Organ segmentation in an image

Abstract

Described herein are devices, systems and methods for organ mask generation, including generating a composite centroid mask, identifying first and second intensity thresholds, in a first segmentation pass (i) setting to zero pixels of the image having an intensity below the first threshold and (ii) setting to zero pixels of the image corresponding to objects whose centroids are outside the composite centroid mask, resulting in an initial organ mask, in a second segmentation pass setting to zero pixels (i) having an intensity below a second threshold that is less than the first threshold and (ii) setting to zero pixels corresponding to objects whose centroids are outside the initial organ mask, resulting in a second organ mask, and expanding and filling the second organ mask to generate the organ mask.

Description

Technical Field

[0001] Priority Claim This application claims the benefit of U.S. Provisional Patent Application No. 63 / 197,877, filed on Jun. 7, 2021, which is hereby incorporated by reference in its entirety.

[0002] The embodiments described herein relate to medical images and devices, systems, methods, and machine-readable media for isolating organs within medical images.

Background Art

[0003] Medical image scans, such as magnetic resonance imaging (MRI) scans and X-ray computed tomography (CT or CAT) scans, are procedures that can be used to obtain information about the internal structure of an object such as a patient. Medical image scans can be used to detect signs of cancer. Cancer in some organs, such as the pancreas, can be difficult to detect with medical image scans due to the position of the organ within the body and the homogeneity of the surrounding tissue.

[0004] Finding an organ such as the pancreas in a medical image scan may be part of a process for assessing its health. The process of finding the organ can be time-consuming for a person such as a radiologist viewing the scan, and it can be difficult for the radiologist to reliably find the boundaries of the organ. Thus, there is a need for systems and methods for isolating organs in medical image scans.

Brief Description of the Drawings

[0005]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

DETAILED DESCRIPTION OF THE INVENTION

[0006] A CAT scan is a procedure in which an object (e.g., a patient) is irradiated from several directions with penetrating (e.g., X-ray) radiation from a radiation source, and a scan image of the radiation that has penetrated through the detector is formed in each instance, forming a plurality of scan images that can each be represented as a two-dimensional array. Since radiation can be attenuated at different rates in different types of substances, each point in each image may correspond to the transmitted radiation intensity according to the attenuation rate of the composition of the substances on the path along which the radiation travels from the radiation source to the detector. From the combination of scan images, unprocessed scan data (e.g., a three-dimensional model of the "density" of the object) can be formed. As used herein, "density" within an object is any property that varies within the object and is measured in a medical image scan. For example, with respect to a CAT scan, "density" may refer to the local attenuation rate of the penetrating radiation, and with respect to an MRI scan, "density" may refer to the density of atoms having nuclear resonance at the frequency of the probe radio frequency (RF) signal in the presence of the applied magnetic field.

[0007] Some examples are described in the context of CAT scans or MRI scans of human patients in this disclosure, but the invention is not limited thereto. In some embodiments, other types of scans that provide three-dimensional density data, such as positron emission tomography scans, or scans of objects other than human patients can be processed similarly. For other types of scans, density can be defined accordingly. For example, in the case of a positron emission tomography scan, density may be the density of nuclei that are attenuated by beta-plus emission. As used herein, the term "object" includes anything that can be scanned, including, without limitation, human patients, animals, plants, inanimate objects, and combinations thereof.

[0008] When the object being imaged is a human patient (or other living entity), a contrast agent may be used (e.g., injected into the patient or ingested orally by the patient) to selectively change the density of some tissues. The contrast agent may, for example, contain a relatively opaque substance (i.e., relatively opaque to the radiation passing through). The density of the tissue containing the contrast agent may consequently increase, and may increase to an extent depending on the concentration of the contrast agent within the tissue.

[0009] Conventional segmentation methods are performed using machine learning (ML) techniques such as convolutional neural networks (CNNs) or some kind of label fusion. These segmentation methods have been successful in segmenting larger organs (e.g., the liver or kidneys), but are less successful in segmenting smaller organs, at least in part, due to the small size, shape, intensity, and variability in data statistics in smaller organs. Unlike larger organs, smaller organs such as the pancreas are more easily deformable. This is at least in part because the soft tissue forming the pancreas may be pushed to conform to the surrounding surface. This deformation causes the boundaries to (often) blur and blend into other adjacent organs. Additionally, smaller organs such as the pancreas may be displayed in a limited number of scans / slices. In the scans / slices where the pancreas is displayed, other organs may overlap the pancreas. This overlap of other organs results in only a part of the smaller organ being shown in a given scan / slice.

[0010] Embodiments relate to heuristic-based segmentation rather than ML-based. Embodiments can use multiple scan paths / thresholds when segmenting small organs from an image. Embodiments can use 3D correlation with the synthetic centroid to assist in the partitioning or regrouping of groups of pixels.

