Automatic inner, middle and outer ear segmentation

Abstract

The invention relates to a method for automatically segmenting the cochlea of a person comprised in an image (M) consisting of voxels (v), each voxel comprising an intensity value indicative of the intensity of the respective voxel (v).

Description

[0001] Prior related applications

[0002] This application claims the benefit of priority of the earlier filed European Patent Application No. 22159818.8 filed on March 2, 2022.

[0003] Detailed Description

[0004] The invention relates to a method for detecting and segmenting a human cochlea and to a corresponding system and computer program.

[0005] The inner ear (auris interna) is the innermost part of the human ear and comprises two important functional parts: the cochlea required for converting sound pressure from the outer ear into electrochemical impulses that are transmitted to the brain via the auditory nerve for sound perception, and the vestibular system as a means for maintaining balance. Due to the fact that these structures are quite tiny in size, surgical procedures performed on these organs need to be performed with the utmost precision.

[0006] A typical surgical procedure performed on the human ear is acoustic neuroma surgery to remove acoustic neuromas, particularly through one of the following: translabyrinthine, in which the mastoid bone and bones in the inner ear are removed to gain access to the ear canal to remove the tumor; retrosigmoid / suboccipital, in which the surgeon makes an incision behind the mastoid portion of the ear through an opening in the skull; middle cranial fossa, in which the surgeon removes the tumor from the upper surface of the internal auditory canal outside the inner ear.

[0007] Another typical surgical procedure is cochlear implant (CI) ear surgery. Such surgically implanted electronic devices help restore sound perception to people with severe hearing loss, such as caused by damage or defects in the inner ear. In particular, such CIs can be configured to directly stimulate the auditory nerve to send information to the brain.

[0008] In particular, during natural hearing, the ear anatomy mechanically converts sound into vibrations of the basilar membrane in the cochlea, which stimulate nerves connected to the spiral ganglion and ultimately the auditory nerve, with higher frequencies resulting in stimulation of more basal spiral ganglion nerves and lower frequencies stimulating more apical spiral ganglion nerves. Therefore, a CI can be configured to use this so-called natural tonotopy by applying an electric field to more apical (basal) spiral ganglion nerves to induce perceived lower (higher) frequency sounds.

[0009] In this way, CI induces hearing by stimulating auditory nerve pathways within the cochlea using, for example, an implanted electrode array. The processor of the CI can be programmed to process the sound detected by the microphone and send a signal to each electrode. The CI electrode array is configured so that each electrode stimulates a neural pathway corresponding to a predefined frequency bandwidth when correctly positioned. However, in surgery, the CI electrode array is blindly inserted into the cochlea, which has its insertion path guided only by the walls of the spiral-shaped cochlear cavity. Since the final position of the electrode is usually unknown, the only option when programming the CI is to assume that the electrode is optimally positioned in the cochlea and use a default frequency allocation table.

[0010] Yet another surgical procedure is congenital atresia ear reconstruction, specifically aimed at rebuilding parts such as the ear canal, tympanic membrane (ear membrane), ossicular chain (middle ear hearing bones).

[0011] Additionally, implantation of osseointegrated bone conduction hearing systems allows for the surgical placement of bone-anchored hearing aids into the skull to transmit sound through the bone to the inner ear.

[0012] Additionally, using procedures such as stapedectomy / stapedotomy, the stapes may be removed and replaced with a prosthesis.

[0013] Yet another surgical procedure, tympanoplasty, may be used on the eardrum and / or middle ear bones to restore the middle ear hearing mechanism.

[0014] All of the implantation / surgical procedures outlined above benefit greatly from precise knowledge of the location and shape of the individual parts of the ear.

[0015] The problem to be solved by the present invention is therefore to provide a method, a system and a computer program for reliably detecting and segmenting a human cochlea.

[0016] This problem is solved by a method having the features of claim 1 , a system having the features of claim 17 and a computer program having the features of claim 18 .

[0017] Preferred embodiments of these aspects of the invention are set out in the corresponding dependent claims and described below.

[0018] According to claim 1, a method for automatically segmenting a human cochlea comprised in an image M consisting of voxels v, each voxel comprising an intensity value indicating the intensity of the corresponding voxel v, is disclosed, the method comprising the following steps:

[0019] - Provide a voxel v with the same origin and size as the image M E1 The empty first gradient accumulator image E1 is composed of each voxel v of the first gradient accumulator image E1.E1 is associated with a voxel v of the image M, and for each voxel v of the image M a gradient vector g pointing in the direction of the maximum possible intensity increase of the image M and a first gradient line extending in the direction of the gradient vector g are calculated, the first gradient line starting from a first starting point p start,1 and terminates at the first endpoint p end,1 , where each voxel v of the first gradient accumulator image E1 is E1 The intensity value of each voxel v of the first gradient accumulator image E1 increases by a predefined constant increment E1 The corresponding voxel v of the image M intersects with the corresponding first gradient line,

[0020] - Provide a voxel v with the same origin and size as the image M E2 The empty second gradient accumulator image E2 is composed of each voxel v of the second gradient accumulator image E2. E2 is associated with a voxel v of the image M, and for each voxel v of the image M a gradient vector g pointing in the direction of the maximum possible intensity increase of the image M and a second gradient line extending in the opposite direction of the gradient vector g is calculated, the second gradient line starting at a second starting point p start,2 and terminates at the second endpoint p end,2 , where each voxel v of the second gradient accumulator image E2 is E2 The intensity value of each voxel v of the second gradient accumulator image E2 is increased by a predefined constant increment E2 The corresponding voxel v of the image M intersects with the corresponding second gradient line,

[0021] - Provide a voxel v with the same origin and size as the image M E3 The third gradient accumulator image E3 is composed of an empty third gradient accumulator image E3, each voxel v of the third gradient accumulator image E3 E3 is associated with a voxel v of the image M, and for each voxel v of the image M a gradient vector g pointing in the direction of the maximum possible intensity increase of the image M and a third gradient line extending in the direction of the gradient vector g is calculated, the third gradient line starting at a third starting point p start,3 and terminates at the third endpoint p end,3 , where each voxel v of the first gradient accumulator image E3 is E3 The intensity value of each voxel v of the first gradient accumulator image E3 is increased by a predefined constant increment E3 The corresponding voxel v of the image M intersects the corresponding third gradient line, and

[0022] - Use the first, second and third gradient accumulator images E1, E2, E3 to form a cochlear descriptor image E consisting of voxels Vc descriptor , and select the cochlear descriptor image Edescriptor The descriptor image E in the cochlea is larger than descriptor The intensity values ​​of all other voxels are the number P ≥ 1 of the intensity values ​​of the voxel vc as a prediction of the possible position of the center of the cochlea in the image M.

[0023] In particular, according to an embodiment of the present invention, said selection of said number P ≥ 1 of voxels vc comprises thresholding the cochlear descriptor image E using a value obtained by multiplying the brightest voxel (i.e., highest intensity) of the descriptor image by a factor in the range of 0.6 to 0.8. descriptor , where the factor is preferably 0.7. Thresholding will create one or several 3D objects comprising the highest cumulative density, where each voxel in the voxels vc corresponds to the center of such an object. This strategy is advantageous because most cochlear gradients will not converge to one voxel, but to several neighboring voxels. Therefore, in order to enhance the accuracy of detection (especially on noisy images), one preferably searches for the object with the highest intensity.

