Systems and methods for automatic estimation of maximum vertical pocket

Abstract

A maximum vertical pocket (MVP) estimation system is provided. The MVP estimation system includes a controller. The controller is configured to receive imaging data of a uterus of a patient. The imaging data includes a plurality of frames. The controller is further configured to detect, via a pocket detector, a plurality of pocket candidates from the plurality of frames. Each of the plurality of pocket candidates has a pocket depth. The controller is further configured to determine, via a pocket quality analyzer, a plurality of explainable pockets from the plurality of pocket candidates. The controller is further configured to determine, via a pocket depth estimator, an estimated MVP depth based on the plurality of explainable pockets.

Description

SYSTEMS AND METHODS FOR AUTOMATIC ESTIMATION OF MAXIMUM VERTICAL POCKETField of the Disclosure

[0001] The present disclosure is generally directed to fetal ultrasound diagnostics, and, more particularly, to systems and methods for automatic estimation of maximum vertical pocket depth of amniotic fluid within a uterus.Background

[0002] Maximum Vertical Pocket (MVP) evaluation is considered a reliable method for assessing amniotic fluid volume around the fetus as seen on an ultrasound scan. Amniotic fluid volume can be an important indicator of fetal health. MVP assessment is typically performed by an experienced clinician manually assessing and measuring a pocket for a maximal depth of amniotic fluid which is free of umbilical cord and fetal parts. This manual assessment typically requires high quality ultrasound scans to accurately determine an MVP depth.Summary of the Disclosure

[0003] The present disclosure is generally directed to systems and methods for automatic estimation of a maximum vertical pocket (MVP) of amniotic fluid within a uterus. In particular, the present disclosure provides systems and methods for automatically estimating depth of an MVP based on ultrasound imaging data, such as cine scan data comprising a plurality of frames, in a manner analogous to a manual evaluation performed by an experienced clinician. The imaging data is processed by a pocket detector, such as a “you only look once” (YOLO) neural network, to identify pocket candidates and evaluate their associated pocket depths. The pocket candidates are then processed by a pocket quality analyzer to generate a subset of explainable pockets by filtering out pockets which would have been ignored by a clinician during manual evaluation. A pocket depth estimator then processes the remaining subset of explainable pockets to determine an estimated MVP depth. Accordingly, the disclosed system enables the estimation of MVP depth based on low-to-moderate quality ultrasound scans captured by inexperienced or amateur sonographers (such as midwives with minimal training).

[0004] The pocket quality detector includes a series of filters to remove pockets from the subset of explainable pockets. The filters include a confidence filter, an anatomy filter, an echogenicityfilter, a continuity filter, and a pair of fan-beam intersection filters. The confidence filter processes the pocket candidates through a deep learning model trained on historic pocket data, and passes the pocket candidates most similar to the historic pockets to remove irregular or unusual pockets. The anatomy filter removes pocket candidates which correspond to non-uterine anatomies of the patient, such as their heart or bladder, as these anatomies may resemble pockets of amniotic fluid in the imaging data. The fan-beam intersection filter then removes any pocket candidates which appear to intersect or overlap the edges of a fan-beam shape of the frames. The echogenicity filter evaluates the echogenicity of the pocket candidates and filters out the pocket candidates having an echogenicity above an echogenicity threshold. The echogenicity filter may also evaluate a thickness of the pocket candidates and remove pocket candidates with a thickness below a thickness threshold. The continuity filter evaluates successive frames to remove pocket candidates present in irregular or erroneous frames. Once the subset of explainable pockets is generated, the pocket depth estimator analyzes the pocket depths of the explainable pockets to determine the estimated MVP depth. In some examples, the estimated MVP depth corresponds to the 99thpercentile of pocket depths of the explainable pockets.

[0005] Generally, in one aspect, an MVP estimation system is provided. The MVP estimation system includes a controller. The controller is configured to receive imaging data of a uterus of a patient. The imaging data includes a plurality of frames.

[0006] The controller is further configured to detect, via a pocket detector, a plurality of pocket candidates from the plurality of frames. Each of the plurality of pocket candidates has a pocket depth.

[0007] The controller is further configured to determine, via a pocket quality analyzer, a plurality of explainable pockets from the plurality of pocket candidates.

[0008] The controller is further configured to determine, via a pocket depth estimator, an estimated MVP depth based on the plurality of explainable pockets.

[0009] According to an example, the pocket quality analyzer includes a confidence filter. The confidence filter is configured to remove one or more of the plurality of pocket candidates having a confidence score less than a confidence threshold. The confidence score is assigned via a confidence scoring model. The confidence model may be a deep learning model. The confidence model may be trained with historical pocket data.

