examined, and a present value of D is determined. which is the identified maximum local thickness of the shell Ki during the iteration considered and previous iterations; one of said iterations is carried out at least one and preferably a plurality of particular iterations, socalled iterations with examination, such that for each of them:  one starts from a voxel X for which the local thickness d (X, E (i1)) of the shell Ki has already been calculated, and is less than the present value of D,  a new voxel Y of BKi, preferably contiguous with XO, is chosen, and The maximum possible local thickness of the hull Ki to the voxel Y is examined to determine if it can be greater than the present value of D, by calculating the sum S = d (X, E (i1)) + d (X, Y), then: If S is greater than the present value of D, the local thickness of Ki is computed at voxel Y, and if the value obtained DY is greater than D, D is updated. to the value DY. If, on the other hand, the value of S obtained is less than or equal to D, the present value of D is not changed, and the local thickness of Ki is not calculated at voxel Y. voxels of BKi, the local thickness has been calculated, and / or the maximum possible local thickness of Ki in one of the said iterations with examination is evaluated, the final updated value of the parameter D is assigned to the maximum local thickness EPLMi of Ki. This method uses an inequality of the sides of a triangle: one of the sides can not have a dimension greater than the sum of the dimensions of the two other sides. Thus, if the sum of the distance from X to Y (easy to calculate) and the thickness of the shell Ki to the voxel X is less than or equal to the present value of D, the thickness of the shell at the point Y will be necessarily also less than or equal to D, which makes it unnecessary to calculate it, and thus reduces the calculation time. Thus, it is clear in FIG. 3 that if the present value of D is the local thickness 40 of Ki at voxel X0, and if the sum S = d (X, E (i1)) + d (X , Y), which is equal to the value 42 of the local thickness of the hull Ki at voxel X, increased by the distance d (X, Y) from voxel X to voxel Y, is less than or equal to D, local thickness of the hull Ki voxel Y can not be greater than D. Therefore, to calculate the maximum local thickness EPLMi of the hull Ki, it is useless to calculate the local thickness of the hull Ki at the point Y, this which reduces the calculation time.
Referring now to FIG. 4, it is advantageous to carry out a calculation step by iterations of the local thickness of the shell Ki into a determined voxel To of BKi as follows: a recovery of BK (i1) by a plurality of subsets SEj, possibly having two by two of the common intersections, each of these subsets SEj being formed by voxels, and preferably all the voxels of BK (i1) located at a maximum distance Rj a voxel of SEj, called anchor of SEj, and noted ANj; the distances of To are calculated for each of the anchors ANj, and the one whose distance to T0 is minimal is determined, ie AN * corresponding to a subset SE *; the minimum distance d (To, SE *) of To is calculated to any voxel of SE *, that is d *; the value of a distance parameter d which is updatable and not increasing is initialized as a function of the number of iterations of a third iterative process; the third iterative process is carried out in which each of the subsets SEj different from SE * is successively examined, by evaluating the value of the difference Diff = d (To, ANj) Rj. If Diff is smaller than the value of the parameter d, the calculation of the smallest distance of To is performed on any voxel of SEj, ie d (To, SEj), and if the value obtained dj is less than , we update d to the value dj. If, on the contrary, the value obtained dj is greater than or equal to d, d is not updated. If the difference Diff is greater than or equal to the value of the parameter d, we do not perform the computation of d (To, SEj). In all cases, we continue the third iterative process until the examination of all sets SEj; the final value of the parameter d is then assigned to the minimum distance d (To, E (i1)), that is to say to the local thickness of Ki in the voxel To. Here again, the inequality is used. sides of a triangle: one of the sides can not have a dimension greater than the sum of the dimensions of the other two sides: In Figure 4 are represented in particular different subassemblies each formed by a portion of curve (representing a portion of BK (i1) in 2D) delimited by a circle, around an anchor. The subset SE * is formed by the portion of curve included in the circle whose center is the AN * anchor. Also represented are the subsets SE2, whose voxels are in a radius R2 around the anchor AN2, SE4, whose voxels are in a radius R4 around the anchor AN4, and SEj, whose voxels are in a radius Rj around the anchor ANj. AN * is the anchor closest to To. The distance d (To, E (i1)), or local thickness of Ki in To, is identical to the minimal distance of To at edge BK (i1) . To evaluate this distance, we will be