[0024] According to an embodiment of the method, forming the cochlear descriptor image corresponds to calculating the Hadamard product of the first, second and third gradient accumulator images (E1, E2, E3)

[0025] Furthermore, according to an embodiment of the method, the respective gradient lines comprise a length extending from the respective starting point to the respective end point, wherein the length of each first gradient line is greater than the length of each third gradient line, and wherein the length of each third gradient line is greater than the length of each second gradient line.

[0026] Furthermore, according to an embodiment of the method, the respective first, second and third starting points p are calculated according to the following formulas: start,1 、p start,2 、p start,3 :

[0027] p start，i =position(v M )+start i gi = 1, 2, 3

[0028] and wherein the respective first, second and third endpoints p are calculated according to the following formula end,1 、p end,2 、p end,3 :

[0029] p end，i =position(v M )+end i gi = 1, 2, 3

[0030] Where position(vM ) represents the position of the receiving voxel (v) of the image (M).

[0031] Furthermore, according to an embodiment of the method, start1 is in the range from 2.7 mm to 3.3 mm, wherein in particular start1 is equal to 3.0 mm.

[0032] Furthermore, according to an embodiment of the method, start2 is in the range from -3.3 mm to -2.7 mm, wherein in particular start1 is equal to -3.0 mm.

[0033] Furthermore, according to an embodiment of the method, start3 is in the range from 0.45 mm to 0.55 mm, wherein in particular start3 is equal to 0.5 mm.

[0034] Furthermore, according to an embodiment of the method, end1 is in the range from 4.5 mm to 5.5 mm, wherein in particular end1 is equal to 5.0 mm.

[0035] Furthermore, according to an embodiment of the method, end2 is in the range from -1.8 mm to -2.2 mm, wherein in particular end2 is equal to -2.0 mm.

[0036] Furthermore, according to an embodiment of the method, end3 is in the range from 1.8 mm to 2.2 mm, wherein in particular end3 is equal to 2.0 mm.

[0037] Furthermore, in an embodiment, the constant increment corresponds to the natural number 1.

[0038] Furthermore, according to an embodiment of the method, the corresponding gradient vector is a normalized gradient vector.

[0039] Furthermore, according to an embodiment of the method, the image (M) is one of the following: a CT image, an MRI image, a cone beam CT image.

[0040] In addition, according to an embodiment of the method, the method further comprises the following steps:

[0041] - providing a statistical shape model of a human ear, said statistical shape model comprising at least a portion of the cochlea corresponding to the cochlea of ​​the human ear.

[0042] In particular, according to an embodiment of the present invention, the SSM comprises, in addition to the cochlea, at least one, several or all of the following further parts of the ear: three semicircular canals, stapes, incus, malleus, tympanic membrane.

[0043] Statistical Shape Models (SSM) are known in image analysis. In the context of the present invention, SSM is used in particular for the segmentation of images of the cochlea.

[0044] The SSM comprises / describes an average shape of at least the cochlea (and in particular the other part(s) present in the SSM), and indicates variations of the training data set relative to the average shape. In particular, according to an embodiment of the present invention, the average shape is calculated based on at least 20 training data sets, each training data set corresponding to a 3D image of an ear of a different person, the 3D image comprising at least a part of the cochlea and in particular the other part(s) described (see above).

[0045] In particular, to compute the changes, rotations and translations are removed from each training dataset.

[0046] In addition, principal component analysis (PCA) is used to extract the variation from the training data set, in particular, by computing a covariance matrix from a training data set matrix containing the training data set. The eigenvectors corresponding to the eigenvalues ​​of the covariance matrix correspond to the direction and amount of variation seen in the training data set. The principal modes of variation are computed based on the first N eigenvectors and associated eigenvalues, where the number N depends on the expected cumulative variance, i.e., the amount of variation that can be explained by the first N eigenvectors. The possible shapes x of the SSM can then be taken as the average shape around The weighted linear combination of these eigenvectors Ψ And is obtained, where the corresponding weight w is represented as the feature weight.

[0047] According to yet another embodiment, the method further comprises the steps of:

[0048] - For each voxel (vc) of the P number of voxels:

[0049] o define a volume in the image (M) with the corresponding voxel (vc) as center,

[0050] o Segmenting the cochlea within the volume with the help of a Hessian-based enhancement filter, thereby obtaining a segmented cochlear region of the image M, the region corresponding to the cochlea in the image M. In particular, the Hessian-based filter can be configured to detect dark tubular structures, wherein in particular α=0.1, β=0.5 and γ=100. The filter is configured in a multiscale mode, wherein σ varies between 0.3 mm and 0.7 mm in 3 steps, for example σ=0.3, σ=0.5 and σ=0.7. In particular, the segmented cochlear region can be converted into a 3D mesh, for example by applying a marching cubes algorithm (see, for example, William E. Lorensen, Harvey E. Cline: Marching Cubes: A high resolution 3D surface construction algorithm in Computer Graphics, Vol. 21, Nr. 4, July 1987) to the segmented cochlear region in order to obtain a 3D surface of the cochlea,

[0051] o Fitting said cochlear portion of the statistical shape model to the 3D surface of the cochlea. In particular, the fitting may comprise testing a range of cochlear sizes, i.e. varying the first variation pattern responsible for the size (scale), wherein in particular testing the following values ​​for the first entry of w in : -3.0, -2.0, -1.0, 0.0, 1.0, 2.0, 3.0, where for each generated cochlea a rotation about its center is applied in order to find the best orientation,

[0052] o calculating a surface similarity measure (e.g. root mean square difference) indicative of the similarity between the surface of said cochlear portion of the statistical shape model and the surface of the segmented cochlear region of the image (M),

[0053] - determining a voxel Vc among the P number of voxels as the center of the cochlea, for which a surface similarity measure satisfies a predefined criterion, wherein in particular a first voxel vc among the P number of voxels is selected as the center of the cochlea, for which a surface similarity measure falls below a predefined threshold.

[0054] In particular, the segmentation of an image corresponds to classifying whether a voxel of the image belongs to a certain structure or not. In particular, the following Hessian-based enhancement filter can be used in this context.

[0055] Furthermore, the corresponding volume may be a cubic or spherical volume, but may also comprise any other suitable shape, in particular the volume is a cube having an edge length of at least 20 mm.

[0056] In addition, according to an embodiment of the method, the method further comprises the following steps:

[0057] - fitting a cochlear portion of a statistical shape model to the segmented cochlear region using the determined center of the cochlea, the statistical shape model further comprising semicircular canal portions corresponding to the three semicircular canals of the ear,

[0058] - predicting the positions of the three semicircular canals in the image (M) as the positions of the three semicircular canals in the semicircular canal portion of the statistical shape model,

[0059] - define a volume that includes the predicted location, and

[0060] - segmenting the three semicircular canals within the volume (e.g. with the help of a Hessian-based enhancement filter) thereby obtaining segmented semicircular canal regions of the image M corresponding in particular to the three semicircular canals in the image M. In particular, the used Hessian-based enhancement filter is configured to detect dark tubular structures, wherein in particular α=0.5, β=0.1, Y=100 and σ is between 0.3 mm and 0.7 mm in 3 steps (e.g. σ=0.3, σ=0.5 and σ=0.7).