[0010] According to an example, the pocket quality analyzer includes an anatomy filter. The anatomy filter is configured to remove, from of the plurality of pocket candidates, one or more pocket candidates which correspond to non-uterine anatomies of the patient.

[0011] According to an example, the pocket quality analyzer includes a fan-beam intersection filter. The fan-beam intersection filter is configured to remove, from the plurality of pocket candidates, one or more of the plurality of pocket candidates which intersect with a fan-beam of a corresponding frame of the plurality of frames.

[0012] According to an example, the pocket quality analyzer includes a continuity filter. The continuity filter is configured to remove one or more of the plurality of pocket candidates based on successive frames of the plurality of frames.

[0013] According to an example, the pocket quality analyzer includes an echogenicity filter. The echogenicity filter is configured to remove one or more of the plurality of pocket candidates having an echogenicity greater than an echogenicity threshold. The echogenicity of one of the plurality of pocket candidates may be determined based on an intensity profile corresponding to a region of interest surrounding the one of the plurality of pocket candidates. The echogenicity may be determined by dividing a peak valley value of the intensity profile by a minimum edge intensity value of the intensity profile. The echogenicity filter may be further configured to remove one or more of the plurality of pocket candidates having a pocket thickness less than a thickness threshold.

[0014] According to an example, the pocket detector is a YOLO neural network.

[0015] According to an example, the estimated MVP corresponds to a 99thpercentile pocket depth of the plurality of explainable pockets.

[0016] According to an example, the imaging data is cine scan data.

[0017] Generally, in another example, a method for MVP estimation is provided. The method includes receiving, via a controller, imaging data of a uterus of a patient. The imaging data includes a plurality of frames.

[0018] The method further includes detecting, via a pocket detector of the controller, a plurality of pocket candidates from the plurality of frames, wherein each of the plurality of pocket candidates has a pocket depth.

[0019] The method further includes determining, via a pocket quality analyzer of the controller, a plurality of explainable pockets from the plurality of pocket candidates.

[0020] The method further includes determining, via a pocket depth estimator of the controller, an estimated MVP depth based on the plurality of explainable pockets.

[0021] In various implementations, a processor or controller may be associated with one or more storage media (generically referred to herein as “memory,” e.g., volatile and non-volatile computer memory such as RAM, PROM, EPROM, EEPROM, floppy disks, compact disks, optical disks, magnetic tape, SSD, etc.). In some implementations, the storage media may be encoded with one or more programs that, when executed on one or more processors and / or controllers, perform at least some of the functions discussed herein. Various storage media may be fixed within a processor or controller or may be transportable, such that the one or more programs stored thereon can be loaded into a processor or controller so as to implement various aspects as discussed herein. The terms “program” or “computer program” are used herein in a generic sense to refer to any type of computer code (e.g., software or microcode) that can be employed to program one or more processors or controllers.

[0022] It should be appreciated that all combinations of the foregoing concepts and additional concepts discussed in greater detail below (provided such concepts are not mutually inconsistent) are contemplated as being part of the inventive subject matter disclosed herein. In particular, all combinations of claimed subject matter appearing at the end of this disclosure are contemplated as being part of the inventive subject matter disclosed herein. It should also be appreciated that terminology explicitly employed herein that also may appear in any disclosure incorporated by reference should be accorded a meaning most consistent with the particular concepts disclosed herein.

[0023] These and other aspects of the various embodiments will be apparent from and elucidated with reference to the embodiment(s) described hereinafter.Brief Description of the Drawings

[0024] In the drawings, like reference characters generally refer to the same parts throughout the different views. Also, the drawings are not necessarily to scale, emphasis instead generally being placed upon illustrating the principles of the various embodiments.

[0025] FIG. 1 is an illustration of a maximum vertical pocket estimation system, in accordance with an example.

[0026] FIG. 2 is a block diagram of a controller of a maximum vertical pocket estimation system, in accordance with an example.

[0027] FIG. 3 is a block diagram of a pocket quality analyzer of the maximum vertical pocket estimation system of FIG. 2, in accordance with an example.

[0028] FIG. 4 is a block diagram of an echogenicity filter of the pocket quality analyzer of FIG. 3, in accordance with an example.

[0029] FIG. 5 is a frame of an ultrasound scan showing pocket candidates detected by the pocket detector of the maximum vertical pocket estimate system, in accordance with an example.

[0030] FIG. 6 is a frame of an ultrasound scan showing an expanded region-of-interest of a pocket candidate, in accordance with an example.

[0031] FIG. 7 is an intensity profile of a pocket candidate, in accordance with an example.