interested in the distances of To to the different subsets SEj of BK (i1). For each of the subsets SEj, excluding SE2, SE * and SE4, the smallest distance from this set to To will be at least equal to the difference Diff = d (To, ANj) Rj, by virtue of the inequality of the sides d 'a triangle. Indeed, the voxel of SEj closest to To is Uj. Now the distance from Uj to To can not be less than Diff = d (To, ANj)  Rj, otherwise we would have d (To, ANj)> d (To, Uj) + Rj, which is contrary to inequality of the triangle [ToANjUj]. Thus, it can be seen that the subsets SEj shown in FIG. 4, apart from the three subsets SE2, SE * and SE4, each have a value of Diff which is greater than d *. Therefore, when looking for the minimum distance from To to BK (i1), it is useless to calculate the minimum distance of To from each of these subsets, which will necessarily be greater than d *. This thus makes it possible to reduce the calculation time of the local thickness of the shell Ki to the voxel To of BKi. Referring now to Figure 5, which schematically shows a pixelated 2D image corresponding to a section of the digital volume VN, which has been previously cut into eight volume portions PVj, performing a VN overlay. Each portion of PVj volume corresponds to a portion of the actual volume cut in three planes perpendicular two by two including an axial plane (or transverse) XY, a sagittal plane YZ, and a coronal plane XZ. These XY, YZ, and XZ planes correspond to frontier digital planes of a voxel thickness, formed of border voxels typically belonging to two adjacent PVj digital volume portions, the intersection of the three digital planes corresponding to a central voxel which can advantageously belong to the EO origin space, or correspond to the centroid (voxel of mean coordinates) among the voxels of the EO origin space. In FIG. 5, corresponding to an axial section, perpendicular to the Z axis, there are therefore four PVj portions of the digital volume, ie 44, 46, 48, and 50 comprising, for two adjacent PVj portions, a border digital plane portion. common XZ or YZ, formed of border voxels. These four digital volume portions have in common a digital line portion 52.  The first iterative process is performed in parallel in each of the digital volume portions 44, 46, 48, 50, and the first iterative process is stopped when observes a predetermined evolution of one or a plurality of magnitudes Gi, j. This makes it possible to define sets E 1, j, ie respectively 54, 56, 58, 60. The first process also makes it possible to determine second order frontier voxels 62, which are frontier with at least two sets E 1, j different from the sets 54, 56, 58, and 60. An assembly formed by the first order frontier voxels 64, which belong to a set Ei, j of a portion of digital volume PVj, and do not belong to a set set Ei, k of another portion of digital volume PVk, with k different from j. At least one additional iteration is then carried out according to the first iterative process, in the digital volume VN, starting from the set of first order frontier voxels as the origin space, each voxel of this first set retaining, its local cost, for obtain at least one ECO complementary digital space portion assembly comprising border voxels. In FIG. 5, ECO thus comprises the assemblies 66, obtained starting from a border voxel of order 1. The method also typically comprises a step of merging sets obtained by the first iterative process. In particular, it is advantageous to carry out a first fusion of the sets Ei, j before carrying out the additional iteration. In this additional iteration, we can notably search in space VN voxels whose cost is the same as that of sets Ei, j. During this additional iteration, we do not take into account all the voxels resulting from this first fusion, which are already selected to be integrated in the target segmentation Ec. Then the ECO set obtained is fused with the set resulting from the first fusion, to form the target segmentation Ec. It can be seen in FIG. 5 that the one or more iterations make it possible to select and integrate in the target segmentation portions 66 of the lesion that would not have been selected if the first iterative process had been performed integrally in the different portions. PVj volume in parallel. Thus, it avoids a performance defect related to a complete process of the process with partition of the digital volume. This method of digital data processing CT or MRI, partially performed in parallel in space portions after partition of the global digital space, and partially in this global digital space VN, can be implemented very general, according to other methods of treatment of CT or MRI data, and not exclusively for the execution of the aforementioned method. Referring now to FIG. 6, which shows very schematically a patient 68 lying on a table 70. A scanning or MRI device 72 makes it possible to send digital information to a processor 76 connected to a screen 74. The set also includes a computer 78 comprising a physical medium 79 on which is recorded a computer