[0061] Furthermore, the three semicircular canals in image M may be annotated based on the statistical shape model and its fitting into the segmented cochlear region and the segmented semicircular canal region.

[0062] Again, the volume used may be a cubic volume, in particular a cube, a spherical volume or any other suitable volume.

[0063] In addition, according to an embodiment of the method, the method further comprises the following steps:

[0064] - Fitting the cochlear part and the semicircular canal part of the statistical shape model to the segmented cochlear region corresponding to the cochlea in the image and to the segmented semicircular canal regions corresponding in particular to the three semicircular canals in the image M.

[0065] Advantageously, this step improves the ear SSM shape fit to the anatomy and makes further predictions more accurate.

[0066] In addition, according to an embodiment of the method, the method further comprises the following steps:

[0067] - predicting the positions of the incus, malleus and stapes in the image as the positions of the incus, malleus and stapes in the ossicular portion of the statistical shape model, the ossicular portion of the statistical shape model corresponding to the incus, malleus and stapes of the ear,

[0068] - define a volume that includes the predicted location, and

[0069] - Segmenting the incus, malleus and stapes within the volume, thereby obtaining a segmented ossicular region of the image M corresponding to the incus, malleus and stapes in the image M.

[0070] Furthermore, in accordance with an embodiment of the present invention, annotating the incus, malleus and stapes in the image is performed based on a statistical shape model and its fitting into the segmented cochlear region, the segmented semicircular canal region and the segmented ossicle region.

[0071] Again, the volume used may be a cubic volume, in particular a cube, a spherical volume or any other suitable volume.

[0072] In addition, according to an embodiment of the method, the method further comprises the following steps:

[0073] - predicting the position of the eardrum as the position of the eardrum in the eardrum portion of the statistical shape model,

[0074] The tympanic membrane part of the statistical shape model corresponds to the eardrum,

[0075] - define a volume that includes the predicted location, and

[0076] - segmenting the eardrum within the volume (e.g. with the help of a Hessian-based enhancement filter), thereby obtaining a segmented eardrum region of the image M corresponding to the eardrum in the image M,

[0077] - In particular, this Hessian filter can be configured to detect bright plate-like structures,

[0078] In particular, α = 0.1, β = 1.0, Y = 50 and σ is between 0.2 ... 0.4 in 2 steps,

[0079] For example, σ=0.2 and σ=0.4.

[0080] Furthermore, according to an embodiment, the annotation of the tympanic membrane in the image may be guided based on a statistical shape model and its fitting into at least the segmented cochlear region, the segmented semicircular canal region, and the segmented tympanic membrane region.

[0081] Again, the volume used may be a cubic volume, in particular a cube, a spherical volume or any other suitable volume.

[0082] In particular, the eardrum is the end of the external auditory canal. Once it has been segmented, one knows where the ear canal ends and also its orientation (in particular, a best fit plane of the eardrum will be sufficient to predict the external auditory canal direction).

[0083] In particular, in an embodiment, once the end of the external canal end and its orientation are known, a cylindrical region of interest is defined (which best describes the shape of the canal). With the defined shape, the cortical bone representing the canal is then segmented. This bone can be segmented using simple thresholding or with the help of a Hessian filter (configured as a bright plate-like structure).

[0084] In addition, according to an embodiment of the method, the method further comprises the following steps:

[0085] - predicting the area, orientation and limits of the position of the external auditory canal based on the segmented eardrum area of ​​the image M corresponding to the eardrum in the image M,

[0086] - define the volume of the area that includes the prediction,

[0087] - Segmenting the external auditory canal within the volume (eg with the help of a Hessian based enhancement filter) thereby obtaining a segmented external auditory canal region of the image M corresponding to the external auditory canal in the image M.

[0088] Furthermore, in an embodiment, annotating the external auditory canal in the image may be performed based on a statistical shape model and its fitting into at least the segmented tympanic membrane region.

[0089] Again, the volume used may be a cubic volume, in particular a cube, a spherical volume or any other suitable volume.

[0090] In addition, according to an embodiment of the method, the method further comprises the following steps:

[0091] - predicting the position and orientation of the outer surface of the temporal bone based on the segmented external auditory canal region of the image M corresponding to the external auditory canal in the image M,

[0092] - define a volume that includes the predicted location, and

[0093] - With the help of thresholding, the temporal bone is segmented within the volume, thereby obtaining a segmented temporal bone region of an image (M) including the outer surface of the temporal bone (and in particular adjacent layers of about 5 mm thickness), and the outer surface is detected using the direction of the external auditory canal. In particular, a surface having a normal pointing in the direction of the external canal (with a deviation within 40 degrees) is considered to be an outer surface.

[0094] Furthermore, according to an embodiment, annotation of the temporal bone in the image may be performed.Again, the volume used may be a cubic volume, in particular a cube, a spherical volume or any other suitable volume.

[0095] In addition, according to an embodiment of the method, the method further comprises the following steps:

[0096] - predicting the location of the internal auditory canal based on a segmented cochlear region of the image corresponding to the cochlea in the image, and

[0097] - define a volume that includes the predicted location, and

[0098] - segmenting the internal auditory canal within the volume, thereby obtaining a segmented internal auditory canal region of said image M corresponding to the internal auditory canal in the image.

[0099] Furthermore, in embodiments of the invention, annotating the inner auditory canal in the image may be performed.Again, the used volume may be a cubic volume, in particular a cube, a spherical volume or any other suitable volume.

[0100] In addition, according to an embodiment of the method, the method further comprises the following steps:

[0101] - predicting the location of the facial nerve based on the segmented cochlear region of the image M corresponding to the cochlea and the segmented semicircular canal regions corresponding to the three semicircular canals in the image,

[0102] - define a volume that includes the predicted location, and

[0103] - segmenting the facial nerves within the volume (eg with the help of a Hessian based enhancement filter) thereby obtaining segmented facial nerve regions of the image M corresponding to the facial nerves in said image.

[0104] Furthermore, according to an embodiment, annotating facial nerves in the segmented image is performed.

[0105] In addition, according to an embodiment of the method, the method further includes the step of graphically visualizing at least one of the following on a display for a user: a segmented cochlear region, a segmented semicircular canal region, a segmented ossicular region, a segmented tympanic membrane region, a segmented external auditory canal region, a segmented temporal bone region, a segmented internal auditory canal region, and a segmented facial nerve region.

[0106] Furthermore, according to a further aspect of the present invention, a system for automatically segmenting a human cochlea included in an input image M consisting of voxels V, each voxel comprising an intensity value indicating the intensity of the corresponding voxel, wherein the system comprises at least one processor and a display, and the system is suitable for executing the method according to the present invention.

[0107] In particular, the system is configured to display one, several or all of the following segmented regions of image M on the display: a segmented cochlear region, a segmented semicircular canal region, a segmented ossicular region, a segmented tympanic membrane region, a segmented external auditory canal region, a segmented temporal bone region, a segmented internal auditory canal region, and a segmented facial nerve region.