[0032] FIG. 8 is a schematic diagram of a controller of a maximum vertical pocket estimation system.

[0033] FIG. 9 is a schematic diagram of a processor of the controller of FIG. 8.

[0034] FIG. 10 is a flow chart of a method for maximum vertical pocket estimation, in accordance with an example.Detailed Description of Embodiments

[0035] The present disclosure is generally directed to systems and methods for automatic estimation of a maximum vertical pocket (MVP) of amniotic fluid within a uterus. In particular, in its various embodiments and implementations, the present disclosure provides systems and methods for automatically estimating depth of an MVP based on ultrasound imaging data, such as cine scan data including a plurality of frames, in a manner analogous to a manual evaluation performed by an experienced clinician. The imaging data is processed by a pocket detector, such as a “you only look once” (YOLO) neural network, to identify pocket candidates and evaluate their associated pocket depths. The pocket candidates are then processed by a pocket quality analyzer to generate a subset of explainable pockets by filtering out pockets which would have been ignored by a clinician during manual evaluation. A pocket depth estimator then processes the remaining subset of explainable pockets to determine an estimated MVP depth.

[0036] Turning now to the figures, FIG. 1 illustrates a non-limiting example of an MVP estimation system 10 according to various embodiments of the present disclosure. As shown inFIG. 1, the system 10 generally includes a controller 100, a display screen 200, and an ultrasound scanner 300. The ultrasound scanner 300 is used by a clinician or other healthcare professional to scan a uterus U of a patient P. In the non-limiting example of FIG. 1, the controller 100, the display screen 200, and the ultrasound scanner 300 are integrated into a single device, such as an ultrasound cart. However, in other examples, one or more of the components of the system 10 may be separated from the other components. For example, the display screen 200 could be embodied as a tablet computer screen or a television screen configured to receive visual depictions of the imaging data 102 captured by the ultrasound scan. The system 10 described in this disclosure enables inexperienced healthcare professionals or amateur sonographers (such as midwives with minimal training) to generate an estimated MVP depth 118 by capturing low to mid quality ultrasound scans. The estimated MVP depth 118 generated by the system 10 is explainable, such that clinician would be able to logically explain the estimated MVP depth 118 based on the frames 104 provided. Accordingly, the estimated MVP depth 118 should approximately match such an estimate observed and measured by an experienced clinician.

[0037] FIG. 2 shows a block diagram of the controller 100 of the MVP estimation system 10. As shown in FIGS. 7 and 8, the controller 100 generally includes a processor 125 for processing data, a memory 175 for storing data, and a transceiver 185 for transmitting and / or receiving data. Broadly, the controller 100 is configured to receive imaging data 102 from the ultrasound scanner 300 and generate an estimated MVP depth 118. Once generated, the estimated MVP depth 118 may then be displayed on the display screen 200. The imaging data 102 is generated from an ultrasound scanning protocol, such as a cine scan. In some examples, the scanning protocol may be a three-by-three or five-by-five blind sweep designed to cover most of the uterus U. Each scan includes a plurality of frames 104 showing the uterus U at different positions.

[0038] As shown in FIG. 2, the controller 100 includes three primary aspects: a pocket detector 106, a pocket quality analyzer 112, and a pocket depth estimator 116. The pocket detector 106 is configured to generate one or more pocket candidates 108 in each frame 104 of the imaging data 102. Examples of pocket candidates 108 in a frame 104 are shown in FIG. 5. The pocket candidates 108 are regions of the frame 104 representing amniotic fluid within the uterus U of the patient P. As shown in FIG. 5, the regions of amniotic fluid are relatively dark, meaning the scanned region is hypoechogenic. As further illustrated in FIG. 5, each of the pocket candidates 108 may bedefined by a pocket depth 110. In some examples, the pocket depth 110 may be provided in centimeters.

[0039] The pocket detector 106 may be any model capable of identifying pocket candidates 108. In some examples, the pocket detector 106 may be a deep learning, neural network model, such as a “you only look once” (YOLO) neural network. The pocket detector 106 may be trained by historical pocket data 128. The historical pocket data 128 may include previously generated ultrasound frames paired with identified pockets of amniotic fluid. The pocket detector 106 is typically trained offline prior to processing new imaging data 102.