[0011] Embodiments can include frame normalization that excludes external parts (if any) from scans / slices. The terms frame, scan, slice, and image are all used synonymously herein. Embodiments can use a composite centroid mask that includes centroids around a composite centroid rather than a single centroid point. Embodiments can include an iterative approach to find thresholds for each scan / slice that takes into account the variation in MRI intensity. Embodiments can include another normalization operation that normalizes the pixels of the frame with respect to the position and image size of the frame. Embodiments can use the centroid mask to retain or remove one or more objects within the frame. Embodiments can include a boundary test that helps to distinguish a small organ from other parts within the frame. Embodiments can include identifying a "central slice", for example, by correlating detection slices to find a central reference frame. Embodiments can include removing high-average objects to remove veins, bones, or other high-intensity objects within the frame. Embodiments may include separating edges to separate the boundary to a neighboring organ or other object. Embodiments can include filling and joining organ mask portions, for example, to recombine damaged organ parts (from three-dimensional (3D) correlation).

[0012] FIG. 1 shows a block diagram of a system 100 for performing a scan, processing, and displaying the results according to one or more embodiments. The system 100 includes a scanner 110, a processing circuit 115, a display 120 for displaying an image or a sequence of images in the form of a movie (or "video"), and one or more input devices 125 such as a keyboard or a mouse that an operator (or, for example, a radiologist) can use to operate the system 100. The input device 125 can be used by the operator to affect the processing of the displayed image. The processing circuit 115, the display 120, and the input device 125 may be part of an integrated system, or may be, for example, separate and part of a distributed system having a processing circuit 115 communicatively coupled to the display 120 and the input device 125. In some embodiments, the server stores the images, the client retrieves the images, and the image processing is performed on the server, on the client, or on both.

[0013] The scanner 110 can generate a scan of the inner portion of an entity. The scans of the entity can be provided and analyzed together as a group of scans. For example, a first scan of an object (e.g., a patient) can be performed before a contrast agent is injected, and some subsequent scans of the object can be performed at various times (e.g., at regular intervals) after the injection of the contrast agent as the concentration of the contrast agent changes. The rate at which the concentration of the contrast agent first increases, the peak concentration reached, and the rate at which the concentration of the contrast agent subsequently decreases may all depend on the tissue into which the contrast agent is injected or the type of tissue being targeted.

[0014] In some embodiments, various methods can be used to generate an image from medical image scan data and assist in using the medical image scan as a diagnostic tool. A series of steps, namely the "operations" shown in FIG. 2 and described in more detail below, can be used, for example, to isolate a target organ (e.g., an organ suspected of having a tumor) and form a video or series of images in which the isolated target organ is more readily apparent than in the raw scan data.

[0015] The processing circuit 115 can include hardware, software, firmware, or a combination thereof configured to perform the operations of method 200. The hardware can include electrical or electronic components such as one or more transistors, resistors, capacitors, diodes, inductors, switches, power supplies, memory devices, oscillators, multiplexers, logic gates (e.g., AND, OR, XOR, NOT, buffer, etc.), amplifiers, analog-to-digital converters, digital-to-analog converters, processing units (e.g., central processing unit (CPU), field programmable gate array (FPGA), application specific integrated circuit (ASIC), graphics processing unit (GPU), etc.).

[0016] The display 120 can include a monitor, television, projector, screen (e.g., touch screen or non-touch screen), etc. configured to visually present image data to an operator. The display 120 can receive image data from the processing circuit 115 and provide a display of the image data.

[0017] FIG. 2 shows, by way of example, a block diagram of one embodiment of a method 200 for organ segmentation. The method 200 shown includes preprocessing a frame 220 at operation 222, resulting in a preprocessed frame 224, and at operation 226, executing a first segmentation pass over the processed frame 224, resulting in an initial segmentation frame 228, and at operation 230, executing a second segmentation pass over the initial segmentation frame 228, resulting in an organ mask 232.

[0018] The method 200 can be executed for each frame of a sequence of slices of a given entity. In some embodiments, operations 222 and 226 can be executed on each frame, and then operation 230 can be executed on the initial segmentation frame.

[0019] Aspects of the preprocessing operation 222 are described with respect to FIG. 3. Aspects of the first segmentation operation 226 are described with respect to FIG. 7. Aspects of the second segmentation operation 230 are described with respect to FIG. 8.

[0020] The result of the method 200 is the organ mask 232. The organ mask 232 may be a 3D mask of the organ through the frame 220. The 3D mask may be a composite of 2D masks provided from executing the method 200. An example of the mask 232 is provided in FIG. 11.

[0021] FIG. 3 shows, by way of example, a diagram of aspects of operation 222. Operation 222 may include receiving the frame 220. At operation 330, the frame 220 can be normalized. Operation 330 results in a preprocessed frame 224. Operation 330 is described in more detail with respect to FIG. 4.