[0108] In particular, the corresponding segmented regions can be superimposed on the image M on the display and / or can be displayed in an isolated manner.

[0109] According to a further aspect of the invention, a computer program is disclosed, the computer program comprising instructions which, when the program is executed on a system according to the invention, cause the system to perform the steps of the method according to the invention.

[0110] According to another aspect of the present invention, a computer-readable (in particular non-transitory) storage medium comprises instructions which, when executed by the system according to the invention (or by a computer), cause the system according to the invention (or the computer) to perform the steps of the method according to the present invention.

[0111] Furthermore, yet another aspect of the invention relates to a data carrier signal carrying the computer program according to the invention.

[0112] Further features and advantages of the invention and embodiments of the invention will be described below with reference to the drawings, in which

[0113] Figure 1 An illustration of the cochlea and its characteristic shapes indicated by means of different circles is shown;

[0114] Figure 2 The geometric principle of automatically detecting the center of the cochlea is shown;

[0115] Figure 3 Schematically shows an empty first gradient accumulator image E1 for detecting the center of the cochlea in a medical image;

[0116] Figure 4 shows fitting the SSM of the ear to the cochlea in the image using a surface similarity metric based on the center of the detected cochlea, which also allows detecting whether the cochlea belongs to the left or right ear;

[0117] Figure 5 Predictions of the locations of the semicircular canals of the segmented cochlea and their segmentation are shown;

[0118] Figure 6 The labeling of the semicircular canals based on the ear SSM adjusted to the segmented cochlea and semicircular canals is illustrated;

[0119] Figure 7 It explains predicting the location of the tympanic membrane and segmenting the tympanic membrane;

[0120] Fig. 8A It illustrates, for example, the temporal bone projections;

[0121] Figure 8B It illustrates, for example, facial neural projections;

[0122] Fig. 9 An embodiment of a system according to the present invention is shown; and

[0123] Fig.10 An embodiment of a system according to the invention is shown.

[0124] The invention provides a method for automatically detecting and segmenting a human cochlea comprised in an image M consisting of voxels v, each voxel comprising an intensity value indicating the intensity of the respective voxel v. Such an image may be acquired by a suitable medical imaging device which may be part of the system according to the invention.

[0125] To detect the center 11 of the cochlea 10, as in Figure 3 As indicated in the figure, the voxel v E1 The empty first gradient accumulator image E1 is composed of, where each voxel v of the first gradient accumulator image E1 E1 is associated with a voxel v of the image M. Then, for each voxel v of the image M, a gradient vector g pointing in the direction of the maximum possible intensity increase of the image M and a first gradient line extending in the direction of the gradient vector g are calculated, the first gradient line starting from the first starting point p start,1 and terminates at the first endpoint p end,1 (See Figure 2 ), where each voxel v of the first gradient accumulator image E1 E1 The intensity value of each voxel v of the first gradient accumulator image E1 increases by a predefined constant increment E1 The corresponding voxel v of the image M intersects the corresponding first gradient line.

[0126] In addition, the voxel v is generated E2 The empty second gradient accumulator image E2 (see Figure 3 ), each voxel v of the second gradient accumulator image E2 E2 Also associated with a voxel v of the image M, wherein for each voxel v of the image M a gradient vector g pointing in the direction of the maximum possible intensity increase of the image M and a second gradient line extending in the opposite direction of the gradient vector g are calculated, the second gradient line starting at a second starting point p start,2 and terminates at the second endpoint p end,2 (See Figure 2 ), where each voxel v of the second gradient accumulator image E2 is E2 The intensity value of each voxel v of the second gradient accumulator image E2 is increased by a predefined constant increment E2 The corresponding voxel v of the image M intersects the corresponding second gradient line.

[0127] Similarly, provided by the voxel v E3The empty third gradient accumulator image E3 (see Figure 3 ), each voxel v of the third gradient accumulator image E3 E3 is associated with a voxel v of the image M, wherein for each voxel v of the image M a gradient vector g pointing in the direction of the maximum possible intensity increase of the image M and a third gradient line extending in the direction of the gradient vector g are calculated, the third gradient line starting from a third starting point p start,3 and terminates at the third endpoint p end,3 (See Figure 2 ), where each voxel v of the first gradient accumulator image E3 is E3 The intensity value of each voxel v of the first gradient accumulator image E3 is increased by a predefined constant increment E3 The corresponding voxel v of the image M intersects the corresponding third gradient line.

[0128] The cochlear descriptor image E consisting of voxels vc is then generated by combining the first, second and third gradient accumulator images E1, E2, E3 descriptor , and select the cochlear descriptor image E descriptor The descriptor image E in the cochlea is larger than descriptor The intensity values ​​of all other voxels whose number P ≥ 1 are used as a prediction of the position of the center of the cochlea in the image M. In particular, the process described above can be used to determine a candidate for the center 11 of the cochlea 10 .

[0129] The reason for using three accumulator images E1 , E2 , E3 is based on the following observation: In general, the cochlear shape can be described as a snail shape and consists of two and a half turns. A relatively large number of the normals of the cochlear surface points will therefore pass through the actual center 11 of the cochlea 10 .

[0130] The main principle behind the cochlear center detection algorithm can thus be understood as the task of finding the point in space that contains the maximum number of surface normal intersections.

[0131] exist Figure 2 This principle for detecting the center of a circle with a known radius is illustrated in . Here, a short line is defined that starts before the center of the circle and ends after the center of the circle. Such an algorithm allows detecting circles of defined sizes.

[0132] This principle can be used to design algorithms for cochlear center detection, since the average radius of the cochlea and its variation are known.

[0133] In the following, as an example, a pseudocode for detecting the center of a circle is stated. The algorithm requires the following as input:

[0134] - Include the center to be detected (e.g. the center of the cochlea, here a circle for simplicity)

[0135] an image M, wherein the image M may be a CT, MR or cone beam CT volume image containing the ear / cochlea;

[0136] - an empty image E, in particular with the same origin and size as M. The image spacing (voxel size) can be different from the image M. This image is called the accumulator image;

[0137] -start distance - the distance from the voxel position to the start of the line (along the voxel gradient direction);

[0138] -end distance - the distance from the voxel position to the end point of the line (along the direction of the voxel gradient),

[0139] And produces the accumulated gradient vector intersection image as output.

[0140] The corresponding pseudo code can be expressed as:

[0141]

[0142] By recognizing that the shape of the cochlea defines Figure 1 With three such circles schematically shown in , the circle center detection algorithm can be easily adapted to the case of the cochlea described initially.

[0143] In particular, the cochlea can be modeled as the three different circles as follows:

[0144] The first circle C1 captures the outer wall of the basal turn of the cochlea 10. Anatomically, it represents the change from cochlear bone (high intensity) to cochlear fluid (low intensity), with these gradient vectors pointing towards the center 11 of the cochlea.

[0145] The second circle C2 captures the inner wall of the basal turn of the cochlea 10. Anatomically, it represents the change from cochlear fluid (low intensity) to cochlear bone (high intensity), with these gradient vectors pointing outwards away from the cochlear center 11, but aligned with the center.