program comprising the code instructions for performing the aforementioned method steps performed by computer. Computer 78 may communicate with a computer server to transfer information and / or digital data. The scanner or MRI device 72 may communicate with the processor 76 and the computer 78 via links 82, 84, and 86. The computer 78 may communicate with the processor 76 via links 88 and 90, and link 92 with the server 80, which can send information via a link 96. The communications can be of any type, for example by wired, optical, wireless, wifi mode or other wireless mode etc ... The set scan or MRI also includes a physical or digital pointer 98 allowing a medical operator to communicate with the computer 78 and receive feedback information relating to a segmentation of interest, for displaying an image of this segmentation of interest on the screen 74 or on a second screen 75. The invention is not limited to the embodiments shown and other embodiments will become apparent to those skilled in the art. It is possible in particular to insert or add additional process steps, for example steps to eliminate from the digital starting space digital artifacts that do not correspond to parts of the body, or a organ, for example from anatomical considerations, and / or eliminate the support table of the patient, and generally use any characteristic known in the field of CT or MRI.
Claims (20)
 REVENDICATIONS1. A method of producing at least one information relating to a particular area of at least a part of a body (68) from digital CT or magnetic resonance imaging data, said MRI of this at least part of body in which:  this at least part of the body (68) being disposed within a real volume VR, this real volume VR is associated with a digital volume VN formed of voxels, each voxel of VN corresponding to a volume elementary reality of VR,  each voxel is associated with a numerical value correlated with a physical effect exerted by a scanner or MRI device (72) on the material contained in the elementary real volume corresponding to this voxel, characterized in that the following steps are carried out by means of a computer (78):  for each voxel, a strictly positive parameter correlated with the numerical value associated with this voxel is defined, the numerical values associated with the x voxels corresponding to the particular body area being lower on average than the numerical values associated with the voxels corresponding to at least a portion of the body areas adjacent to the particular body area;  We define a subset EO (0) of VN, said origin space, for example chosen by a medical operator, and comprising at least one voxel; a cost of a voxel Oi of the origin space EO is defined to a given voxel A along a determined path of neighbor voxels two by two joining Oi to A, as the sum of the local scores of the voxels of this path , without counting the score of Oi, each local score of a voxel of the path different from Oi being multiplied by a geometric coefficient, this coefficient being for example correlated to a relative position between this voxel of the path and the preceding neighbor voxel of this path ; a cost of each given voxel A is defined as the minimum cost among all the paths joining any voxel of the EO origin space and the voxel A; by successive iterations of a first iterative process, an arithmetic sequence of sets Ei of voxels of VN is defined, in which E0 is the origin space E0, and for any integer i included in an interval [1, M], with M being> 1, each set Ei is the set of voxels whose cost is less than or equal to a predetermined cost Ci, Ci being a predetermined arithmetic sequence of strictly increasing positive values as a function of i; for each value of iz 1, we define a set called hull of rank i, denoted by Ki (16, 18, 20, 22, 24), corresponding to all the voxels belonging to Ei and not belonging to E (i 1), the shell KO being defined as identical to the original space EO; for each shell Ki (16, 18, 20, 22, 24) is defined, with iz 0, an outer edge BKi (4, 6, 8, 10, 12) of this shell, formed by all the voxels of Ei (16, 18, 20, 22, 24) which are adjacent to a voxel not belonging to Ei; for each value of i 1, a value Gi is determined which is correlated with a thickness of the shell Ki (16, 18, 20, 22, 24), for example a maximum local thickness EPLMi or an average thickness EPmi; a first segmentation, said target segmentation of the volume VN corresponding to the choice of a set Ec, is determined, automatically or at the request of a medical operator, with C being the integer É M, the computer determining the value of C in conjunction with the observation of a predetermined evolution of the value of Gi, for example obtaining a value of EPLMi, or EPmi, or EPLMi / EPLM (i1), or Epmi / EPm (i1) below a predetermined threshold; optionally, complementary sought medical information relating to the target segmentation, which is displayed and / or stored on a physical medium and / or broadcast on a telecommunication network, automatically or on request of a medical operator.
