WO2022140960A1 - 一种卵泡跟踪方法和系统 - Google Patents
一种卵泡跟踪方法和系统 Download PDFInfo
- Publication number
- WO2022140960A1 WO2022140960A1 PCT/CN2020/140317 CN2020140317W WO2022140960A1 WO 2022140960 A1 WO2022140960 A1 WO 2022140960A1 CN 2020140317 W CN2020140317 W CN 2020140317W WO 2022140960 A1 WO2022140960 A1 WO 2022140960A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- follicle
- region
- area
- ultrasound image
- ovarian
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 189
- 238000002604 ultrasonography Methods 0.000 claims abstract description 398
- 230000002611 ovarian Effects 0.000 claims abstract description 283
- 238000012360 testing method Methods 0.000 claims abstract description 15
- 238000007689 inspection Methods 0.000 claims description 55
- 230000011218 segmentation Effects 0.000 claims description 54
- 230000003325 follicular Effects 0.000 claims description 51
- 239000000523 sample Substances 0.000 claims description 25
- 238000005070 sampling Methods 0.000 claims description 5
- 230000010354 integration Effects 0.000 claims 1
- 238000001514 detection method Methods 0.000 abstract description 11
- 238000004422 calculation algorithm Methods 0.000 description 40
- 230000008569 process Effects 0.000 description 22
- 238000013135 deep learning Methods 0.000 description 21
- 235000013601 eggs Nutrition 0.000 description 20
- 238000010801 machine learning Methods 0.000 description 19
- 238000003709 image segmentation Methods 0.000 description 14
- 238000012544 monitoring process Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 11
- 238000011161 development Methods 0.000 description 9
- 230000018109 developmental process Effects 0.000 description 9
- 230000003993 interaction Effects 0.000 description 9
- 238000012285 ultrasound imaging Methods 0.000 description 9
- 239000000243 solution Substances 0.000 description 8
- 238000013519 translation Methods 0.000 description 8
- 230000006870 function Effects 0.000 description 6
- 238000002372 labelling Methods 0.000 description 6
- 230000009466 transformation Effects 0.000 description 6
- 238000005295 random walk Methods 0.000 description 5
- 239000000284 extract Substances 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000001131 transforming effect Effects 0.000 description 4
- 241000270295 Serpentes Species 0.000 description 3
- 239000003814 drug Substances 0.000 description 3
- 229940079593 drug Drugs 0.000 description 3
- 238000002347 injection Methods 0.000 description 3
- 239000007924 injection Substances 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000004590 computer program Methods 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006698 induction Effects 0.000 description 2
- 208000000509 infertility Diseases 0.000 description 2
- 230000036512 infertility Effects 0.000 description 2
- 231100000535 infertility Toxicity 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 210000001672 ovary Anatomy 0.000 description 2
- 230000016087 ovulation Effects 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 238000002187 spin decoupling employing ultra-broadband-inversion sequences generated via simulated annealing Methods 0.000 description 2
- 238000011426 transformation method Methods 0.000 description 2
- 206010042573 Superovulation Diseases 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 238000002592 echocardiography Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 210000001161 mammalian embryo Anatomy 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000027758 ovulation cycle Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000000638 stimulation Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
- 238000002054 transplantation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10132—Ultrasound image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Definitions
- the present application relates to the technical field of follicle tracking, and more particularly to a follicle tracking method and system.
- IVF infertility
- the key step in IVF is egg retrieval.
- the patient's natural cycle usually only has one dominant follicle to obtain one embryo, so in order to improve the success rate of transplantation, it is necessary to use controlled superovulation to enhance and improve ovarian function, so as to achieve the goal of obtaining multiple healthy eggs without being restricted by the natural cycle .
- the timing of egg retrieval is critical.
- a follicle tracking method comprising:
- the growth trend graph is displayed.
- the growth trend graph is a growth parameter graph, wherein the growth parameter graph takes the inspection time as the first coordinate and the growth parameter as the second coordinate; or,
- the growth trend graph is a list of the growth parameters corresponding to the different inspection times.
- determining the follicle area corresponding to the target follicle on the ultrasound images of the at least three different inspection times respectively includes:
- the follicle area corresponding to the target follicle is determined in the second ultrasonic image in the at least three ultrasonic images of different inspection times, and the second follicle area is obtained;
- the follicle region corresponding to the target follicle is determined in the third ultrasound image of the at least three ultrasound images at different inspection times, and the third follicle region is obtained.
- determining the follicle area corresponding to the target follicle, and obtaining the first follicle area including:
- the ultrasound image is a three-dimensional ultrasound image
- the follicle area corresponding to the target follicle is measured in the first ultrasound image of the at least three ultrasound images at different inspection times based on the image features of the follicle. Segmentation is performed to obtain the first follicular area, including:
- the corresponding regions of the target follicle on the plurality of two-dimensional slice images are synthesized to obtain the first follicle region.
- the multiple two-dimensional slice images of the first ultrasound image are all two-dimensional slice images in the first ultrasound image, or,
- the multiple two-dimensional slices of the first ultrasound image are sampled images obtained by sampling the first ultrasound image according to a first preset rule, and the target follicles on the multiple two-dimensional slices are integrated.