[0146] The third circle C3 captures the outer wall of the second-and-a-half turn of the cochlea. Anatomically, it represents the change from cochlear bone (high intensity) to cochlear fluid (low intensity), with these gradient vectors again pointing towards the center 11 of the cochlea.

[0147] In pseudocode, this can be implemented as follows:

[0148] With an input image M, we compute an accumulator image E for each circle i :

[0149] That is, for circle C1, the accumulator image is:

[0150] E1=Circle Center Detection Algorithm (M, 3.0, 5.0).

[0151] For the second circle C2, the gradient is pointing outwards, so we use a negative distance:

[0152] E2=Circle Center Detection Algorithm (M, -3.0, -2.0).

[0153] For the third circle C3, one can use:

[0154] E3=Circle Center Detection Algorithm (M, 0.5, 2.0).

[0155] Then for example the cochlear descriptor image is computed as the multiplication of the 3 accumulated images:

[0156] E descriptor =E1*E2*E3

[0157] For example, E descriptor The brightest voxel of may be selected as representative of the center 11 of the cochlea 10. Other processes for selecting candidates for the center 11 of the cochlea as described herein may also be applied.

[0158] A preferred variant of finding a candidate for the center 11 of the cochlea 10 is to threshold the cochlear descriptor image E with a value obtained by multiplying the brightest voxel of the cochlear descriptor image by a factor ranging from 0.6 to 0.8 (preferably 0.7) (i.e. only considering the corresponding brightest fraction of the voxel). descriptor This will typically result in a descriptor image comprising a plurality of 3D objects of highest cumulative density, where each candidate for the center 11 of the cochlea is the center of such an object.

[0159] like Figure 4 As indicated in , in an embodiment of the invention, these possible centers of the cochlea may be tested as follows in order to find the true center 11 of the cochlea 10:

[0160] a. cropping around the corresponding candidate for the center of the cochlea using a box B1 of size, for example, 20x20x20 mm (other volume shapes and sizes may also be used),

[0161] b. automatically segmenting the tubular structure in said box B1 with the help of a Hessian-based enhancement filter, wherein the segmented structure is preferably converted into a 3D mesh, in particular by applying a marching cubes algorithm to the segmented structure, in order to obtain a 3D surface of the segmented structure,

[0162] c. automatically fitting an ear statistical shape model (SSM) (i.e. only the cochlear part) into the segmented data D1, in particular by fitting the cochlear part to said 3D surface, wherein both left and right cochlear ear SSMs are fitted, since initially it is not known to which side (left or right ear) the cochlea belongs,

[0163] d. Based on the surface similarity measure, automatically determine which candidate represents the center 11 of the cochlea 10 (ie results in the best surface similarity). This also yields the side of the cochlea (left or right).

[0164] also, Figure 5 The prediction of the positions of the semicircular canals 12 of the segmented cochlea 10 and their segmentation are illustrated:

[0165] a. To automatically predict the position and segment the semicircular canals 12, the ear SSM is automatically adjusted to the segmented cochlea 10, which produces a prediction of the position of the semicircular canals 12,

[0166] b. Then crop a volume such as box B2 around the predicted area, and

[0167] c. Finally, segmentation is automatically performed in the volume to segment the semicircular canals 12 in image M.

[0168] like Figure 6 As shown in , by adjusting the ear SSM to the cochlea 10 and the semicircular canals 12 , the semicircular canals 12 can be marked (annotated).

[0169] Furthermore, the cochlear and semicircular canal parts of the statistical shape model SSM can be automatically fitted to the segmented cochlea 10 and to the segmented semicircular canals 12, based on which the positions of the incus 5, the malleus 6 and the stapes 7 can be predicted and then segmented in the volume B4 (see also Fig. 8A ).

[0170] Furthermore, once the ear SSM is fitted into the cochlea 10 and the semicircular canals 12, the approximate position of the eardrum 13 is known. Figure 7 This is illustrated in an exemplary manner in . Again, a volume B3 is defined which includes the predicted position, and the eardrum 13 is automatically segmented within the volume B3, so that a segmented eardrum region of the image M is obtained.

[0171] also, Fig. 8A The inner ear 1 and the middle ear 2 are shown, as well as the Eustachian tube 3 and the auricle 4. Fig. 8A As explained in , the area, orientation and limits of the position of the external auditory canal 14 can be predicted based on the segmented tympanic membrane 13 (see above). After a volume B5 including the predicted area has been defined, the external auditory canal 14 can be automatically segmented within the volume B5. Based on the segmented external auditory canal 14, the position and orientation of the outer surface 15 of the temporal bone 16 can be automatically predicted. After a volume B6 including the predicted position has been defined, the temporal bone 16 (the bone always has a high intensity) can be automatically segmented within the volume B6 with the help of thresholding. The direction of the external auditory canal 14 can be used to detect the outer surface 15.

[0172] Furthermore, since the internal auditory canal 17 is located below the cochlea 10, the position of the internal auditory canal 17 can also be automatically predicted based on the segmented cochlea 10. Having defined a volume B7 including the predicted position, the internal auditory canal can be automatically segmented within the volume B7.

[0173] Furthermore, the location of the facial nerve 18 can be automatically predicted based on the segmented cochlea 10 and the segmented semicircular canals 12. Figure 8B The volume B8 including the predicted position shown in FIG. 1 may automatically segment the facial nerve 18 within the volume B8 .

[0174] The segmented images can be advantageously used for a variety of different applications, such as planning surgery, monitoring during surgery, image-guided surgery, diagnosis and patient education, in particular showing the planned surgery and / or treatment to the patient, in particular before surgery, and the like.

[0175] also, Fig. 9 An embodiment of a system 100 according to the invention is shown, the system 100 being configured to automatically detect and segment a cochlea 10 of a human ear comprised in an input image M consisting of voxels v, each voxel v comprising an intensity value indicating the intensity of the respective voxel. Preferably, the system 100 comprises at least one processor 101 and a display 102, the at least one processor 101 and the display 102 being adapted to perform the method according to the invention.

[0176] Preferably, the system is configured to display one, several or all of the following segmented regions of the image M on the display 102: a segmented cochlear region 10 and / or its center 11, a segmented semicircular canal region 12, a segmented ossicular region 5, 6, 7, a segmented tympanic membrane region 13, a segmented external auditory canal region 14, a segmented temporal bone region 16, a segmented internal auditory canal region 17, and a segmented facial nerve region 18.

[0177] The system 100 is preferably configured to allow input data such as (one or more) images M and a statistical shape model SSM to be input. The system 100 according to the invention can be implemented by a general purpose computer and a display connected thereto, on which a computer program can be executed, causing the computer to perform the method according to the invention. However, the system can also include or consist of hard-wired components / functions configured to perform the method according to the invention. In addition, Fig.10 An embodiment of a system 100 according to the present invention is shown, which is customized towards performing image-guided surgical procedures. Fig. 9 In addition to the components described together, the system 100 further comprises at least one surgical instrument 103 for performing a surgical operation on the patient's ear, in particular on one of the segmented parts of the ear such as the cochlea 10, the semicircular canals 12, the ossicles 5, 6, 7, the tympanic membrane 13, the external auditory canal 14, the temporal bone 16, the internal auditory canal 17, the facial nerve 18. In particular, the surgical instrument can be driven by an actuator, which is configured to be controlled by a doctor or to be fully automatically controlled by means of an operating device. The operating device can be connected to at least one processor / computer 101 via an interface, and the actuator can be controlled by means of at least one processor / computer 101.

