Ultrasound Data Segmentation

Abstract

A method for segmenting a target anatomical structure in ultrasound data. Scan-converted ultrasound data is obtained in a scan-converted space in a Cartesian coordinate system. The scan-converted ultrasound data is transformed into descanned ultrasound data in a descanned space in a toroidal coordinate system. The descanned ultrasound data is an estimate of the ultrasound data obtained in the original acquisition procedure. Segmentation of the target anatomical structure can then be performed on the ultrasound data in the descanned space. The resulting segmented data is rescanned back to the Cartesian coordinate system and displayed with the ultrasound data.

Description

[Technical Field]

[0001] The present invention relates to methods for improving ultrasound imaging, and in particular to anatomy segmentation of ultrasound data. [Background technology]

[0002] Image segmentation is used in digital image processing to divide a digital image or volume into multiple segments (e.g., groups of pixels / voxels), each covering a region of the image or volume. Typically, the various segments are identified and distinguished based on similar shared graphical properties (e.g., color, intensity, or texture) of the pixels / voxels contained in the segmented region.

[0003] In the field of medical imaging, segmentation helps to identify various anatomical structures or features, such as organs, air and fluid passages (e.g., blood vessels or the digestive tract), valves, or cavities. The resulting segment contours allow for faster assessment of the medical image or volume by a clinician (enabling faster diagnostic or treatment decisions). Segmentation also helps feed into subsequent processing techniques.

[0004] One important area of ​​medical imaging is ultrasound imaging, which plays a key role in the evaluation and diagnosis of large areas of the body. Diagnosis, treatment, and planning are often based on clearly delineating relevant anatomical sites or features within ultrasound images, and segmentation can be used to automate this delineation process.

[0005] Automated segmentation and quantification of organs from 3D ultrasound volumes is an active area of ​​research but has not yet reached clinical routine. Several methods have been proposed, including the use of classification techniques and landmark-based approaches. However, these methods are often inaccurate and typically contain segmentation leaks.

[0006] Automated segmentation of 3D organs from 3D ultrasound arrays presents a major problem related to image acquisition: the acquired volumes are stored as rectangular volumes, which contain information only within a so-called cone shape, resulting in an inefficient representation of the target anatomical structures. Summary of the Invention [Problem to be solved by the invention]

[0007] More precisely, ultrasound data acquisition methods have the following drawbacks, which affect the training of automated segmentation algorithms, especially when deep learning is required: (i) Inefficient use of memory due to the need to store data outside the cone shape. When using current AI methods for segmentation, memory space is a key factor for success. In some acquired ultrasound volumes, up to 60% of voxels are black and represent no information (these are outside the cone-shaped beam range). (ii) There is a learning bias due to the location of the acquisition cone boundary. (iii) Data augmentation requires complex transformations. Data augmentation is an essential component of deep learning to achieve high performance. Indeed, deep learning requires large and abundant training datasets to enable application to new images and avoid overfitting. In the absence of a large dataset, a well-established technique is to apply geometric / grayscale transformations to the training dataset to generate more training images. To preserve the realistic aspects of the image, including the cone shape, while mimicking the possible presence of organs elsewhere in the image, complex transformations must be carefully applied to the Cartesian domain. While traditional image processing approaches can use masks to ignore information outside the cone, this is not feasible using new segmentation techniques involving deep learning.

[0008] Therefore, a method to overcome these acquisition shortcomings is needed with the ultimate goal of training automatic segmentation algorithms in ultrasound imaging.

[0009] CN107909585 discloses a method for segmenting the vascular intima of intravascular ultrasound images.

[0010] Lo Vercio, Lucas et al., "Assessment of image features for vessel wall segmentation in intravascular ultrasound images," International Journal of Computer Assisted Radiology and Surgery, Vol. 11, No. 8, pp. 1397-1407 (XP036013989), discloses a method for segmenting blood vessel walls from intravascular ultrasound images.

[0011] Nicolas Toussaint et al., "Weakly supervised localization for fetal ultrasound images," Arxiv.org, Cornell University Library (XP080901171), discloses a method for detecting and localizing fetal anatomical regions within 2D ultrasound images. [Means for solving the problem]

[0012] The invention is defined by the claims.