[0040] Once the pocket detector 106 generates the pocket candidates 108, the pocket candidates 108 are provided to pocket quality analyzer 112. As will be explained in more detail with respect to FIGS. 3 and 4, the pocket quality analyzer 112 includes a plurality of filters configured to remove pockets candidates 108 which would not be selected by an experienced clinician performing a manual analysis. For example, the filters may remove pocket candidates 108 which overlap on non-uterine anatomy of the patient (such as their bladder), pocket candidates 108 which are too thin, pocket candidates 108 which intersect a fan-beam 136 of the frame 104, etc. Accordingly, the pocket quality analyzer 112 generates a plurality of explainable pockets 114 which would have been manually selected by an experience clinician. Each of these explainable pockets 114 has an associated pocket depth 110.

[0041] The pocket depths 110 of the explainable pockets 114 are then provided to the pocket depth estimator 116 to determine the estimated MVP depth 18. In some examples, the estimated MVP depth 118 may simply be the greatest pocket depth 110 of all of the explainable pockets 114 of all of the frames 104 of the imaging data 102. In other examples, the pocket depth estimate 116 may perform a statistical analysis to generate the estimated MVP depth 118. In some examples, the estimated MVP depth 118 is the 99thpercentile pocket depth 158 of all of the pocket depths 110 of all of the frames 104. By using the 99thpercentile value instead of the absolute maximum, the pocket depth analyzer 116 may filter out extreme outliers potentially representative of anomalies or system glitches. In other examples, other types of statistical analyses may be performed to determine the estimated MVP depth 118.

[0042] FIG. 3 illustrates various aspects of the pocket quality analyzer 112 shown in FIG. 2. The non-limiting example of FIG. 3 depicts the pocket quality analyzer 112 as including a confidence filter 120, an anatomy filter 130, a first fan-beam intersection filter 134a, anechogenicity filter 140, a second fan-beam intersection filter 134b, and a continuity filter 138. The pocket quality analyzer 112 is generally configured to receive the pocket candidates 108 produced by the pocket detector 106 and determine a subset of explainable pockets 114 from the pocket candidates 108. As described above, the various filters of the pocket quality analyzer 112 are configured to remove pocket candidates 108 which would typically be ignored by an experienced clinician manually evaluating MVP in a frame 104.

[0043] The confidence filter 120 processes the pocket candidates 108 using a confidence scoring model 126 to generate a confidence score 122 for each pocket candidate 108. The confidence scoring model 126 may be a deep learning model trained on historic pocket data 128 of a wide array of patients. Accordingly, the confidence score 122 represents the similarity of the pocket candidate 108 to previously evaluated pockets of amniotic fluid. The confidence filter 120 may then utilize a confidence threshold 124 to remove pocket candidates 108 with a confidence score 122 below the confidence threshold 124. In some examples, rather than evaluating the confidence score 122 against a constant threshold, the confidence filter 120 may remove all pocket candidates 108 except for the pocket candidates 108 with the highest (such as the 10 highest) confidence scores 122. The remaining pocket candidates 108 are then outputted by the confidence filter 120 as a confidence filter subset 108a of the pocket candidates 108. The confidence filter subset 108a includes the pocket candidates 108 most similar to the historic pockets with irregular or unusual pockets removed.

[0044] The confidence filter subset 108a is then provided to the anatomy filter 130 for further filtering. The anatomy filter 130 is configured to remove pocket candidates 108 corresponding to non-uterine anatomies of the patient P. For example, certain anatomical aspects of the patient P or their fetus (such as a bladder or heart) may appear in the frame 104 of the imaging data 102 as hypoechoic, similar to pockets of amniotic fluid. When visualized, these hypoechoic regions are shown as dark areas of the frame 104. Accordingly, the pocket detector 108 may mistake a section of the bladder or the heart of the patient P or fetus as a pocket of amniotic fluid. The anatomy filter 130 may receive non-uterine anatomy data 132 descriptive of these non-uterine regions to remove pocket candidates 108 corresponding to these regions. The remaining pocket candidates 108 are then outputted by the anatomy filter 130 as an anatomy filter subset 108b of the pocket candidates 108.

[0045] The anatomy filter subset 108b is then provided to the first fan-beam intersection filter 134a. An example fan-beam 134 of a frame 104 is shown in FIG. 5. The first fan-beam intersection filter 134b removes any pocket candidates 108 which appear to intersect or overlap one or more edges of a fan-beam 136. As shown in FIG. 5, an upper right corner of one pocket candidate 108i intersects the edge of the fan-beam 136. Such a fan-beam-overlapping pocket candidate 108i would typically be ignored by an experience clinician, as they would be unable to accurately measure the depth of the overlapping pocket. Thus, the first fan-beam intersection filter 134a removes pocket candidate 108i. The remaining pocket candidates 108 are then outputted by the first fan-beam intersection filter 134a as a first fan-beam filter subset 108c of the pocket candidates 108.