[0178] In particular, here, the system 100 is preferably configured to perform one, several or all of the following functions:

[0179] - automatically segmenting and in particular annotating in at least one image M at least one structure or structures of the ear as described herein (such as the cochlea 10, the semicircular canals 12, the ossicles 5, 6, 7, the tympanic membrane 13, the external auditory canal 14, the temporal bone 16, the inner ear

[0180] tract 17, facial nerve 18);

[0181] - automatically depicting a graphical visualization of the at least one structure or the plurality of structures (e.g. on a display

[0182] display 102);

[0183] - performing an automatic registration of the at least one structure to the current position of the patient or the

[0184] Automatic registration of multiple structures to the patient's current position;

[0185] - displaying such registered structure(s), particularly in concert with a real-time video representation of the patient during surgery (e.g. on display 102 and / or on another display);

[0186] display);

[0187] - Automatic tracking of at least one surgical instrument relative to the patient and relative to the registered structure(s), in particular enabling full visualization of the surgical site.

[0188] In the following, further aspects of the invention and embodiments thereof are listed as items. These items can also be formulated as claims of the invention. Reference numerals in brackets refer to the figures described in this text.

[0189] Item 1: A method for automatically segmenting a human cochlea (10) included in an image (M) composed of voxels (v), each voxel including an intensity value indicating the intensity of the corresponding voxel (v), the method comprising the following steps:

[0190] - Provided by voxels (v E1 ), each voxel (v E1 ) is associated with a voxel (v) of the image (M), and for each voxel (v) of the image (M) a gradient vector (g) pointing in a direction of maximum possible intensity increase towards the image (M) and a first gradient line extending in the direction of the gradient vector (g), the first gradient line starting from a first starting point (p start,1 ) and terminates at the first endpoint (p end,1 ), where each voxel (v E1 ) is increased by a predefined constant increment, each voxel (v E1 ) intersects the corresponding voxel (v) of the image (M) with the corresponding first gradient line,

[0191] - Provided by voxels (v E2 ), each voxel (v E2 ) is associated with a voxel (v) of the image (M), and for each voxel (v) of the image (M) a gradient vector (g) pointing in a direction of maximum possible intensity increase towards the image (M) and a second gradient line extending in the opposite direction of the gradient vector (g), the second gradient line starting at a second starting point (p start,2 ) and terminates at the second endpoint (p end,2 ), where each voxel (v E2 ) is increased by a predefined constant increment, each voxel (v E2 ) intersects the corresponding voxel (v) of the image (M) with the corresponding second gradient line,

[0192] - Provided by voxels (v E3 ) is an empty third gradient accumulator image (E3), each voxel (v E3 ) is associated with a voxel (v) of the image (M), and for each voxel (v) of the image (M) a gradient vector (g) pointing in a direction of maximum possible intensity increase towards the image (M) and a third gradient line extending in the direction of the gradient vector (g), the third gradient line starting at a third starting point (p start,3 ) and terminates at the third endpoint (p end,3 ), where each voxel (v E3 ) is increased by a predefined constant increment, each voxel (v E3 ) intersects the corresponding voxel (v) of the image (M) of FIG. 1 , and

[0193] - Using the first, second and third gradient accumulator images (E1, E2, E3) to form a cochlear descriptor image (E) composed of voxels (Vc) descriptor ), and select the cochlear descriptor image (E descriptor ) has a larger descriptor image than the cochlear descriptor image (E descriptor ) as a prediction of the position of the center (11) of the cochlea (10) in the image (M) by combining the intensity values ​​of all other voxels of the voxel (Vc) whose number is P ≥ 1.

[0194] Item 2: The method according to Item 1, wherein a cochlear descriptor image (E descriptor ) corresponds to computing the Hadamard product of the first, second and third gradient accumulator images (E1, E2, E3)

[0195] Item 3: A method according to one of the preceding items, wherein the corresponding gradient line comprises a gradient from the corresponding starting point (p start,1 、p start,2 、p start,3 ) extends to the corresponding end point (p end,1 、p end,2 、p end,3 ), wherein the length of each first gradient line is greater than the length of each third gradient line, and wherein the length of each third gradient line is greater than the length of each second gradient line.

[0196] Item 4: A method according to one of the preceding items, wherein the respective first, second and third starting points p are calculated according to the following formulas: start,1 、p start,2 、p start,3 :

[0197] p start，i =position(v)+start i gi = 1, 2, 3

[0198] and wherein the respective first, second and third endpoints p are calculated according to the following formula end,1 、p end,2 、p end,3 :

[0199] p end，i =position(v)+end i gi = 1, 2, 3

[0200] where position(v) denotes the position of the recipient voxel (v) of the image (M), and where -start1 is in the range from 2.7 mm to 3.3 mm, where in particular start1 is equal to 3.0 mm,

[0201] -start2 is in the range from -3.3 mm to -2.7 mm, where in particular start1 is equal to -3.0 mm,

[0202] - start3 is in the range from 0.45 mm to 0.55 mm, wherein in particular start3 is equal to 0.5 mm,

[0203] -end1 is in the range from 4.5 mm to 5.5 mm, wherein in particular end1 is equal to 5.0 mm, -end2 is in the range from -1.8 mm to -2.2 mm, wherein in particular end2 is equal to -2.0 mm,

[0204] - end3 is in the range from 1.8 mm to 2.2 mm, wherein in particular end3 is equal to 2.0 mm.

[0205] Item 5: A method according to one of the preceding items, wherein the image (M) is one of the following: a CT image, an MRI image, a cone beam CT image.

[0206] Item 6: The method according to one of the preceding items, wherein the method further comprises the following steps:

[0207] - providing a statistical shape model (SSM) of the ear, said statistical shape model (SSM) comprising at least a cochlear portion corresponding to the cochlea (10) of the ear.

[0208] Item 7: The method according to Item 6, wherein the method further comprises the following steps:

[0209] - For each voxel (Vc) of said number P of voxels:

[0210] o define a volume (B1) in the image (M) having the corresponding voxel (vc) as center,

[0211] o segmenting the cochlea (10) within the volume (B1) with the help of a Hessian-based enhancement filter, thereby obtaining segmented cochlear regions of said image (M), wherein a 3D surface of the cochlea is generated from the segmented cochlear regions, in particular by using a marching cubes algorithm,

[0212] o fitting a statistical shape model (SSM) of said cochlear portion to a 3D surface,

[0213] o calculating a surface similarity measure indicating the similarity between the surface of the cochlear portion of the statistical shape model (SSM) and the 3D surface,

[0214] - determining a voxel (vc) of said number P of voxels as the center (11) of the cochlea (10) for which a surface similarity measure satisfies a predefined criterion, wherein in particular the first voxel (vc) of said number P of voxels is selected as the center of the cochlea,

[0215] The surface similarity metric falls below a predefined threshold.