[0022] The preprocessed frame 224 is provided to an operation 332 that identifies a region of interest 334. The region of interest 334 may include a target organ. The region of interest 334 for the pancreas may be near the patient's spine. The region of interest 334 can be guaranteed to include a target organ (e.g., the pancreas). The region of interest 334 may include a region close to the central region within the body.

[0023] The composite centroid mask 338 can be generated by an operation 336. The operation 336 is described in more detail with respect to FIG. 5. The threshold 342 used in the first and second splitting operations 226, 230 can be determined by an operation 340. The operation 340 is described in more detail with respect to FIG. 6.

[0024] FIG. 4 shows, as an example, a block diagram of an embodiment of a method for performing the operation 330 of FIG. 3. The operation 330 shown includes identifying a reference frame 440. The reference frame 440 is a frame among an array of frames. The reference frame 440 may be a random frame of the array of frames or a frame selected according to a criterion. The reference frame 440 may include a frame that can be used by an image analyzer to visually inspect and manually approximate a central reference point of a target organ.

[0025] At an operation 442, the dimensions of the body frame, including the width and the physical length of the body frame, can be determined. The width and length of the body frame are shown as (X, Y) 444. The dimensions of the body frame can be determined based on the resolution information in the header of the reference frame 440. The body frame is a spatial location within the reference frame 440 where the display of the internal part of the body is expanded.

[0026] At an operation 446, the central position 448 of the organ can be predicted. The central position 448 of the organ is shown as (X0, Y0). The central position 448 and the dimensions 444 of the body frame can be used at an operation 460 to enlarge / reduce a frame among the array of frames.

[0027] Different frames 450 can be received or retrieved. In operation 452, the edges of the body frame 454 within the second frame 450 can be identified. This information can be determined based on the header information of the frame 450. The edges of the body frame 454 can be defined by four points that define a rectangle within the frame 450. The body frame 454 can be guaranteed to contain the target organ.

[0028] The edges of the body frame 454 can be used to determine the body frame dimensions 458 in operation 456. The dimensions 458 of the body frame can be determined based on [XMIN, XMAX, YMIN, YMAX].

[0029] In operation 460, the body frame dimensions 458 of the body frame and the body frame dimensions 444 of the reference frame 440 can be used to scale the second frame 450. The second frame 450 can be scaled to match the scale of the reference frame 440. In operation 460, the center of the organ can be determined. The center of the organ can be estimated as the center of the tip of the organ. The center of the organ can be estimated in the same way as operation 446, but using additional information indicating the center position of the organ within the reference frame 440.

[0030] Based on the location of the center of the organ determined in operation 460 and the body frame dimensions 458 determined in operation 456, in operation 462, an angle 464 indicating the rotation angle of the organ can be determined. The organ angle 464 can be determined as a function of the ratio of the width and length of the body (in the axial view).

[0031] FIG. 5 shows, by way of example, a diagram of one embodiment of a method of performing operation 336. The result of operation 336 can include a composite centroid mask 562. The composite centroid mask 562 provides an outer shape that can be used to identify pixels within the frame 450 associated with the organ.

[0032] Operation 336 can include generating the synthetic organ tip 564 in operation 556 and generating the synthetic organ body 566 in operation 558. Since organs are often displayed as two chunks within a given frame, the synthetic organ tip 564 can be generated separately from the synthetic organ body 566. This also allows different parts of the organ to be bent or rotated separately. The tip 564 can bend or rotate differently from the body 566.

[0033] Operation 556 can be performed based on the organ angle 464, frame dimension 458, and center position 466 determined in operations 462, 456, and 460 respectively. This allows the organ mask to be oriented as a function of the MRI scan parameters and sized and shaped as a function of the MRI scan parameters.

[0034] Similarly, operation 558 can be performed based on the organ angle 464, frame dimension 458, and center position 466 determined in operations 462, 456, and 460 respectively. This allows the organ mask to be oriented as a function of the MRI scan parameters and sized and shaped as a function of the MRI scan parameters. The synthetic centroid mask 562 can be generated in operation 560 by combining the tip 564 and body 566 of the organ. The synthetic centroid mask 562 can then be used to assist in identifying the pixels associated with the organ during segmentation.

[0035] FIG. 6 shows, by way of example, a diagram of one embodiment of a method for performing operation 340. Operation 340 can identify a first pixel intensity threshold 678 and a second pixel intensity threshold 680. Operation 340 can include receiving or retrieving the preprocessed frame 224. Next, operation 340 can proceed to determine the thresholds 678, 680 for each of the frames 224.