[0013] According to an example according to one aspect of the present invention, there is provided a method for segmenting a target anatomical structure in a 3D ultrasound volume, the method comprising: obtaining a scan-converted ultrasound volume in a scan-converted space in a Cartesian coordinate system; transforming the scan-converted ultrasound volume into a descanned ultrasound volume in descanned space in a toroidal coordinate system, the descanned ultrasound volume comprising an estimate of the ultrasound volume obtained by the original acquisition procedure; and performing segmentation of the target anatomical structure in the ultrasound volume in the descanned space to generate segmentation data.

[0014] Ultrasound volumes typically contain "empty" data that lacks meaningful information. The ultrasound images used to generate the ultrasound volume are typically represented as cone-shaped images when acquired using phased arrays or curved arrays, but are displayed on a rectangular (or square) screen. Therefore, the majority of the pixels (or voxels in a 3D volume) are empty and contain no meaningful information.

[0015] The descanned space represents the data size required to store the scan-converted ultrasound volume in the original coordinate system. For example, in Cartesian coordinates, the image data is stored in a matrix, with each matrix element representing the color of a pixel in RGB. However, in the case of a scan-converted ultrasound image, the significant data due to the acquisition of the ultrasound volume is cone-shaped, so a certain percentage of the data stored in the matrix (for scan-converted ultrasound images) is null data.

[0016] The ultrasound volume is acquired in scan-transformed space (e.g., a cube containing the ultrasound volume cone) in Cartesian coordinate space with null data, as this represents how the image will actually be seen, and therefore is more intuitive for the clinician.

[0017] However, due to the presence of null data, segmentation algorithms may, during training, give more weight to regions with cone shapes and null data rather than segmenting the target anatomical structures. Therefore, scan-converted ultrasound volumes are "descanned." This involves performing an inverse scan conversion on the scan-converted ultrasound volume. This is done by transforming the scan-converted ultrasound volume into a toroidal coordinate system (hence considered a descanned coordinate system) that represents how the ultrasound volume was acquired. By transforming the scan-converted ultrasound volume into a coordinate system that represents how it was acquired, the percentage of null data can be significantly reduced.

[0018] A segmentation algorithm can then be trained and applied to the descanned ultrasound volume in descanned space with better accuracy and precision than in acquisition space.

[0019] Segmentation methods are, for example, based on convolutional neural networks. Descan conversion methods improve convolutional neural network-based segmentation methods by using memory efficiently.

[0020] The method may further include, after segmentation has been performed, transforming the descanned ultrasound volume in the descanned space back to the scan-converted space.

[0021] Once the descanned ultrasound volume has been segmented for the target anatomical structure (e.g., kidney, fetus), the segmentation data (in descanned space) can be transformed back to a Cartesian coordinate system so that it can be overlaid on the original image. The segmentation transformation thus provides visualization of the segmented portions in their correct Cartesian geometric configuration. In this way, the clinician can analyze the ultrasound volume without warping, yet still have the advantage of the data being segmented (e.g., to calculate the area / volume of the segmented target geometric structure).

[0022] The above method is a scan-converted ultrasound volume in scan-converted space; a descanned ultrasound volume in descanned space; a segmented ultrasound volume in descanned space, or The method may further include displaying one or more of the segmented ultrasound volumes in the scan-converted space.

[0023] The method may further include determining a volume of the target anatomical structure based on the segmentation of the target anatomical structure.

[0024] The descanned coordinate system is a toroidal coordinate system.

[0025] The toroidal coordinate system is used because it best describes the acquisition of an ultrasound volume. For example, in phased arrays and curved arrays, the ultrasound probe rotates around the center of the rotor (i.e., the motor that spins the ultrasound probe) as the 3D volume is acquired. Therefore, the toroidal coordinate system results in a low percentage of null data.

[0026] The method may further include estimating an acquisition geometry of the ultrasound volume to derive a transformation for converting the scan converted ultrasound volume to a descanned ultrasound volume, wherein estimating the acquisition geometry comprises: estimating the optical center of acquisition; estimating the mechanical center of acquisition; estimating the minimum depth of acquisition; estimating the maximum depth of acquisition; Estimating the angle of view, and and estimating the acquisition sweep angle.

[0027] In typical imaging acquisitions, the imaging lines of the scan-converted ultrasound volume may not be stored with the scan-converted ultrasound volume. Therefore, it may not be possible to perform descanning directly because the acquisition geometry required to perform the transformation is not present. In this case, the acquisition geometry must be estimated from the scan-converted ultrasound volume before transforming the scan-converted ultrasound volume into descanned space.