[0046] The first fan-beam filter subset 108c is then provided to the echogenicity filter 140. The echogenicity filter 140 will be described in greater detail with reference to FIG. 4. Broadly, the echogenicity filter 140 determines an echogenicity 142 of each of the remaining pocket candidates 108 and filters out the pocket candidates 108 having an echogenicity 142 above an echogenicity threshold 144. The echogenicity filter 140 may also evaluate a pocket thickness 154 of each of the remaining pocket candidates 108 and remove pocket candidates 108 with a pocket thickness 154 below a thickness threshold 156. The remaining pocket candidates 108 are then outputted by the echogenicity filter 140 as an echogenicity filter subset 108d of the pocket candidates 108.

[0047] The echogenicity filter subset 108d is then provided to the second fan-beam intersection filter 134b. As part of the echogenicity filter 140, an expanded region of interest 148 for each pocket candidate 108 is generated. These regions of interest 148 expand each of the pocket candidates 108 by an offset amount in both the vertical and horizontal directions. In some examples, this offset amount may be 1.5 centimeters. FIG. 6 illustrates an example expanded region of interest 148a corresponding to the pocket candidate 108a in FIG. 5. As will be described with reference to FIGS. 4 and 6, an intensity profile 146 of each expanded region of interest 148 is generated to evaluate the echogenicity 142 of the corresponding pocket candidate 108. The second fan-beam 134b removes pocket candidates 108 with expanded regions of interest 148 intersecting or overlapping with the fan beam 136. The remaining pocket candidates 108 are then outputted by the second fan-beam intersection filter 134b as a second fan-beam filter subset 108e of the pocket candidates 108.

[0048] The second fan-beam filter subset 108e is then provided to the continuity filter 138. The continuity filter 138 evaluates successive series of frames 104 to remove pocket candidates 108 shown in irregular or erroneous frames 104. The continuity filter 138 may perform this analysis by evaluating consecutive frames 104 for pocket candidates 108 of similar pocket depth 110. Outlier pocket candidates 108 are then removed by the continuity filter 138, and the remaining pocket candidates 108 are then outputted by the continuity filter 138 as the plurality of explainable pockets 114. The explainable pockets 114 are then provided to the pocket depth estimator 116 to generate an estimated MVP depth 118.

[0049] FIG. 4 illustrates various aspects of the echogenicity filter 140 shown in FIG. 3. The non-limiting example of FIG. 3 depicts the pocket quality analyzer 112 as including an image enhancer 160, a dominant edge extractor 162, a region-of-interest expander 164, an intensity profile extractor 166, an intensity well filter 168, and a pocket thickness filter 170. The echogenicity filter 140 is generally configured to receive the first fan-beam filter subset 108c of pocket candidates 108 from the first fan beam filter 134a and the determine echogenicity filter subset 108d of pocket candidates 108 to be provided to the second fan-beam intersection filter 134b. As with the other filters described above, the echogenicity filter 140 is configured to remove pocket candidates 108 which would typically be ignored by an experienced clinician manually evaluating MVP depth in a frame 104.

[0050] The image enhancer 160 processes the frames 104 containing the pocket candidates 108 by increasing the visual contrast between the portions of the frames 104 within the pocket candidates 108 (and therefore representative of amniotic fluid) and the portions of the frames 104 outside of the pocket candidates 108. The contrast is increased by darkening the amniotic fluid portion of the frames 104 and lightening the other, non-amniotic fluid portions. The image enhancer 160 then outputs a series of enhanced pocket candidates 108cl within each frame 104.

[0051] The enhanced pocket candidates 108cl are then provided to a dominant edge extractor 162. The dominant edge extractor 162 is configured to output the dominant edges 172 of each pocket candidate 108 following the image enhancement.

[0052] The dominant edges 108c3 of each pocket candidate 108 are then provided to a region- of-interest expander 164. The region-of-interest expander 164 generates an expanded region-of- interest 148 for each pocket candidate 108. In some examples, the expanded region-of-interest 148 for each pocket candidate 108 encompasses the area 1.5 centimeters around the pocket candidate108 in both vertical and horizontal directions. FIG. 6 illustrates a frame 104 with an example expanded region-of-interest 148a corresponding to pocket candidate 108a of FIG. 5.

[0053] The expanded regions-of-interest 148 of each pocket candidate 108 is then provided to an intensity profile extractor 166. The intensity profile extractor 166 generates an intensity profile 146 for each expanded region-of-interest 148. An example intensity profile 146 is shown in FIG. 7. The intensity profile 146 depicts the intensity (or brightness) over a range of pixels of a region- of-interest 148 of the frame 104. An intensity profile 146 representing a pocket of amniotic fluid clear of non-amniotic material or imaging processing artifacts will appear as a well or valley shape, as seen in FIG. 7. The intensity profile 146 may be defined by a peak valley value 150 and two edge intensity values 152a, 152b. The edge intensity values 152a, 152b correspond to dominant edges 172a, 172b of the pocket candidate 108 as determined by the dominant edge extractor 162.