- the corresponding area of includes: performing three-dimensional interpolation on the segmentation result of the sampled image to obtain the first follicle area.
- the ultrasound image is a three-dimensional ultrasound image
- the follicle area corresponding to the target follicle is measured in the first ultrasound image of the at least three ultrasound images at different inspection times based on the image features of the follicle. Segmentation is performed to obtain the first follicular area, including:
- three-dimensional segmentation is performed on the follicle area corresponding to the target follicle in the first ultrasonic image in the at least three three-dimensional ultrasonic images of different inspection times, so as to obtain the first follicle area.
- determining the follicle area corresponding to the target follicle, and obtaining the first follicle area including:
- a follicle area corresponding to the target follicle is determined based on the first ovarian area, and a first follicular area is obtained.
- determining the follicle area corresponding to the target follicle, and obtaining the first follicle area including:
- the growth parameters of each first candidate follicle region are acquired, and the first candidate follicle region whose growth parameter meets the first preset condition is determined as the first follicle region.
- determining the follicle area corresponding to the target follicle, and obtaining the first follicle area including:
- the growth parameters of each first candidate follicle region are acquired, and the first candidate follicle region with the maximum growth parameter is determined as the first follicle region.
- the follicle region corresponding to the target follicle is determined in the second ultrasound image of the at least three ultrasound images of different inspection times, and the second follicle region is obtained, include:
- the follicle region corresponding to the target follicle is determined in the second ultrasound image of the at least three ultrasound images of different inspection times, and the second follicle region is obtained, include:
- the first ovarian region is registered with the second ovarian region.
- registering the first ovarian region with the second ovarian region comprises:
- the orientation of the second ovarian region in the second ultrasound image is adjusted, so that the adjusted second ovarian region is in the second ultrasound image.
- the orientation in the second ultrasound image is the same as the orientation of the first ovarian region in the first ultrasound image; or,
- the size of the second ovarian region is adjusted relative to the first ovarian region so that the adjusted second ovarian region is the same size as the first ovarian region.
- registering the first ovarian region with the second ovarian region includes:
- a second learning model is used to determine the corresponding relationship between the second ovarian region and the first ovarian region, and the second ovarian region is adjusted according to the corresponding relationship; or,
- An instruction from the operator to adjust the second ovarian area according to the first ovarian area is received, and the second ovarian area is adjusted according to the instruction.
- the second ultrasound image is matched with the template of the first ovarian region, and a second ovarian region of the ovarian tissue in the second ultrasound image is determined according to the matching result.
- the growth parameters include at least one of the following: volume, path length and growth rate.
- a follicle tracking system comprising:
- a transmitting circuit used to excite the ultrasonic probe to transmit ultrasonic waves to the ovarian tissue of the measured object
- a receiving circuit configured to excite the ultrasonic probe to receive the echo of the ultrasonic wave to obtain the echo signal of the ultrasonic wave
- a follicle tracking method comprising:
- the ovarian tissue includes the target follicle
- the growth trend graph is displayed.
- a follicle tracking system comprising:
- a transmitting circuit used to excite the ultrasonic probe to transmit ultrasonic waves to the ovarian tissue of the measured object
- a receiving circuit configured to excite the ultrasonic probe to receive the echo of the ultrasonic wave to obtain the echo signal of the ultrasonic wave
- a processor for executing the follicle tracking method as described in the above embodiments.
- a single target follicle can be tracked and monitored, and a growth trend graph of the single target follicle can be finally obtained, and the operator can accurately evaluate the optimal egg retrieval time according to the growth trend graph of the target follicle. , effectively improve work efficiency and accuracy.
- a follicle tracking method comprising:
- a follicle area corresponding to the target follicle is determined in the second ultrasound image, and a second follicle area is obtained.
- the follicle area corresponding to the target follicle is determined in the first ultrasound image to obtain the first follicle area, including:
- the ultrasound image is a three-dimensional ultrasound image
- the follicle region corresponding to the target follicle is segmented in the first ultrasound image based on the image features of the follicle to obtain the first follicle region, including: :
- the corresponding regions of the target follicle on the plurality of two-dimensional slice images are synthesized to obtain the first follicle region.
- the multiple two-dimensional slice images of the first ultrasound image are all two-dimensional slice images in the first ultrasound image, or,
- the multiple two-dimensional slices of the first ultrasound image are sampled images obtained by sampling the first ultrasound image according to a first preset rule, and the target follicles on the multiple two-dimensional slices are integrated.
- the corresponding area of includes: performing three-dimensional interpolation on the segmentation result of the sampled image to obtain the first follicle area.
- the ultrasound image is a three-dimensional ultrasound image
- the follicle region corresponding to the target follicle is segmented in the first ultrasound image based on the image features of the follicle to obtain the first follicle region, including: :
- three-dimensional segmentation is performed on the follicle area corresponding to the target follicle in the first ultrasound image to obtain the first follicle area.