 2. The method as claimed in claim 1, in which the complementary sought medical information relating to the target segmentation is determined, this sought information belonging to the group formed by a maximum dimension, a volume, a density, a surfacetovolume ratio, a volume ratio. on a surface, and a statistical element, for example an average value or a standard deviation of a medical quantity.
 3. Method according to claim 1 or 2 wherein the volume of a shell Ki (16, 18, 20, 22, 24) is defined in relation to the number of voxels of this shell, and in which the size Gi is correlated. to the average thickness EPmi of the shell Ki (16, 18, 20, 22, 24), for example is a function of the volume / surface ratio of the outer edge of the shell Ki (16, 18, 20, 22 , 24), in particular is equal to this ratio volume / outer edge area.
 4. Method according to any one of claims 1 to 3 wherein, if it is found that in a first range of values of i, either INT1 the value of Gi evolves in a range of VG values of determined DeltaG amplitude around of a mean value Gm, then that for one or more values of i greater than the values of INT1, the value of Gi exits the range of values VG, and deviates from the average value Gm by a value of deviation VE greater than a determined threshold, optionally correlated with DeltaG, a value of C is determined so that for i belonging to the interval [1, C], the value of Gi does not deviate from the average value Gm of a value at least equal to the difference value for at most two values of i, and preferably at most one value of i.
 5. Method according to any one of claims 1, 2 and 4 wherein the magnitude Gi is correlated to the maximum local thickness EPLMi of the hull Ki (16, 18, 20, 22, 24), for example is equal or proportional to EPLMi, and C is determined as the value of i for obtaining a value of the maximum local thickness EPLMi or of a ratio EPLMi / EPLM (i1) below a predetermined threshold .
 6. Method according to any one of claims 1 to 5, wherein there is carried out a step of calculating the maximum local thickness EPLMi of the shell Ki (16, 18, 20, 22, 24), by means of a second iterative process relating to a parameter D, of updatable value and not decreasing as a function of the number of iterations, in which: a starting voxel X0 of BKi is chosen, calculates the smallest distance d (X0, E (i) 1)) between X0 and any voxel Z of the outer edge BK (i1) of the shell K (i1), i.e. the local thickness of Ki at voxel X0, that is, and we initialize the value of D to Do; at each iteration, the local thickness of Ki is computed or examined in a new voxel of BKi, different from a voxel for which the local thickness of Ki has already been calculated or examined, and a present value of D is determined. which is the identified maximum local thickness of the shell Ki during the iteration considered and previous iterations; one of the said iterations is carried out at least one and preferably a plurality of particular iterations, called iterations. with examination, such that for each of them:  We start from a voxel X for which the local thickness d (X, E (i1)) of the shell Ki has already been calculated, and is less than the discounted value of D,  A new voxel Y of BKi, preferably contiguous to X0, is chosen and the maximum possible local thickness of shell Ki is examined in voxel Y to determine if it can be greater than to the present value of D, by calculating the sum S = d (X, E (i1 )) + d (X, Y), then:  If S is greater than the present value of D, we calculate the local thickness of Ki at voxel Y, and if the value obtained DY is greater than D, we update D to the value DY. If, on the other hand, the value of S obtained is less than or equal to D, the present value of D is not changed, and the local thickness of Ki is not calculated at voxel Y. when for each of the voxels of BKi the local thickness has been calculated, and / or the maximum possible local thickness of Ki in one of said examination iterations has been calculated, the final updated value of the parameter D is assigned to the maximum local thickness EPLMi of Ki.