[0054] The intensity profiles 146 are then provided to the intensity well filter 168. The intensity well filter 168 is configured to remove pocket candidates 108 having an echogenicity 142 greater than an echogenicity threshold 144. This filtering corresponds to the manual review performed by an experienced clinician who would ignore pocket candidates 108 having non-amniotic material or imaging artifacts. Generally, the echogenicity 142 is determined by comparing the intensity of the valley of the intensity profile 146 by the intensity of the edges. In one example, the echogenicity 142 is calculated by dividing a minimum edge intensity value 152 by a peak valley value 150. In the example of FIG. 7, the minimum edge intensity value 152 corresponds to the second edge 172b having a second edge intensity value 152b. Thus, the minimum edge intensity value 152 is approximately 120 joules. Further, the peak valley value 150 is approximately 30 joules. Thus, the echogenicity 142 of the intensity profile 146 is approximately 0.25. In other examples, rather than using the peak valley value 150, the echogenicity 142 may be determined by dividing the minimum edge intensity value 152 by a certain percentile value (such as 70thpercentile) of the valley. The pocket candidates 108 with an echogenicity 142 below the echogenicity threshold 144 are then outputted as an echogenic subset 108c2.

[0055] The echogenic subset 108c2 is then provided to the pocket thickness filter 170. The pocket thickness filter 170 is configured to remove pocket candidates 108 having a pocket thickness 154 less than a thickness threshold 156. This filtering corresponds to the manual review performed by an experienced clinician who would ignore very thin pocket candidates 108. The pocket thickness 154 is illustrated in the intensity profile 146 of FIG. 7 as the distance between thetwo dominant edges 152a, 152b. The remaining pocket candidates are then outputted as the echogenicity filter subset 108d to be provided to the second fan-beam intersection filter 134b as shown in FIG. 3.

[0056] FIG. 8 schematically illustrates the controller 100 previously depicted in FIGS. 1 and 2. The controller 100 includes the processor 125, the memory 175, and the transceiver 185. The processor 125 is shown in more detail in FIG. 9. The memory 175 is configured to store imaging data 102, the plurality of pocket candidates 108, the estimated MVP depth 118, the confidence threshold 124, the historical pocket data 128, the non-uterine anatomy data 134, the fan beam data 136, the echogenicity threshold 144, and the pocket thickness threshold 156. The imaging data 102 includes the plurality of frames 104. The pocket candidates 108 include, for each pocket candidate 108, the pocket depth 110, the confidence score 122, the echogenicity 142, the expanded region- of-interest 148, and an intensity profile 146 (including the peak valley value 150 and the minimum edge intensity value 152). The pocket candidates 108 may also include the subset of explainable pockets 114 as well as a 99thpercentile value 158 of the pocket depth of the explainable pockets 158.

[0057] FIG. 9 is a schematic diagram of the processor 125 of the controller 100 shown in FIG. 8. The processor 125 is configured to execute the pocket detector 106, the pocket quality analyzer 112, and the pocket depth estimator 116. The pocket quality analyzer 112 includes the confidence filter 120 (configured to execute the confidence scoring model 126), the anatomy filter 130, the first and second fan beam intersection filters 134a, 134b, the continuity filter 138, and the echogenicity filter 140. The echogenicity filter 140 is configured to execute the image enhancer 160, the dominant edge extractor 162, the region-of-interest expander 164, the intensity profile extractor 166, the intensity well filter 168, and the pocket thickness filter 170.

[0058] FIG. 10 is a flow chart of a method 900 for maximum vertical pocket estimation. Referring to FIGS. 1-10, the method 900 includes, in step 902, receiving, via a controller, imaging data 102 of a uterus U of a patient P. The imaging data 102 includes a plurality of frames 104.

[0059] The method 900 further includes, in step 904, detecting, via a pocket detector 106 of the controller 100, a plurality of pocket candidates 108 from the plurality of frames 104, wherein each of the plurality of pocket candidates 108 has a pocket depth 110.

[0060] The method 900 further includes, in step 906, determining, via a pocket quality analyzer 112 of the controller 100, a plurality of explainable pockets 114 from the plurality of pocket candidates 108.