[0216] Item 8: The method according to Item 7, wherein the method further comprises the following steps:

[0217] - fitting a cochlear portion of a statistical shape model (SMM) to the segmented cochlear region using the determined center (11) of the cochlea (10), the statistical shape model (SSM) further comprising semicircular canal portions corresponding to the three semicircular canals (12) of the ear,

[0218] - predicting the positions of the three semicircular canals (12) in the image (M) as the positions of the three semicircular canals (12) in the semicircular canal portion of the statistical shape model (SSM),

[0219] - defining a volume (B2) including the predicted location, and

[0220] - Segmenting the three semicircular canals (12) within the volume (B2) thereby obtaining the segmented semicircular canal regions of the image (M).

[0221] Item 9: The method according to Item 8, wherein the method further comprises the following steps:

[0222] - Fitting the cochlear and semicircular canal parts of a statistical shape model (SSM) to the segmented cochlear region and to the segmented semicircular canal region.

[0223] Item 10: The method according to Item 9, wherein the method further comprises the following steps:

[0224] - predicting the positions of the incus (5), the malleus (6) and the stapes (7) in the image (M) as the positions of the incus (5), the malleus (6) and the stapes (7) in the ossicular portion of the statistical shape model (SSM), the ossicular portion of the statistical shape model corresponding to the incus (5), the malleus (6) and the stapes (7) of the ear,

[0225] - defining a volume (B4) including the predicted position, and

[0226] - Segmenting the incus, malleus and stapes within the volume, thereby obtaining a segmented ossicular region of the image (M) corresponding to the incus, malleus and stapes in the image (M).

[0227] Item 11: The method according to Item 9 or 10, wherein the method further comprises the following steps:

[0228] - predicting the position of the eardrum (13) as the position of the eardrum (13) in the eardrum portion of a statistical shape model (SSM), the eardrum portion of the statistical shape model corresponding to the eardrum (13) of the ear,

[0229] - defining a volume (B3) including the predicted location, and

[0230] - Segmenting the eardrum (13) within the volume (B3) thereby obtaining the segmented eardrum region of the image (M).

[0231] Item 12: The method according to Item 11, wherein the method further comprises the following steps:

[0232] - predicting the area, orientation and limits of the position of the external auditory canal (14) based on the segmented eardrum area of ​​the image (M),

[0233] - define the volume (B5) of the area including the prediction,

[0234] - Segmenting the external auditory canal (14) within the volume (B5), thereby obtaining the segmented external auditory canal region of the image (M).

[0235] Item 13: The method according to Item 12, wherein the method further comprises the following steps:

[0236] - predicting the position and orientation of the outer surface (15) of the temporal bone (16) based on the segmented external auditory canal region of the image (M),

[0237] - defining a volume (B6) including the predicted position, and

[0238] - Segmenting the temporal bone (16) within the volume (B6) with the help of thresholding, thereby obtaining the segmented temporal bone region of the image (M), and using the direction of the external auditory canal (14) to detect the external surface.

[0239] Item 14: The method according to item 7 or according to one of items 8 to 13 when referring back to item 7, wherein the method further comprises the following steps:

[0240] - predicting the location of the internal auditory canal (17) based on the segmented cochlear region of the image, and

[0241] - defining a volume (B7) including the predicted position, and

[0242] - Segmenting the internal auditory canal (17) within the volume (B7) thereby obtaining the segmented internal auditory canal region of said image (M).

[0243] Item 15: The method according to items 7, 9 and 12, wherein the method further comprises the following steps:

[0244] - predicting the position of the facial nerve (18) based on the segmented cochlear region, the segmented semicircular canal region and the segmented external auditory canal region of the image (M),

[0245] - defining a volume (B8) including the predicted position, and

[0246] - Segmenting the facial nerve (18) within the volume (B8) thereby obtaining the segmented facial nerve region of the image (M).

[0247] Item 16: A method according to one of Items 7 to 15, wherein the method further includes a step of graphically visualizing for a user on a display (102) at least one of the following: a segmented cochlear region, a segmented semicircular canal region, a segmented ossicular region, a segmented tympanic membrane region, a segmented external auditory canal region, a segmented temporal bone region, a segmented internal auditory canal region, and a segmented facial nerve region.

[0248] Item 17: A system (100) for automatically segmenting a human cochlea (10) included in an input image (M) consisting of voxels (v), each voxel comprising an intensity value indicating the intensity of the corresponding voxel, wherein the system (100) comprises at least one processor (101) and a display (102), and the at least one processor (101) and the display (102) are suitable for performing a method according to one of items 1 to 16.

[0249] Item 18: A computer program comprising instructions which, when executed on a system (100) according to Item 17, cause the system (100) to perform the steps of a method according to one of Items 1 to 16.

[0250] Item 19: A computer-readable storage medium comprising instructions, which, when executed by the system according to Item 17, cause the system according to claim 17 to perform the steps of the method of one of claims 1 to 16.

[0251] Item 20: A data carrier signal carrying the computer program of Item 18.

Claims

1. A method for automatically segmenting a human cochlea (10) comprised in an image (M) composed of voxels (v), each voxel comprising an intensity value indicating the intensity of the corresponding voxel (v), the method comprising the following steps: - Provided by voxels (v E1 ) is an empty first gradient accumulator image (E1), each voxel (v E1 ) is associated with a voxel (v) of said image (M) and a gradient vector (g) pointing in the direction of the maximum possible intensity increase towards said image (M) is calculated for each voxel (v) of said image (M) and a first gradient line extending in the direction of the gradient vector (g), the first gradient line starting from a first starting point (p start,1 ) and terminates at the first endpoint (p end,1 ), Wherein each voxel (v E1 ) is increased by a predefined constant increment, each voxel (v E1 ) intersects the corresponding voxel (v) of the image (M) of FIG. - Provided by voxels (v E2 ) is an empty second gradient accumulator image (E2), each voxel (v E2 ) is associated with a voxel (v) of said image (M) and a gradient vector (g) pointing in the direction of the maximum possible intensity increase towards said image (M) is calculated for each voxel (v) of said image (M) and a second gradient line extending in the opposite direction of the gradient vector (g), the second gradient line starting from a second starting point (p start,2 ) and terminates at the second endpoint (p end,2 ), Wherein each voxel (v E2 ) is increased by a predefined constant increment, each voxel (v E2 ) intersects the corresponding voxel (v) of the image (M) of FIG. - Provided by voxels (v E3 ) is an empty third gradient accumulator image (E3) composed of each voxel (v E3 ) is associated with a voxel (v) of the image (M), and for each voxel (v) of the image (M) a gradient vector (g) pointing in a direction of maximum possible intensity increase towards the image (M) and a third gradient line extending in the direction of the gradient vector (g), the third gradient line starting from a third starting point (p start,3 ) and terminates at the third endpoint (p end,3 ), wherein each voxel (v E3 ) is increased by a predefined constant increment, each voxel (v E3 ) intersects the corresponding voxel (v) of the image (M) of FIG. 1 , and - using the first, second and third gradient accumulator images (E1, E2, E3) to form a cochlear descriptor image (E) composed of voxels (vr) descriptor ), and select the cochlear descriptor image (E descriptor ) has a value greater than the cochlear descriptor image (E descriptor ) as a prediction of the position of the center (11) of the cochlea (10) in the image (M) by using the intensity values ​​of all other voxels of the voxel (vc) whose number is P ≥ 1.