 7. Method according to any one of claims 1 to 6, wherein there is carried out a step of calculation by iterations of the local thickness of the shell Ki (16, 18, 20, 22, 24) in a given voxel To of BKi in which:  a recovery of BK (i1) by a plurality of subsets SEj, possibly having two by two of the common intersections, each of these subsets SEj being formed by voxels, and of preferably all the voxels of BK (i1) located at a maximum distance Rj of a voxel of SEj, said anchor of SEj, and denoted ANj; the distances of To are calculated for each of the anchors ANj, and the one whose distance to T0 is minimal is determined, ie AN * corresponding to a subset SE *; the minimum distance d (To, SE *) of To is calculated to any voxel of SE *, that is d *; the value of a distance parameter d, updatable and not increasing, is initialized as a function of the number of iterations of a third iterative process; the third iterative process is carried out in which each of the subsets SEj different from SE * is successively examined, by evaluating the value of the difference Diff = d (To, ANj) Rj. If Diff is smaller than the value of the parameter d, we compute the smallest distance from To to any voxel of SEj, that is d (To, SEj), and if the value obtained dj is less than d, we update d to the value dj. If, on the contrary, the value obtained dj is greater than or equal to d, d is not updated. If the difference Diff is greater than or equal to the value of the parameter d, we do not perform the computation of d (To, SEj). In all cases, we continue the third iterative process until the examination of all sets SEj; the final value of the parameter d is then assigned to the local thickness of the shell Ki at the determined voxel To of BKi (16, 18, 20, 22, 24), that is to say at the minimum distance d ( To, E (i1)).
 The method according to any one of claims 1 to 6, wherein the first iterative process for determining the sets Ei is stopped when the number of iterations reaches the value N = C corresponding to the target segment Ec, or the value N = C + A corresponding to the set E (c., to), in which A is a predetermined positive integer.
 9. A method according to any one of claims 1 to 8, wherein:  a recovery of the digital volume VN is carried out by eight portions of digital volume PVj corresponding to portions of the actual volume delimited by three perpendicular planes two by two of which one axial (or transverse) plane XY, a sagittal plane YZ, and a coronal plane XZ; the first iterative process is carried out in parallel in each of the digital volume portions PVj to determine sets Ei, j (54, 56, 58, 60), Ki, j, BKi, j and the evaluation of a magnitude Gi , j, calculating in parallel these different elements in each of the volume portions PVj, respectively as Ei, Ki, BKi and Gi in each of these portions of digital volume PVj; When a predetermined evolution of one or a plurality of the magnitudes Gi, j is determined, a set (64) of frontier voxels of order 1, which are the border voxels belonging to a set Ei, j d, are determined. a portion of digital volume PVj and which do not belong to a set Ei, k of another portion of digital volume PVk, with k different from j. at least one additional iteration is carried out according to the first iterative process, in the digital volume VN, starting from said set of firstorder frontier voxels as the origin space, each voxel of this first set retaining its local cost to obtain at least an ECO complementary set (66); the method also comprising at least one step of merging sets obtained by the first iterative process.
 10. Method according to any one of claims 1 to 9, wherein is determined and displayed on a screen (74), automatically or at the request of a medical operator, and optionally is produced on a film support, at least one image socalled primary digital obtained from Ec, for example a 2D image corresponding to an axial, or sagittal, or coronal section of a body part corresponding to Ec, or a 3D view of a body part corresponding to Ec.