[0061] The method 900 further includes, in step 908, determining, via a pocket depth estimator 116 of the controller 100, an estimated MVP depth 118 based on the plurality of explainable pockets 114.

[0062] All definitions, as defined and used herein, should be understood to control over dictionary definitions, definitions in documents incorporated by reference, and / or ordinary meanings of the defined terms.

[0063] The indefinite articles “a” and “an,” as used herein in the specification and in the claims, unless clearly indicated to the contrary, should be understood to mean “at least one.”

[0064] The phrase “and / or,” as used herein in the specification and in the claims, should be understood to mean “either or both” of the elements so conjoined, i.e., elements that are conjunctively present in some cases and disjunctively present in other cases. Multiple elements listed with “and / or” should be construed in the same fashion, i.e., “one or more” of the elements so conjoined. Other elements may optionally be present other than the elements specifically identified by the “and / or” clause, whether related or unrelated to those elements specifically identified.

[0065] As used herein in the specification and in the claims, “or” should be understood to have the same meaning as “and / or” as defined above. For example, when separating items in a list, “or” or “and / or” shall be interpreted as being inclusive, i.e., the inclusion of at least one, but also including more than one, of a number or list of elements, and, optionally, additional unlisted items. Only terms clearly indicated to the contrary, such as “only one of’ or “exactly one of,” or, when used in the claims, “consisting of,” will refer to the inclusion of exactly one element of a number or list of elements. In general, the term “or” as used herein shall only be interpreted as indicating exclusive alternatives (i.e. “one or the other but not both”) when preceded by terms of exclusivity, such as “either,” “one of,” “only one of,” or “exactly one of.”

[0066] As used herein in the specification and in the claims, the phrase “at least one,” in reference to a list of one or more elements, should be understood to mean at least one element selected from any one or more of the elements in the list of elements, but not necessarily including at least one of each and every element specifically listed within the list of elements and notexcluding any combinations of elements in the list of elements. This definition also allows that elements may optionally be present other than the elements specifically identified within the list of elements to which the phrase “at least one” refers, whether related or unrelated to those elements specifically identified.

[0067] It should also be understood that, unless clearly indicated to the contrary, in any methods claimed herein that include more than one step or act, the order of the steps or acts of the method is not necessarily limited to the order in which the steps or acts of the method are recited.

[0068] In the claims, as well as in the specification above, all transitional phrases such as “comprising,” “including,” “carrying,” “having,” “containing,” “involving,” “holding,” “composed of,” and the like are to be understood to be open-ended, i.e., to mean including but not limited to. Only the transitional phrases “consisting of’ and “consisting essentially of’ shall be closed or semi-closed transitional phrases, respectively.

[0069] The above-described examples of the described subject matter can be implemented in any of numerous ways. For example, some aspects may be implemented using hardware, software, or a combination thereof. When any aspect is implemented at least in part in software, the software code can be executed on any suitable processor or collection of processors, whether provided in a single device or computer or distributed among multiple devices / computers.

[0070] The present disclosure may be implemented as a system, a method, and / or a computer program product at any possible technical detail level of integration. The computer program product may include a computer readable storage medium (or media) having computer readable program instructions thereon for causing a processor to carry out aspects of the present disclosure.

[0071] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium may be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer readable storage medium includes the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in agroove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0072] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network may comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0073] Computer readable program instructions for carrying out operations of the present disclosure may be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, configuration data for integrated circuitry, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++, or the like, and procedural programming languages, such as the “C” programming language or similar programming languages. The computer readable program instructions may execute entirely on the user’s computer, partly on the user's computer, as a standalone software package, partly on the user’s computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider). In some examples, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate arrays (FPGA), or programmable logic arrays (PLA) may execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0074] Aspects of the present disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to examples of the disclosure. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer readable program instructions.

[0075] The computer readable program instructions may be provided to a processor of a, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions may also be stored in a computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram or blocks.

[0076] The computer readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process, such that the instructions which execute on the computer, other programmable apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0077] The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various examples of the present disclosure. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the blocks may occur out of the order noted in the Figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block ofthe block diagrams and / or flowchart illustration, and combinations of blocks in the block diagrams and / or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts or carry out combinations of special purpose hardware and computer instructions.

[0078] Other implementations are within the scope of the following claims and other claims to which the applicant may be entitled.