[0036] In operation 660, the maximum intensity value 662 inside the body frame can be identified. The maximum intensity value 662 is the highest value. Multiple maximum intensity values may exist within a given frame or within multiple different frames. The maximum intensity value 662 can be determined across all normalized frames 224. In operation 664, the pixel intensity values of a given frame can be normalized (e.g., scaled up / down) based on the maximum intensity value. When all frames are normalized to the maximum intensity value 662, the detection threshold 678 and the second threshold 680 can be made executable for all of the normalized frames 224. Different maximum intensity values can be determined for each patient scan array.

[0037] In operation 666, the minimum area detection (MAD) inside the composite centroid mask 562 can be determined. In operation 666, an initial test detection threshold can be set. For better performance of the techniques herein, the test detection threshold is set too high and repeatedly reduced until the MAD is met.

[0038] In operation 668, the composite centroid mask 562 can be placed on frame 682 such that the mask covers the center position 466. The number of pixels within the mask 562 that are greater than the detection threshold can be totaled in operation 670 and compared to the MAD. This can be repeated for each scaled up / down and normalized frame 682. The number 672 of scaled up / down and normalized frames 682 that contain a number of pixel intensities greater than the MAD can be determined in operation 670.

[0039] The number 672 of scaled up / down and normalized frames 682 can be compared to the minimum number of frames in operation 674. The minimum number of frames can be determined experimentally. The minimum number of frames can help ensure that the organ segmentation functions accurately. Without a sufficient number of frames, the generated masks can become inaccurate, fragmented, or a combination thereof.

[0040] If the number 672 of the enlarged / reduced, normalized frames 682 includes the number of pixel intensity values within the centroid mask that is greater than the detection threshold and does not exceed the MAD, the detection threshold can be reduced in operation 682. If the number 672 of the enlarged / reduced, normalized frames 682 includes the number of pixel intensity values within the centroid mask that is greater than the detection threshold and exceeds the MAD, the detection threshold 678 can be set to the frame in operation 682. The second threshold 680 can be set in operation 676. The second threshold 680 can be set to be a scalar value of the detection threshold 678. The second threshold 680 is used in the second splitting path, while the detection threshold 678 is used in the first splitting path. Both the first splitting path and the second splitting path are described with respect to FIGS. 7 and 8.

[0041] The frame 450 can be received and combined in operation 770. Defining the boundary of the frame 450 can include keeping the pixel intensity values within the region of interest at their current values and setting all other pixel intensity values of the frame 450 to a specified value (e.g., zero (0)). Operation 770 can reduce the search space of the frame to reduce the overall processing. Operation 770 can hold the central part of the body within the frame around the spine. This is where the target organ usually resides.

[0042] In operation 772, a bounded frame 794 may be normalized, resulting in a normalized bounded frame 796. Operation 772 may include moving the region of interest within frame 794 by a column or pixel row to better center the region of interest. In operation 774, the pixels of the normalized bounded frame can be compared to a detection threshold 678. Any pixel having an intensity value below the detection threshold can be removed from the normalized bounded frame 796 (e.g., the intensity value is set to 0). The result of operation 774 is a partially segmented frame 798. It is partially segmented because the frame likely includes pixels corresponding to not only the target organ but also other structures in the field of view within frame 798.

[0043] In operation 776, small objects can be removed, resulting in a partially segmented frame 702 with small objects removed. Removing small objects in operation 776 may include counting the number of adjacent pixels having an intensity greater than a specified value (e.g., zero (0)). If the number of adjacent pixels is less than a specified pixel count threshold (e.g., 5, 10, 15, 20, a larger or smaller number, or some number in between), those pixels can be removed (e.g., their intensity values can be set to zero ((0))).

[0044] In operation 778, an object containing a sufficient number of adjacent pixels (not removed in operation 776) may have its average intensity checked. Checking the average intensity may include determining the average of the pixel intensities of the pixels constituting the object. If the average of the pixel intensities is higher than the average pixel intensity of the pixels of the target organ, then those pixels can be removed. In some embodiments, the average needs to exceed a threshold greater than the average pixel intensity of the pixels of the target organ. A further segmented frame 702 is the result of operation 778.

[0045] The centroid mask 562 can be used in operation 780 to identify an object within a further divided frame 702 that is likely part of the organ of interest. Operation 780 can include identifying and retaining an object having a centroid (central pixel) within the composite centroid mask 562. Any object having a centroid outside the centroid mask 562 can be deleted in operation 782. The retained object can be marked as "not to be deleted", such as by setting a bit associated with the object.

[0046] Operation 782 can include testing for objects outside all four sides of the centroid mask 562. A temporary mask can be generated for each side of the centroid mask 562. If an object is not shown as being retained and contains a pixel within one of the temporary masks, the object can be deleted. The result of the operation is the initially divided frame 228.