 11. The method of claim 10, wherein is automatically determined and displays at least one primary digital image 2D corresponding to an axial section, or sagittal, or coronal of a body part, the image corresponding to the segmentation Ec.
 12. The method of claim 10 or 11, wherein, after determination and display of this primary digital image, is made, at the request of a medical operator, a display of another digital image corresponding to the same view of the body as that of the primary image, this other image, called secondary image, corresponding to a set E (c + p) said segmentation of interest, P being a positive or negative integer, the computer realizing if necessary one or more complementary iterations , according to the first iterative process, making it possible to reach the iteration rank C + P.
 13. Method according to any one of claims 1 to 12, wherein is stored in at least one memory of the computer (78) and / or diffuses on a telecommunications network at least one medical information, for example corresponding data. to the sets Ei and / or data corresponding to the target segmentation and / or data corresponding to the segmentation of interest obtained by the method according to claim 12.
 14. A computer program characterized in that it comprises program code instructions for the execution of the steps of the method according to any of claims 1 to 13, which are performed by means of a computer (78) .
 A method of computerassisted tomography or computerassisted MRI (78) of at least a portion of a body, for visualizing a particular area of that at least part of the body, characterized in that one determines and displays at least one primary digital image of the target segmentation obtained by the method according to any one of claims 10 to 12, and / or a secondary image of the segmentation of interest obtained by the method according to claim 12.
 16. A method of scanning or MRI according to claim 15, wherein performing at least the following steps:  one determines and displays one or more digital image (s) scanography or MRI of a part at least the body, said initial image (s), preferably one or more 2D image (s);  by means of digital or physical pointing means, one or more pixels of this or these initial image (s) are selected, to which (s) are associated one or more voxels of VN, which together define the origin space E0; the steps of the method according to any one of claims 10 to 12 are carried out in order to determine and display said primary image; a variation P is determined of the number of iterations of the basic iterative process, corresponding to a set E (c + p) said segmentation of interest, P being a positive or negative integer, less than C in absolute value if P is negative, thanks to adjustable means (98) for modifying the number of iterations; if necessary, using the computer one or more sets Ei are defined by continuing the basic iterative process so as to be able to define the set E (c + p); a secondary image corresponding to this segmentation of interest E (c + p) is displayed
 17. A CT or MRI set comprising a CTscan or MRI device (72) and a computer (78), separate or integrated with this device, characterized in that this installation is able to implement the CT or MRI method according to claim 15 or 16.
 An assembly according to claim 17, wherein the computer (78) comprises a physical medium on which is recorded a computer program according to claim 14.
 19. The assembly of claim 17 or 18, for carrying out the method according to claim 16, comprising adjustable means (98) for modifying the number of iterations, these adjustable means belonging to the group formed by a physical pointer movable to key (s) for click (s) and / or wheel, a thumbscrew or rotary ball, a touch device, for example a touch screen or a touch pad, an optical pointer, a tilting pointing device, and a graphic tablet.
 20. Scanning image or MRI on a filmbased physical medium, characterized in that it was produced from the target segmentation determined during the execution of the method according to one of claims 1 to 13, or from of the segmentation of interest determined during the execution of the method according to claim 12.