[0028] If the optical and mechanical centers are separate, a toroidal transformation results.

[0029] This estimate can be obtained by image analysis of a scan-converted ultrasound volume in a Cartesian coordinate system.

[0030] The present invention also provides a computer program product comprising computer program code means which, when executed on a computing device having a processing system, causes the processing system to perform all the steps of the above method.

[0031] The present invention also provides a system for segmenting a target anatomical structure within a descanned acoustic volume, the system comprising: receiving a scan-converted ultrasound volume in a scan-converted space in a Cartesian coordinate system; Transforming the scan-converted ultrasound volume into a descanned ultrasound volume in a descanned space in a toroidal coordinate system, the descanned ultrasound volume comprising an estimate of the ultrasound volume obtained by the original acquisition procedure; and performing segmentation of the target anatomical structure in the ultrasound volume in the descanned space to generate segmentation data.

[0032] The processor may be further configured to convert the segmentation data (in descanned space) back to scan-converted space after segmentation has been performed.

[0033] The above system is an ultrasound volume in scan-converted space; ultrasound volume in descanned space, a segmented ultrasound volume in descanned space, or The method may further include displaying one or more of the segmented ultrasound volumes in the scan-converted space.

[0034] The processor may be further configured to determine a volume of the target anatomical structure based on the segmentation of the target anatomical structure.

[0035] The descanned coordinate system is a toroidal coordinate system.

[0036] The processor may be configured to transform the ultrasound volume to a descanned space in a toroidal coordinate system by estimating an acquisition geometry of the ultrasound volume to derive a transformation for converting the scan converted ultrasound volume to a descanned ultrasound volume, and the processor may be configured to: estimating the optical center of acquisition; estimating the mechanical center of acquisition; estimating the minimum depth of acquisition; estimating the maximum depth of acquisition; Estimating the angle of view, and The method is configured to estimate an acquisition geometry based on one or more of estimating the acquisition sweep angle.

[0037] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. [Brief explanation of the drawings]

[0038] For a better understanding of the present invention, and to show more clearly how it may be carried into effect, reference will now be made, by way of example only, to the accompanying drawings in which:

[0039] [Figure 1] FIG. 1 shows an ultrasound image in a rectangular scan-converted space. [Figure 2] Figure 2 shows a histogram of the percentage of significant data in a large database of volumetric kidney acquisitions. [Figure 3] FIG. 3 shows a flow diagram of a method for performing segmentation. [Figure 4] FIG. 4 shows the ultrasound volume transformed into polar coordinates. [Figure 5] FIG. 5 shows an ultrasound volume in scan-converted space. [Figure 6] FIG. 6 shows a toroid in Cartesian and toroidal coordinate systems. [Figure 7] FIG. 7 shows the ultrasound volume in spherical coordinates. [Figure 8] FIG. 8 shows organ segmentation in the descanned and acquisition coordinate systems. [Figure 9] FIG. 9 shows the results of organ segmentation in the toroidal and acquisition coordinate systems. DETAILED DESCRIPTION OF THE INVENTION

[0040] The present invention will now be described with reference to the figures.

[0041] It should be understood that the detailed description and specific examples, while indicating exemplary embodiments of the devices, systems, and methods, are for purposes of illustration only and are not intended to limit the scope of the invention. These and other features, aspects, and advantages of the devices, systems, and methods of the present invention will become better understood from the following description, the appended claims, and the accompanying drawings. It should be understood that the figures are schematic representations only and are not drawn to scale. It should also be understood that the same reference numerals are used throughout the figures to indicate the same or similar parts.

[0042] The present invention provides a method for segmenting a target anatomical structure in ultrasound data. Scan-converted ultrasound data is acquired in a scan-converted space in a Cartesian coordinate system. The scan-converted ultrasound data is transformed into descanned ultrasound data in a descanned space in a toroidal coordinate system. The descanned ultrasound data is an estimate of the ultrasound data obtained in the original acquisition procedure. Segmentation of the target anatomical structure can then be performed on the ultrasound data in the descanned space. The resulting segmented data is rescanned back to the Cartesian coordinate system and displayed with the ultrasound data.