[0047] The organ mask can be defined to include the peripheral pixels of the retained objects of the initial divided frame 338. The organ mask can be expanded in operation 786. Operation 786 can include expanding the organ mask in all directions by a specified number of pixels (e.g., 1, 2, 3, or other number of pixels). Any holes within the organ mask can be filled in operation 788, resulting in the initial organ mask 790 of frame 228.

[0048] In operation 792, the central slice can be identified. The central slice is the slice within an array of frames that is at the center of the frames in which the organ of interest is visible. Operation 792 can include identifying each frame that includes the non-zero initial organ mask 790 and determining the frame that is at the center of the identified frames.

[0049] FIG. 8 shows, by way of example, a block diagram of a method for performing operation 230. Operation 230 can start in the same way as operation 226. Operations 770 and 772 are the same as the operations in FIG. 7 but are performed.

[0050] In operation 884, pixels of the bounded and normalized frame having an intensity less than the second threshold 680 are removed, resulting in a partially segmented frame 880. The 3D mask 802 (an aggregate of 2D organ masks 790 through frame 450) is determined in operation 886 and can be applied to the partially segmented frame 880 in operation 888. Applying the 3D mask 802 can include retaining only the pixels of the object within the partially segmented frame 880 that contain the centroid within the 3D mask 802 and removing the rest of the object.

[0051] (Similar to operation 778) In operation 890, objects having a high average intensity can be removed. In operation 230, operation 890 can help remove tumors, veins, or other parts that are inside the organ mask 802 but may not be part of the organ.

[0052] In operation 892, any object having an elongated connection can be separated. For each row of pixels, if the number of pixels having an intensity value greater than a specified threshold (e.g., zero) is less than a specified number (e.g., 1, 2, 3, or a larger value), those pixels can be set to zero to separate the object. Operation 892 helps remove objects that are in contact with the organ of interest.

[0053] In operation 894, an object having a centroid within the composite centroid mask 562 can be identified. Operation 894 can include retaining the identified object and removing any object that contains a centroid outside the centroid mask 562 (as a result of separating the edges in operation 892).

[0054] The organ mask 802 can be enlarged by operation 896. Operation 896 is similar to operation 786. At operation 898, the holes in the enlarged organ mask 804 can be filled to generate the organ mask 232 for the frame 450. The 3D mask can include a series of 2D organ masks for a given patient. The 3D mask can be displayed on a series of slices to assist a person in analyzing the target organ.

[0055] Figure 9 shows a series of images 900 of exemplary frames 450 before and after some of the operations of Figure 7. Frame 796 shows the result of performing operations 770 and 772 on frame 450. Frame 798 includes the initial organ mask generated after operation 774. Each pixel in the initial organ mask that is greater than the detection threshold 678 can be set to 1, and the rest can be set to zero.

[0056] The partially segmented frame 702 with small objects removed is the result of performing operation 776 on the partially segmented frame 798. The further segmented frame 704 is the result of performing operation 778 on the partially segmented frame 702 with small objects removed. The initial segmentation frame 228 is generated as a result of operation 782. The initial segmentation frame 228 can be used as a basis for generating an organ mask. The organ mask can include the pixels of the initial segmentation frame 228 that have an intensity greater than 0.

[0057] The enlarged organ mask 706 is the result of operation 786. The initial organ mask 790 is the result of operation 788. Image 990 shows the periphery of the organ mask 790 displayed on frame 450.

[0058] FIG. 10 shows a series of images 1000 of an exemplary frame 1010 before and after some of the operations of FIG. 8. The frame 1010 includes an identified region of interest shown by a dashed line. Frame 796 shows the result of performing operations 770 and 772 on frame 1010. Frame 880 includes an initial organ mask generated after operation 884. Each pixel in frame 880 that is greater than the second detection threshold 680 can be set to 1, and the rest can be set to zero.

[0059] Organ mask 882 can be generated as a result of operation 888. Operation 886 can be performed based on organ mask 790 for each scan within the array. Frame 806 can be generated as a result of operation 890. Operations 892 and 894 can be applied to frame 806 to generate organ mask 808. Organ mask 808 can be enlarged by operation 896 to generate an enlarged organ mask 804. The enlarged organ mask 804 can be filled with holes (to combine separated objects) by operation 898 to generate organ mask 232. Image 1012 includes frame 1010 with the perimeter of organ mask 232 shown thereon.

[0060] FIG. 11 shows, by way of example, various images 1100 of a 3D organ mask. The 3D organ mask includes a combination of 2D organ masks generated for each frame 450. The different shades of the 3D organ mask indicate different frames 450 from which the 2D organ masks were generated.