[0079] While various examples have been described and illustrated herein, those of ordinary skill in the art will readily envision a variety of other means and / or structures for performing the function and / or obtaining the results and / or one or more of the advantages described herein, and each of such variations and / or modifications is deemed to be within the scope of the examples described herein. More generally, those skilled in the art will readily appreciate that all parameters, dimensions, materials, and configurations described herein are meant to be exemplary and that the actual parameters, dimensions, materials, and / or configurations will depend upon the specific application or applications for which the teachings is / are used. Those skilled in the art will recognize, or be able to ascertain using no more than routine experimentation, many equivalents to the specific examples described herein. It is, therefore, to be understood that the foregoing examples are presented by way of example only and that, within the scope of the appended claims and equivalents thereto, examples may be practiced otherwise than as specifically described and claimed. Examples of the present disclosure are directed to each individual feature, system, article, material, kit, and / or method described herein. In addition, any combination of two or more such features, systems, articles, materials, kits, and / or methods, if such features, systems, articles, materials, kits, and / or methods are not mutually inconsistent, is included within the scope of the present disclosure.

Claims

ClaimsWhat is claimed is:

1. A maximum vertical pocket (MVP) estimation system (10), comprising a controller (100) configured to: receive imaging data (102) of a uterus (U) of a patient (P), wherein the imaging data (102) comprises a plurality of frames (104); detect, via a pocket detector (106), a plurality of pocket candidates (108) from the plurality of frames (104), wherein each of the plurality of pocket candidates (108) has a pocket depth (110); determine, via a pocket quality analyzer (112), a plurality of explainable pockets (114) from the plurality of pocket candidates (108); and determine, via a pocket depth estimator (116), an estimated MVP depth (118) based on the plurality of explainable pockets (114).

2. The MVP estimation system (10) of claim 1, wherein the pocket quality analyzer (112) comprises a confidence filter (120) configured to remove one or more of the plurality of pocket candidates (108) having a confidence score (122) less than a confidence threshold (124), and wherein the confidence score (122) is assigned via a confidence scoring model (126).

3. The MVP estimation system (10) of claim 2, wherein the confidence scoring model (126) is a deep learning model.

4. The MVP estimation system (10) of claim 2, wherein the confidence scoring model (126) is trained with historical pocket data (128).

5. The MVP estimation system (10) of claim 1, wherein the pocket quality analyzer (112) comprises an anatomy filter (130) configured to remove, from of the plurality of pocket candidates (108), one or more pocket candidates (108) which correspond to non-uterine anatomies (132) of the patient (P).

6. The MVP estimation system (10) of claim 1, wherein the pocket quality analyzer (112) comprises fan-beam intersection filter (134) configured to remove, from the plurality of pocketcandidates (108), one or more of the plurality of pocket candidates (108) which intersect with a fan-beam (136) of a corresponding frame (104) of the plurality of frames (104).

7. The MVP estimation system (10) of claim 1, wherein the pocket quality analyzer (112) comprises a continuity filter (138) configured to remove one or more of the plurality of pocket candidates (108) based on successive frames (104) of the plurality of frames (104).

8. The MVP estimation system (10) of claim 1, wherein the pocket quality analyzer (112) comprises an echogenicity filter (140) configured to remove one or more of the plurality of pocket candidates (108) having an echogenicity (142) greater than an echogenicity threshold (144).

9. The MVP estimation system (10) of claim 8, wherein the echogenicity (142) of one of the plurality of pocket candidates (108) is determined based on an intensity profile (146) corresponding to a region of interest (148) surrounding the one of the plurality of pocket candidates (108).

10. The MVP estimation system (10) of claim 9, wherein the echogenicity (142) is determined by dividing a peak valley value (150) of the intensity profile (146) by a minimum edge intensity value (152) of the intensity profile (146).

11. The MVP estimation system (10) of claim 8, wherein the echogenicity filter (140) is further configured to remove one or more of the plurality of pocket candidates (108) having a pocket thickness (154) less than a thickness threshold (156).

12. The MVP estimation system (10) of claim 1, wherein the pocket detector (106) is a “you only look once” (YOLO) neural network.

13. The MVP estimation system (10) of claim 1, wherein the estimated MVP depth (118) corresponds to a 99thpercentile pocket depth (158) of the plurality of explainable pockets (114).

14. The MVP estimation system (10) of claim 1, wherein the imaging data (102) is cine scan data.

15. A method (900) for maximum vertical pocket (MVP) estimation, comprising: receiving (902), via a controller, imaging data of a uterus of a patient, wherein the imaging data comprises a plurality of frames; detecting (904), via a pocket detector of the controller, a plurality of pocket candidates from the plurality of frames, wherein each of the plurality of pocket candidates has a pocket depth; determining (906), via a pocket quality analyzer of the controller, a plurality of explainable pockets from the plurality of pocket candidates; and determining (908), via a pocket depth estimator of the controller, an estimated MVP depth based on the plurality of explainable pockets.

Images