[0043] FIG. 1 illustrates scan-converted ultrasound data 102 in a rectangular scan-converted space. Scanlines are typically used to create visible ultrasound data 102 (e.g., an image or volume) from raw data based on how the ultrasound data 102 was acquired (based on the acquisition procedure), thereby allowing the data to be visualized in a true Cartesian domain where distances and angles can be measured. To this end, raw data captured by an ultrasound probe is scan-converted into Cartesian coordinates using scanlines to provide a "real" view of the ultrasound data. Furthermore, scan-converted ultrasound data 102 is typically stored in scan-converted space (Cartesian coordinates), and any analysis or data processing performed after acquisition typically must be performed in scan-converted space.

[0044] Scan-converted ultrasound data 102 in Cartesian coordinates typically contains a cone-shaped cone of meaningful data 104 surrounded by "empty" (or meaningless) data 106. Although the majority of scan-converted ultrasound data 102 does not contain meaningful data 104, the nature of image data storage in Cartesian coordinates (typically image bitmaps) requires that all scan-converted space be stored.

[0045] 3D scan-converted ultrasound volumes typically contain a significant proportion of empty data 106 (black voxels) that is not needed for clinical information but is necessary to store the scan-converted ultrasound volume in Cartesian coordinates.

[0046] Figure 2 shows a histogram of the proportion of significant data 104 in a large database of volumetric kidney acquisitions. The large database of volumetric kidney acquisitions contained 621 volumes. The ratio of volumes within the imaging cone (significant data 104) to the total Cartesian volume was calculated for each volumetric kidney acquisition (in percentage %).

[0047] The X-axis shows the ratio (in percentage %, 0.4% scale) of significant data 104 to the total data stored in the scan-converted space, and the Y-axis shows the number of cases in each ratio range.

[0048] Thus, the ratio ranges from 35.6 to 44.8%, with an average ratio of significant data 104 of approximately 38%. Therefore, the ratio of empty data 106 (necessary only for storing data in Cartesian coordinates) in the volumetric kidney acquisition is approximately 62%.

[0049] This means that on average 62% of all data stored (for volumetric kidney acquisitions) is essentially wasted storage space. While data compression can be used to make data storage more efficient, voxels still require memory storage when used in deep learning algorithms. Furthermore, machine learning algorithms (e.g., segmentation algorithms) trained using this data will assign an undue importance to null data 106 due to the high proportion of null data 106 in the transformed ultrasound data 102.

[0050] 3 shows a flow diagram of a method for performing segmentation on a 3D ultrasound volume. Once the 3D ultrasound volume is acquired (step 302), the acquisition geometry is estimated (step 304), if unknown. The acquisition geometry may include estimating the scanlines of the ultrasound data and finding the optical and mechanical centers of the acquisition process.

[0051] Estimating the acquisition geometry (step 304) is based on an estimate of how the ultrasound data was acquired. For example, if an ultrasound volume was acquired using a phased array ultrasound probe rotated around the subject's skin, the ultrasound volume will have a minimum and maximum depth (relative to the ultrasound transducer array), an optical center (where the ultrasound transducers are located), a mechanical center (if the ultrasound probe rotor is present), an angle of view (based on the angle of the ultrasound image acquired by the ultrasound probe), and a sweep angle (based on the rotation of the ultrasound probe around the skin by the ultrasound probe rotor). These acquisition geometry quantities can be estimated from the ultrasound volume. The acquisition geometry can be used to "descan" the volume (step 306).

[0052] De-scanning is the process of "undoing" the scan conversion process to eliminate as much null data as possible. This is done by choosing a coordinate system that simulates the acquisition process. The toroidal coordinate system is a good approximation because it requires two centers (corresponding to the optical and mechanical centers of acquisition in ultrasound imaging).

[0053] Once the volume is descanned, a segmentation algorithm is applied to the volume to segment target anatomical structures (step 308). Working in a toroidal coordinate system for the segmentation process is more powerful than other coordinate systems and provides better performance for automated segmentation. Target anatomical structures include one or more of organs, blood vessels, ventricles, bones, muscles, etc.

[0054] The segmentation algorithm is also trained on the descanned volume. The segmentation data is then converted back to a Cartesian coordinate system (rescanned) (step 310), which makes viewing the volume (e.g., on a display) more intuitive for the physician. For example, the original volume is shown along with the rescanned segmentation data. Optionally, the image data of the original volume can be rescanned along with the segmentation data, but it is preferable to use the original image data (without descanning and rescanning).

[0055] 4 shows an example of how to descan a scan converted ultrasound image 102a for purposes of explaining the general steps. This example uses a 2D ultrasound image rather than an ultrasound volume to simplify the example while illustrating the descanning steps.