[0061] FIG. 12 shows a block diagram of an exemplary form of a machine of one embodiment of a computer system on which instructions for causing a machine to execute any one or more of the methods described herein may be executed. Scanner 110, processing circuitry 115, display 120, input device 125, or other components may include one or more components of machine 1200, or may be implemented using one or more components of machine 1200. In a networked deployment, the machine may operate in the capacity of a server or a client machine in a server-client network environment, or as a peer machine in a peer-to-peer (or distributed) network environment. The machine may be a personal computer (PC), tablet PC, set-top box (STB), personal digital assistant (PDA), cellular telephone, web appliance, network router, network switch, or network bridge, or any machine capable of executing instructions (sequentially or otherwise) that specify actions to be taken by that machine. Further, although only a single machine is shown, the term “machine” shall also be construed to include any collection of machines that individually or jointly execute a set (or multiple sets) of instructions for executing any one or more of the methods described herein.

[0062] Exemplary computer system 1200 includes a processor 1202 (e.g., a central processing unit (CPU), a graphics processing unit (GPU), or both) that communicates with each other via a bus 1208, a main memory 1204, and a static memory 1206. The computer system 1200 may further include a video display unit 1210 (e.g., a liquid crystal display (LCD) or a cathode ray tube (CRT)). The computer system 1200 also includes an alphanumeric input device 1212 (e.g., a keyboard), a user interface (UI) navigation device 1214 (e.g., a mouse), a mass storage device 1216, a signal generating device 1218 (e.g., a speaker), a network interface device 1220, and a wireless 1230 such as Bluetooth®, WWAN, WLAN, and NFC, enabling the application of security management for such protocols.

[0063] The illustrated computer system 1200 includes a sensor 1221 that converts energy from one form to another form of energy. The sensor 221 is a transducer that can convert thermal energy, sound energy, light energy, mechanical energy, or other energy into an electrical signal. The sensor 1221 can generally provide data indicating the characteristics of the environment in which the sensor 1221 is located.

[0064] The mass storage device 1216 includes a machine-readable medium 1222 in which one or more sets of instructions and data structures (e.g., software) 1224 that embody or are utilized by any one or more of the methods or functions described herein are stored. The instructions 1224 may also reside, completely or at least partially, within the main memory 1204 and / or within the processor 1202 during execution by the computer system 1200, and the main memory 1204 and the processor 1202 also constitute a machine-readable medium.

[0065] Machine-readable medium 1222 is illustrated in the exemplary embodiment as a single medium, but the term "machine-readable medium" can include a single medium or multiple media (e.g., a centralized or distributed database, and / or associated cache and server) that store one or more instructions or data structures. The term "machine-readable medium" is also to be construed as including any tangible medium that can store, encode, or carry instructions for execution by a machine, cause a machine to perform any one or more of the methods of the present invention, be utilized by such instructions, or store, encode, or carry data structures associated with such instructions. The term "machine-readable medium" is accordingly to be construed as including, without limitation, solid-state memory as well as optical and magnetic media. Specific examples of machine-readable media include, by way of example, non-volatile memory such as semiconductor memory devices, e.g., erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), and flash memory devices; magnetic disks such as internal hard disks and removable disks; magneto-optical disks; and CD-ROM and DVD-ROM disks.

[0066] Instructions 1224 can further be transmitted or received over communication network 1226 using a transmission medium. Instructions 1224 can be transmitted using either network interface device 1220 and any one of several well-known transfer protocols (e.g., HTTP). Examples of communication networks include local area networks ("LANs"), wide area networks ("WANs"), the Internet, cellular telephone networks, conventional telephone service (POTS) networks, and wireless data networks (e.g., WiFi networks and WiMax networks). The term "transmission medium" is to be construed as including any intangible medium that can store, encode, or carry instructions for execution by a machine, and includes digital or analog communication signals or other intangible media that facilitate communication of such software.

[0067] Additional Notes and Examples Example 1 can include a computer-implemented method for organ segmentation, the method including generating a composite centroid mask, identifying first and second intensity thresholds, and in a first segmentation pass, (i) setting pixels of an image having an intensity less than the first threshold to zero, and (ii) setting pixels of the image corresponding to an object having a centroid outside the composite centroid mask to zero, thereby resulting in an initial organ mask, and in a second segmentation pass, setting pixels (i) having an intensity less than the second threshold and less than the first threshold to zero, and (ii) setting pixels corresponding to an object having a centroid outside the initial organ mask to zero, thereby resulting in a second organ mask, and expanding and filling the second organ mask to generate an organ mask.

[0068] In Example 2, Example 1 can further include generating the organ mask for each frame of an array of scans and combining the generated organ masks to generate a three-dimensional organ mask.

[0069] In Example 3, at least one of Examples 1-2 can further include the first segmentation pass deleting objects in the frame having an average pixel intensity greater than the average pixel intensity of objects having a centroid within the composite centroid mask.

[0070] In Example 4, at least one of Examples 1-3 can further include the first segmentation pass deleting objects outside the centroid mask.