2. The method according to claim 1, wherein: Forming the cochlear descriptor image (E descriptor ) corresponds to calculating the Hadamard product of the first, second and third gradient accumulator images (E1, E2, E3) 3. The method according to one of the preceding claims, wherein: The corresponding gradient line includes the line extending from the corresponding starting point (pstart,1, pstart,2, pstart,3) to the corresponding end point (pend,1, pend,2, p end,3 ), wherein the length of each first gradient line is greater than the length of each third gradient line, and wherein the length of each third gradient line is greater than the length of each second gradient line.

4. The method according to one of the preceding claims, wherein: The corresponding first, second and third starting points p are calculated according to the following formula start,1 、p start,2 、p start,3 : p start,i =position(v)+start i ·g i=1,2,3 And wherein the corresponding first, second and third endpoints pend,1, pend,2 and pend,3 are calculated according to the following formula: p end,i =position(v)+end i ·g i=1,2,3 wherein position(v) represents the position of the receiving voxel (v) of the image (M), and wherein - start1 is in the range from 2.7 mm to 3.3 mm, wherein in particular start1 is equal to 3.0 mm, -start2 is in the range from -3.3 mm to -2.7 mm, where in particular start1 is equal to -3.0 mm, - start3 is in the range from 0.45 mm to 0.55 mm, wherein in particular start3 is equal to 0.5 mm, - end1 is in the range from 4.5 mm to 5.5 mm, wherein in particular end1 is equal to 5.0 mm, -end2 is in the range from -1.8 mm to -2.2 mm, wherein in particular end2 is equal to -2.0 mm, - end3 is in the range from 1.8 mm to 2.2 mm, wherein in particular end3 is equal to 2.0 mm.

5. The method according to one of the preceding claims, wherein: The method further comprises the steps of: - providing a statistical shape model (SSM) of the ear, said statistical shape model (SSM) comprising at least a cochlear portion corresponding to said cochlea (10) of said ear.

6. The method according to claim 5, wherein: The method further comprises the steps of: - For each voxel (Vc) of said number P of voxels: o defining a volume (B1) in said image (M) having said corresponding voxel (Vc) as centre, o segmenting the cochlea (10) within the volume (B1) with the help of a Hessian-based enhancement filter, thereby obtaining segmented cochlear regions of the image (M), wherein a 3D surface of the cochlea is generated from the segmented cochlear regions, in particular by using a marching cubes algorithm, o fitting the cochlear portion of the statistical shape model (SSM) to the 3D surface, o calculating a surface similarity measure indicating the similarity between a surface of the cochlear portion of the statistical shape model (SSM) and the 3D surface, - Determine the voxel (v c ) as the center (11) of the cochlea (10), for which the surface similarity measure satisfies a predefined criterion, wherein in particular the first voxel (v c ) is selected as the center of the cochlea for which the surface similarity measure falls below a predefined threshold.

7. The method according to claim 6, wherein: The method further comprises the following steps: - fitting the cochlear portion of the statistical shape model (SMM) to the segmented cochlear region using the determined center (11) of the cochlea (10), the statistical shape model (SSM) further comprising semicircular canal portions corresponding to the three semicircular canals (12) of the ear, - predicting the positions of the three semicircular canals (12) in the image (M) as the positions of the three semicircular canals (12) in the semicircular canal portion of the statistical shape model (SSM), - defining a volume (B2) including the predicted location, and - Segmenting the three semicircular canals (12) within the volume (B2) thereby obtaining segmented semicircular canal regions of the image (M).

8. The method according to claim 7, wherein: The method further comprises the steps of: - fitting the cochlear portion and the semicircular canal portion of the statistical shape model (SSM) to the segmented cochlear region and to the segmented semicircular canal region.

9. The method according to claim 8, wherein: The method further comprises the steps of: - predicting the positions of the incus (5), the malleus (6) and the stapes (7) in the image (M) as the positions of the incus (5), the malleus (6) and the stapes (7) in the ossicular part of the statistical shape model (SSM), the ossicular part of the statistical shape model corresponding to the incus (5), the malleus (6) and the stapes (7) of the ear, - defining a volume (B4) including the predicted position, and - segmenting the incus, malleus and stapes within the volume, thereby obtaining a segmented ossicular region of the image (M) corresponding to the incus, malleus and stapes in the image (M).

10. The method according to claim 8 or 9, wherein: The method further comprises the steps of: - predicting the position of an eardrum (13) as the position of the eardrum (13) in an eardrum portion of the statistical shape model (SSM), the eardrum portion of the statistical shape model corresponding to the eardrum (13) of the ear, - defining a volume (B3) including the predicted location, and - Segmenting the eardrum (13) within the volume (B3) thereby obtaining a segmented eardrum region of the image (M).

11. The method according to claim 10, wherein: The method further comprises the steps of: - predicting the area, orientation and limits of the position of the external auditory canal (14) based on the segmented eardrum area of ​​the image (M), - define the volume (B5) of the area including the prediction, - segmenting the external auditory canal (14) within the volume (B5), thereby obtaining a segmented external auditory canal region of the image (M), And wherein the method further comprises the following steps: - predicting the position and orientation of the outer surface (15) of the temporal bone (16) based on said segmented external auditory canal region of said image (M), - defining a volume (B6) including the predicted position, and - Segmenting the temporal bone (16) within the volume (B6) with the help of thresholding, thereby obtaining a segmented temporal bone region of the image (M), and using the direction of the external auditory canal (14) to detect the external surface.

12. The method according to claim 6 or according to one of claims 7 to 11 when referring back to claim 7, wherein: The method further comprises the steps of: - predicting the location of the internal auditory canal (17) based on the segmented cochlear region of the image, and - defining a volume (B7) including the predicted position, and - Segmenting the internal auditory canal (17) within the volume (B7) thereby obtaining a segmented internal auditory canal region of the image (M).

13. The method according to claims 6, 8 and 11, wherein: The method further comprises the steps of: - predicting the position of the facial nerve (18) based on the segmented cochlear region, the segmented semicircular canal region and the segmented external auditory canal region of the image (M), - defining a volume (B8) including the predicted position, and - Segmenting the facial nerve (18) within the volume (B8) thereby obtaining a segmented facial nerve region of the image (M).

14. The method according to one of claims 6 to 13, wherein: The method further includes the step of graphically visualizing for a user on a display (102) at least one of the following: the segmented cochlear region, the segmented semicircular canal region, the segmented ossicular region, the segmented tympanic membrane region, the segmented external auditory canal region, the segmented temporal bone region, the segmented internal auditory canal region, and the segmented facial nerve region.

15. A system (100) for automatically segmenting a human cochlea (10) comprised in an input image (M) consisting of voxels (v), each voxel comprising an intensity value indicating the intensity of the respective voxel, wherein: The system (100) comprises at least one processor (101) and a display (102), which are adapted to perform the method according to one of claims 1 to 14.

Images