[0056] FIG. 4a) shows a scan-converted ultrasound image 102a in Cartesian coordinates. Null data 106 is required to store the meaningful image 104a in a rectangular format, such as a bitmap. A target anatomy 402 is shown to help explain how the descanning process works.

[0057] FIG. 4b) shows a scan-converted ultrasound image 102a in which an imaging line 404 has been interpolated. The imaging line 404 may be stored along with the scan-converted ultrasound image 102a. However, in most cases, the stored scan-converted ultrasound data 102 does not include data for the imaging line 404. In this case, the imaging line 404 can be estimated from the scan-converted ultrasound image 102a. Two imaging lines 404 can be estimated from the side edges of the scan-converted ultrasound image 102a. From these imaging lines 404, an optical center 406 can be estimated and additional imaging lines 404 can be constructed. Additionally, from the imaging lines 404, a field angle can be estimated as the angle between the two furthest imaging lines 404 determined by the edges.

[0058] FIG. 4c) shows a descanned ultrasound image 408 in polar coordinates (again, because a simplified 2D image is used as an example). Therefore, the radial component is based on the distance of each data point from the optical center 406 found in FIG. 4b). The angular component of each data point is also found by measuring the angle between the imaging line containing the data point and the edge imaging line of the scan-converted ultrasound image 102a in Cartesian coordinates. Thus, the descanned ultrasound image 408 does not contain null data 106, but can be stored in a rectangular format (e.g., an RGB matrix bitmap). Furthermore, the target anatomical structure 402 represents a higher proportion of the total data (compared to the scan-converted ultrasound image 102a).

[0059] The descanned coordinate system used is instead a toroidal coordinate system. FIG. 5 shows a portion of a toroid 502 in Cartesian and toroidal coordinate systems. This transformation is inspired by the acquisition process, as it performs a type of descan transformation, which may be considered a dedicated preprocessing step for the segmentation algorithm. To mimic ultrasound lines that may generate similar volumes, a Caesar-to-toroidal transformation is applied to the scan-converted ultrasound data 102 using the acquisition geometric configuration parameters. The toroidal coordinate system can be used to represent meaningful 3D ultrasound data 104b.

[0060] A 3D ultrasound volume is typically determined from many 2D ultrasound images. For each 2D ultrasound image (at least two), the optical center 406 can be obtained. The optical centers create a section of a circle, and the center of the circle is the mechanical center 504.

[0061] Once the mechanical center 504 and the optical center 406 are known, the distance between them can be calculated (-R0). Thus, Caesar coordinates (x,y,z) can be transformed into toroidal coordinates (r,φ,θ):

number

[0062] Using equations (1) and (2), we can transform the volume from Caesar coordinates to toroidal coordinates. In the acquisition volume, points are named (x,y,z). To find the corresponding intensity at point (r,φ,θ) in the transformed domain, the above equations calculate (r,φ,θ) as a function of (x,y,z), where (r,φ,θ)∈[R,Θ,Φ].

[0063] [R, Θ, Φ] is the range of values ​​of the parameters selected for descanning. This defines the size of the descanned volume that is computed as input to the segmentation training and testing algorithms.

[0064] The true values ​​of these parameters are usually unknown, since the pre-scan conversion data is not stored. Therefore, these values ​​can be set freely, which is an advantage in learning approaches. For example, for a rough and fast algorithm, small values ​​such as (64,64,64) can be chosen. In practice, it is preferable to choose a larger range of values ​​so that the deep learning algorithm can be trained in memory. It is also possible to define non-identical values ​​for these parameters, such as (128,128,64), which may reflect proportional distances in the original acquired volume.

[0065] Therefore, we can choose multiple of these parameters and apply descanning, which means that when training a network for a given organ, we can scale our method to the available memory on the computer. At the end of this step, we have a 3D dataset with dimensions (R, Θ, Φ), where each voxel carries important information.

[0066] The minimum and maximum values ​​of r can be calculated from the minimum and maximum depth of acquisition, respectively, the value of θ can be calculated from the angle of view, and the value of φ can be calculated from the sweep angle.