[0071] In Example 5, at least one of Examples 1-4 can further include the second segmentation pass separating connected objects by up to a specified number of pixels, thereby resulting in separated objects.

[0072] In Example 6, Example 5 can further include deleting an object among the separated objects whose center of gravity is outside the combined center-of-gravity mask in the second splitting path.

[0073] In Example 7, at least one of Examples 1 to 6 can further include the first and second splitting paths further including normalizing the size and position of the organ in the image.

[0074] Example 8 can include a non-transitory machine-readable medium including instructions that, when executed by a machine, cause the machine to perform operations for organ segmentation, the operations including generating a combined center-of-gravity mask, identifying first and second intensity thresholds, in a first splitting path, (i) setting pixels of an image having an intensity less than the first threshold to zero, and (ii) setting pixels of the image corresponding to an object having a center of gravity outside the combined center-of-gravity mask to zero, thereby resulting in an initial organ mask, and in a second splitting path, (i) setting pixels (i) having an intensity less than the first threshold and less than the second threshold to zero, and (ii) setting pixels corresponding to an object having a center of gravity outside the initial organ mask to zero, thereby resulting in a second organ mask, and expanding and filling the second organ mask to generate an organ mask.

[0075] In Example 9, Example 8 can further include the operations further including generating the organ mask for each frame of an array of scans and combining the generated organ masks to generate a three-dimensional organ mask.

[0076] In Example 10, at least one of Examples 8 to 9 can further include the first splitting path further including deleting an object in the frame having an average pixel intensity greater than the average pixel intensity of an object having a center of gravity within the combined center-of-gravity mask.

[0077] In Example 11, at least one of Examples 8 to 10 can further include deleting an object in which the first division path is outside the centroid mask.

[0078] In Example 12, at least one of Examples 8 to 11 can further include separating the combined object by at most a specified number of pixels for the second division path and the operation, resulting in a separated object.

[0079] In Example 13, Example 12 can further include deleting an object among the separated objects in which the centroid is outside the composite centroid mask for the second division path.

[0080] In Example 14, at least one of Examples 8 to 13 can further include normalizing the size and position of the organ in the image for the first and second division paths.

[0081] Example 15 can include a system including a processing circuit and a memory device storing instructions that, when executed by the processing circuit, cause the processing circuit to perform operations for generating an organ mask, the operations including generating a composite centroid mask, identifying first and second intensity thresholds, in a first division path, (i) setting pixels of an image having an intensity less than the first threshold to zero, (ii) setting pixels of the image corresponding to an object having a centroid outside the composite centroid mask to zero, resulting in an initial organ mask, in a second division path, (i) setting pixels (i) having an intensity less than the second threshold and less than the first threshold to zero, (ii) setting pixels corresponding to an object having a centroid outside the initial organ mask to zero, resulting in a second organ mask, and expanding and filling the second organ mask to generate an organ mask.

[0082] In Example 16, Example 15 can further include that the operation further includes generating the organ mask for each frame of the scan array and generating a three-dimensional organ mask by combining the generated organ masks.

[0083] In Example 17, at least one of Examples 15 to 16 can further include that the first segmentation path further includes deleting an object in the frame that includes an average pixel intensity greater than the average pixel intensity of an object whose center of gravity is within the combined center-of-gravity mask.

[0084] In Example 18, at least one of Examples 15 to 17 can further include that the first segmentation path further includes deleting an object that is outside the center-of-gravity mask.

[0085] In Example 19, at least one of Examples 15 to 18 can further include that the second segmentation path and the operation further include separating the combined objects by up to a specified number of pixels, resulting in separated objects.

[0086] In Example 20, Example 19 can further include that the second segmentation path further includes deleting an object among the separated objects whose center of gravity is outside the combined center-of-gravity mask.

[0087] In Example 21, at least one of Examples 15 to 20 can further include that the first and second segmentation paths further include normalizing the size and position of the organ in the image.

[0088] The appendix provides the disclosure of other organ segmentation techniques that can be used in whole or in part with the embodiments of this specification.

[0089] Embodiments have been described with reference to specific exemplary embodiments, but it will be apparent that various modifications and changes can be made to these embodiments without departing from the broader spirit and scope of the invention. Accordingly, the specification and drawings are to be regarded in an illustrative rather than a limiting sense. The accompanying drawings, which form a part of this specification, illustrate, by way of example and not limitation, specific embodiments in which the subject matter may be practiced. The illustrated embodiments are described in sufficient detail to enable those skilled in the art to practice the aspects of the disclosure disclosed herein. Since other embodiments may be utilized and structural and logical substitutions and changes may be made without departing from the scope of the disclosure, embodiments for carrying out the invention should not be construed in a limiting sense, and the scope of various embodiments is defined only by the appended claims, along with the full scope of equivalents to which such claims are entitled.