Priority Applications (1)
Application Number  Priority Date  Filing Date  Title 

FR1357345A FR3009109A1 (en)  20130725  20130725  Method for producing medical assisted medical images 
Applications Claiming Priority (4)
Application Number  Priority Date  Filing Date  Title 

FR1357345A FR3009109A1 (en)  20130725  20130725  Method for producing medical assisted medical images 
FR1358356A FR3009110B1 (en)  20130725  20130830  Process for the production of medical images assisted by computer 
EP14790158.1A EP3025308A1 (en)  20130725  20140723  Computerassisted method for producing medical images 
PCT/FR2014/051900 WO2015011407A1 (en)  20130725  20140723  Computerassisted method for producing medical images 
Publications (1)
Publication Number  Publication Date 

FR3009109A1 true FR3009109A1 (en)  20150130 
Family
ID=49753328
Family Applications (2)
Application Number  Title  Priority Date  Filing Date 

FR1357345A Pending FR3009109A1 (en)  20130725  20130725  Method for producing medical assisted medical images 
FR1358356A Active FR3009110B1 (en)  20130725  20130830  Process for the production of medical images assisted by computer 
Family Applications After (1)
Application Number  Title  Priority Date  Filing Date 

FR1358356A Active FR3009110B1 (en)  20130725  20130830  Process for the production of medical images assisted by computer 
Country Status (3)
Country  Link 

EP (1)  EP3025308A1 (en) 
FR (2)  FR3009109A1 (en) 
WO (1)  WO2015011407A1 (en) 
Family Cites Families (2)
Publication number  Priority date  Publication date  Assignee  Title 

US5920319A (en) *  19941027  19990706  Wake Forest University  Automatic analysis in virtual endoscopy 
US7379062B2 (en) *  20050801  20080527  Barco Nv  Method for determining a path along a biological object with a lumen 

2013
 20130725 FR FR1357345A patent/FR3009109A1/en active Pending
 20130830 FR FR1358356A patent/FR3009110B1/en active Active

2014
 20140723 WO PCT/FR2014/051900 patent/WO2015011407A1/en active Application Filing
 20140723 EP EP14790158.1A patent/EP3025308A1/en not_active Withdrawn
Also Published As
Publication number  Publication date 

WO2015011407A1 (en)  20150129 
FR3009110B1 (en)  20160101 
FR3009110A1 (en)  20150130 
EP3025308A1 (en)  20160601 
Similar Documents
Publication  Publication Date  Title 

Menze et al.  The multimodal brain tumor image segmentation benchmark (BRATS)  
JP2845995B2 (en)  Region extraction technique  
US20130127848A1 (en)  System and Method for Generating 3D Surface Patches from Unconstrained 3D Curves  
US6813373B1 (en)  Image segmentation of embedded shapes using constrained morphing  
US20030056799A1 (en)  Method and apparatus for segmentation of an object  
US7881878B2 (en)  Systems, devices, and methods for diffusion tractography  
US7379062B2 (en)  Method for determining a path along a biological object with a lumen  
Oktay et al.  Multiinput cardiac image superresolution using convolutional neural networks  
JP4152648B2 (en)  How segmenting a threedimensional image included in the object  
Bitter et al.  Penalizeddistance volumetric skeleton algorithm  
US7990379B2 (en)  System and method for coronary segmentation and visualization  
Toriwaki et al.  Fundamentals of threedimensional digital image processing  
US20070165917A1 (en)  Fully automatic vessel tree segmentation  
Xu et al.  Generating triangulated macromolecular surfaces by Euclidean distance transform  
US6285805B1 (en)  System and method for finding the distance from a moving query point to the closest point on one or more convex or nonconvex shapes  
Marten et al.  Computerassisted detection of pulmonary nodules: performance evaluation of an expert knowledgebased detection system in consensus reading with experienced and inexperienced chest radiologists  
US7242401B2 (en)  System and method for fast volume rendering  
JP4001600B2 (en)  3D image comparing program, a threedimensional image comparison method, and threedimensional image comparison device  
US20110262015A1 (en)  Image processing apparatus, image processing method, and storage medium  
CN101790748A (en)  Method of segmenting anatomic entities in 3d digital medical images  
Liu et al.  Deep convolutional neural network and 3D deformable approach for tissue segmentation in musculoskeletal magnetic resonance imaging  
EP1881453A2 (en)  A medical imageprocessing apparatus and a method for processing medical images  
JP2007528529A (en)  Method and system for identifying the surface of the 3d data set ( "voxel subdivision")  
Jeong et al.  Scalable and interactive segmentation and visualization of neural processes in EM datasets  
US7570802B2 (en)  Automated centerline detection algorithm for colonlike 3D surfaces 