[0067] Once the descanned ultrasound data 408 has been segmented in descanned space, it can be converted back to scan-converted space in Cartesian coordinates using equations (3) and (4).

number

[0068] 6 shows the scan-converted ultrasound volume 102b in scan-converted space. The white cones represent significant data 104b, and the dotted space represents zero values ​​106 that encode no significant information. The large dotted circle outside the scan-converted space represents the mechanical center 504, and the small black circle represents the optical center 406 of the A-plane.

[0069] Using the meaningful data 104b, parameters of the acquisition geometry are estimated, which represent the center of the A-plane imaging beam (optical center 406), the minimum and maximum imaging depth, the field angle at the A-plane, the center of the sweep for imaging other planes (mechanical center 504), and the corresponding sweep angle.

[0070] FIG. 7 illustrates the ultrasound volume 104b in spherical coordinates. A spherical coordinate system is an extreme case of a toroidal coordinate system in which the optical center 406 and the mechanical center 504 are co-located (or approximated as being co-located). In this case, the ultrasound volume 104b can be completely defined in the spherical coordinate system without the need for null data. However, using a spherical coordinate system is not as accurate as using a toroidal coordinate system because it assumes that the optical center 406 and the mechanical center 504 are co-located (which may be unrealistic). Therefore, using a toroidal coordinate system allows for scanned data with different optical and mechanical centers, allowing for effective manipulation of various initial scan conditions.

[0071] FIG. 8 shows organ segmentation 806 in the descanned and acquisition coordinate systems. The image on the left 802 shows an ultrasound image from an ultrasound volume with kidney segmentation in Cartesian coordinates. The image on the right 804 shows a descanned ultrasound image from an ultrasound volume with the corresponding descanned segmentation after the ultrasound volume has been transformed to a toroidal coordinate system. In the image on the right 804, the horizontal axis represents the radius r of the toroidal coordinate system, and the vertical axis represents the poloidal angle θ. The descanned segmentation can be generated for ground truth segmentation and used for network training.

[0072] Training a segmentation algorithm requires a large amount of data. Data augmentation is a method used to create a large amount of data from an initial small amount of data. For example, geometric transformations (e.g., rotation and translation) and color transformations (e.g., grayscale transformation) can be applied to a single ultrasound image / volume to create multiple training images.

[0073] Descanned ultrasound data in a toroidal coordinate system allows for more intuitive and less error-prone geometric transformations to create large databases for training.

[0074] Furthermore, when training a segmentation algorithm, data bias (e.g., data shape) should be avoided if possible. In scan-converted ultrasound data, the cone shape containing significant data is a source of data bias for segmentation algorithms. Segmentation algorithms are also trained with null (meaningless) data. This allows the segmentation algorithm to ignore certain sections of data for future segmentations or to search for certain patterns in the null data.

[0075] The descanned ultrasound volume (e.g., in a toroidal coordinate system) can be completely described without the need to store null data, thus eliminating data bias due to null data.

[0076] This algorithm can be any type of standard deep learning approach, such as U-Net segmentation. At the output of the segmentation, the segmented ultrasound data can be transformed back to a Cartesian coordinate system. For example, the acquisition geometry is used again in the inverse transformation to transform the segmentation mask back from the toroidal coordinate system to the image Cartesian coordinate system.

[0077] FIG. 9 shows the results of organ segmentation 908 in a toroidal coordinate system 904 and a Cartesian coordinate system 902. Similarly, an image segmented without coordinate transformation 902 is shown on the left, and an image segmented using coordinate transformation to the descanned domain 904 is shown on the right. Using a descanned ultrasound image 904 yields more accurate results. The approximate ground truth 906 is shown as a black circle, and the results of the segmentation algorithm 908 are shown as a white perimeter. The striped arrow 910 indicates segmentation leakage in the scan-converted ultrasound image 902, resulting from no descan conversion. A corresponding arrow 912 is shown in the descan-converted image, demonstrating the significant reduction in segmentation leakage in the descanned ultrasound image 904.

[0078] This method has also been successfully applied in the field of fetal abdomen and adult kidney segmentation. The results obtained with this method significantly outperform approaches that do not use transformation to the toroidal domain. The following table shows the results, which show a significant improvement over traditional segmentation (without descanning): [Table 1]

[0079] As previously mentioned, systems use processors to process data. Processors can be implemented in a variety of ways using software or hardware to perform the various functions required. Typically, a processor uses one or more microprocessors that are programmed using software (e.g., microcode) to perform the required functions. A processor can be implemented as a combination of dedicated hardware to perform some functions and one or more programmed microprocessors and associated circuitry to perform other functions.