Claims

1. A computer-implemented method for organ segmentation, comprising: generating a composite centroid mask; identifying first and second intensity thresholds; in a first segmentation pass, (i) setting to zero pixels of an image having an intensity less than the first intensity threshold, and (ii) setting to zero pixels of the image corresponding to an object having a centroid outside the composite centroid mask, thereby resulting in an initial organ mask; in a second segmentation pass, (i) setting to zero pixels of the image having an intensity less than the second intensity threshold and less than the first intensity threshold, and (ii) setting to zero pixels of the image corresponding to an object having a centroid outside the initial organ mask, thereby resulting in a second organ mask; enlarging and filling the second organ mask to generate an organ mask; and wherein the first segmentation pass further comprises removing objects in a frame having an average pixel intensity greater than the average pixel intensity of objects having a centroid within the composite centroid mask.

2. The method of claim 1, further comprising generating the organ mask for each frame of an array of scans and combining the generated organ masks to generate a three-dimensional organ mask.

3. The method of claim 1, wherein the first segmentation pass further comprises removing objects outside the composite centroid mask.

4. The method of claim 1, wherein the second segmentation pass further comprises separating connected objects by up to a specified number of pixels, thereby resulting in separated objects.

5. The method of claim 4, wherein the second segmentation pass further comprises removing objects among the separated objects having a centroid outside the composite centroid mask.

6. The method of claim 1, wherein the first and second segmentation passes further comprise normalizing the size and position of the organ within the image.

7. A non-transitory machine-readable medium including instructions that, when executed by a machine, cause the machine to perform operations for organ segmentation, the operations comprising: generating a composite centroid mask; identifying first and second intensity thresholds; In a first segmentation path, (i) setting pixels of an image having an intensity less than the first intensity threshold to zero, and (ii) setting pixels of the image corresponding to an object having a center of gravity outside the combined center of gravity mask to zero, thereby resulting in an initial organ mask; In a second segmentation path, setting to zero pixels of the image (i) having an intensity less than the second intensity threshold and less than the first intensity threshold, and (ii) setting pixels of the image corresponding to an object having a center of gravity outside the initial organ mask to zero, thereby resulting in a second organ mask; Enlarging and filling the second organ mask to generate an organ mask; including; The first segmentation path further includes removing an object in a frame including an average pixel intensity greater than the average pixel intensity of an object having a center of gravity within the combined center of gravity mask, a non-transitory machine-readable medium.

8. The non-transitory machine-readable medium according to claim 7, wherein the operation further includes generating the organ mask for each frame of the scan array and combining the generated organ masks to generate a three-dimensional organ mask.

9. The non-transitory machine-readable medium according to claim 7, wherein the first segmentation path further includes removing an object outside the combined center of gravity mask.

10. The non-transitory machine-readable medium according to claim 7, wherein the second segmentation path and the operation further include separating combined objects by a maximum of a specified number of pixels, thereby resulting in separated objects.

11. The non-transitory machine-readable medium according to claim 10, wherein the second segmentation path further includes removing an object among the separated objects having a center of gravity outside the combined center of gravity mask.

12. The non-transitory machine-readable medium according to claim 7, wherein the first and second segmentation paths further include normalizing the size and position of the organ in the image.

13. A system comprising: a processing circuit; and a memory device storing instructions that cause the machine to perform operations for generating an organ mask when executed by the machine, wherein the operations include: generating a combined center of gravity mask; identifying first and second intensity thresholds; In a first segmentation path, (i) set pixels of an image having an intensity less than the first intensity threshold to zero, and (ii) set pixels of the image corresponding to an object having a center of gravity outside the composite center-of-gravity mask to zero, thereby resulting in an initial organ mask; In a second segmentation path, set to zero pixels (i) of the image having an intensity less than the second intensity threshold but not less than the first intensity threshold, and (ii) pixels of the image corresponding to an object having a center of gravity outside the initial organ mask, thereby resulting in a second organ mask; Enlarge and fill the second organ mask to generate an organ mask; including; The system further includes the first segmentation path removing an object in a frame including an average pixel intensity greater than the average pixel intensity of an object having a center of gravity within the composite center-of-gravity mask.

14. The system according to claim 13, wherein the operation further includes generating the organ mask for each frame of the scan array and combining the generated organ masks to generate a three-dimensional organ mask.

15. The system according to claim 13, wherein the first segmentation path further includes removing an object outside the composite center-of-gravity mask.

16. The system according to claim 13, wherein the second segmentation path and the operation further include separating combined objects by up to a specified number of pixels, thereby resulting in separated objects.

17. The system according to claim 16, wherein the second segmentation path further includes removing an object among the separated objects having a center of gravity outside the composite center-of-gravity mask.

18. The system according to claim 13, wherein the first and second segmentation paths further include normalizing the size and position of an organ in the image.

Images