[0080] Examples of circuitry that may be used in various embodiments of the present disclosure include, but are not limited to, conventional microprocessors, application specific integrated circuits (ASICs), and field programmable gate arrays (FPGAs).

[0081] In various implementations, a processor may be associated with one or more storage media, such as volatile and non-volatile computer memory, such as RAM, PROM, EPROM, and EEPROM. The storage media may be encoded with one or more programs that, when executed on the one or more processors and / or controllers, perform the necessary functions. The various storage media may be fixed within the processor or controller, or may be transportable so that one or more programs stored thereon can be loaded into the processor.

[0082] Variations of the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the singular does not exclude a plurality.

[0083] A single processor or other unit may fulfill the functions of several items recited in the claims.

[0084] The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.

[0085] The computer program may be stored / distributed on any suitable medium, such as an optical storage medium or a solid-state medium, supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless communication systems.

[0086] It should be noted that when the term "adapted to" is used in the claims or description, it is intended to be equivalent to the term "configured to."

[0087] Any reference signs in the claims should not be construed as limiting the scope.

Claims

1. 1. A method for segmenting a target anatomical structure in a 3D ultrasound volume, comprising: obtaining a scan-converted ultrasound volume in a scan-converted space in a Cartesian coordinate system; Transforming the scan-converted ultrasound volume into a descanned ultrasound volume in a descanned space of a toroidal coordinate system, the two centers of the toroidal coordinate system being at different locations and corresponding to the optical center and mechanical center, respectively, of the ultrasound volume obtained by the original acquisition procedure, thereby reducing the proportion of null data; performing segmentation of a target anatomical structure in the ultrasound volume in the descanned space to generate segmentation data; A method comprising:

2. The method of claim 1 , further comprising the step of transforming the segmentation back into the scan-converted space after the segmentation has been performed.

3. the ultrasound volume in the scan-converted space; the ultrasound volume in the descanned space; a segmented ultrasound volume in the descanned space; or The method of claim 2 , further comprising displaying one or more of the segmented ultrasound volumes in the scan-converted space.

4. The method of claim 1 , further comprising determining a volume of the target anatomical structure based on the segmentation of the target anatomical structure.

5. further comprising estimating an acquisition geometry of the ultrasound volume to derive a transformation for converting the scan converted ultrasound volume to a descanned ultrasound volume, the estimating the acquisition geometry comprising: estimating the optical center of acquisition; estimating the mechanical center of acquisition; estimating a minimum depth of acquisition; estimating the maximum depth of acquisition; estimating the angle of view; and 5. The method of claim 1, further comprising one or more of the steps of: estimating an acquisition sweep angle;

6. A computer program comprising computer program code means which, when executed on a computing device having a processing system, causes said processing system to perform all the steps of the method of any one of claims 1 to 5.

7. 1. A system for segmenting a target anatomical structure in a 3D ultrasound volume, comprising: The system includes a processor, the processor comprising: receiving a scan-converted ultrasound volume in a scan-converted space in a Cartesian coordinate system; Transforming the scan-converted ultrasound volume into a descanned ultrasound volume in a descanned space of a toroidal coordinate system, the two centers of the toroidal coordinate system being at different locations and corresponding to the optical center and mechanical center, respectively, of the ultrasound volume obtained by the original acquisition procedure, thereby reducing the proportion of null data; performing a segmentation of a target anatomical structure in the ultrasound volume in the descanned space to generate segmentation data.

8. The system of claim 7 , wherein the processor transforms the segmentation data back to the scan-converted space after the segmentation is performed.

9. the ultrasound volume in the scan-converted space; the ultrasound volume in the descanned space; a segmented ultrasound volume in the descanned space; or The system of claim 8 , further comprising a display that displays one or more of the segmented ultrasound volume in the scan-converted space.

10. The system of claim 7 , wherein the processor determines a volume of the target anatomical structure based on the segmentation of the target anatomical structure.

11. the processor transforms the ultrasound volume to a descanned space in the toroidal coordinate system by estimating an acquisition geometry of the ultrasound volume to derive a transformation for converting the scan converted ultrasound volume to a descanned ultrasound volume; The processor: estimating the optical center of acquisition; estimating the mechanical center of acquisition; estimating the minimum depth of acquisition; estimating the maximum depth of acquisition; Estimating the angle of view, and estimating an acquisition sweep angle.

Images