- determining the follicle area corresponding to the target follicle to obtain the first follicle area including:
- a follicle area corresponding to the target follicle is determined based on the first ovarian area, and a first follicular area is obtained.
- determining the follicle area corresponding to the target follicle to obtain the first follicle area including:
- the growth parameters of each first candidate follicle area are acquired, and the first candidate follicle area whose growth parameter meets the first preset condition is determined as the first follicle area.
- determining the follicle area corresponding to the target follicle to obtain the first follicle area including:
- the growth parameters of each first candidate follicle area are acquired, and the first candidate follicle area with the maximum growth parameter is determined as the first follicle area.
- the follicle area corresponding to the target follicle is determined in the second ultrasound image, and the second follicle area is obtained, including:
- the follicle area corresponding to the target follicle is determined in the second ultrasound image, and the second follicle area is obtained, including:
- the operation of the operator to identify the follicular region corresponding to the first follicular region in the second ultrasound image is detected, and the identified follicular region is determined as the second follicular region.
- the first ovarian region is registered with the second ovarian region.
- registering the first ovarian region with the second ovarian region comprises:
- the orientation of the second ovarian region in the second ultrasound image is adjusted, so that the adjusted second ovarian region is in the second ultrasound image.
- the orientation in the second ultrasound image is the same as the orientation of the first ovarian region in the first ultrasound image; or,
- the size of the second ovarian region is adjusted relative to the first ovarian region so that the adjusted second ovarian region is the same size as the first ovarian region.
- registering the first ovarian region with the second ovarian region comprises:
- a second learning model is used to determine the corresponding relationship between the second ovarian region and the first ovarian region, and the second ovarian region is adjusted according to the corresponding relationship; or,
- An instruction from the operator to adjust the second ovarian area to be registered according to the first ovarian area is received, and the second ovarian area to be registered is adjusted according to the instruction.
- the second ultrasound image is matched with the template of the first ovarian region, and a second ovarian region of the ovarian tissue in the second ultrasound image is determined according to the matching result.
- the growth parameters include at least one of the following: volume, path length and growth rate.
- a follicle tracking system comprising:
- a transmitting circuit used to excite the ultrasonic probe to transmit ultrasonic waves to the ovarian tissue of the measured object
- a receiving circuit configured to excite the ultrasonic probe to receive the echo of the ultrasonic wave to obtain the echo signal of the ultrasonic wave
- the same target follicle can be tracked and monitored on multiple ultrasound images, and the growth and development of the same target follicle can be clinically achieved, and the operator can accurately assess the growth and development of the same target follicle.
- Optimal egg retrieval time effectively improving work efficiency and accuracy.
- FIG. 1 shows a schematic block diagram of an ultrasound imaging system according to an embodiment of the present application
- FIG. 2 shows a schematic flowchart of a method for tracking follicles according to an embodiment of the present invention
- FIG. 3 shows a schematic diagram of a follicle tracking method according to another embodiment of the present invention.
- FIG. 4 shows a schematic diagram of a follicle tracking method according to another embodiment of the present invention.
- FIG. 5 shows a schematic diagram of a follicle tracking method according to another embodiment of the present invention.
- FIG. 6 shows a schematic diagram of a follicle tracking method according to another embodiment of the present invention.
- FIG. 7 shows a schematic diagram of a follicle tracking method according to another embodiment of the present invention.
- FIG. 8 shows a schematic diagram of a follicle tracking method according to another embodiment of the present invention.
- FIG. 9 shows a schematic diagram of a follicle tracking method according to another embodiment of the present invention.
- FIG. 10 shows a schematic diagram of a follicle tracking method according to another embodiment of the present invention.
- FIG. 11 shows a schematic diagram of a follicle tracking method according to another embodiment of the present invention.
- FIG. 12 shows a schematic flowchart of a method for tracking follicles according to yet another embodiment of the present invention.
- FIG. 1 shows a schematic structural block diagram of an ultrasound imaging system 100 according to an embodiment of the present application.
- the ultrasound imaging system 100 includes an ultrasound probe 110 , a transmit/receive circuit 112 , a processor 114 , and a display 116 . Further, the ultrasound imaging system 100 may further include a beam forming circuit, a transmit/receive selection switch, and the like.
- Ultrasound probe 110 typically includes an array of multiple elements. Every time ultrasonic waves are emitted, all or part of the array elements of the ultrasonic probe 110 participate in the emission of ultrasonic waves. At this time, each array element or each part of the array elements participating in the ultrasonic emission is stimulated by the transmitting pulse and emits ultrasonic waves respectively. A synthetic ultrasound beam in the area where the subject's ovarian tissue is located.
- the transmit/receive circuit 112 may be connected to the ultrasound probe 110 through a transmit/receive selection switch.
- the transmit/receive selection switch may also be called a transmit/receive controller, which may include a transmit controller and a receive controller, and the transmit controller is used to excite the ultrasound probe 110 to transmit ultrasound to the area where the ovarian tissue of the measured object is located via the transmit circuit;
- the receiving controller is configured to receive the ultrasonic echo returned from the region where the ovarian tissue of the test object is located through the ultrasonic probe 110 via the receiving circuit, so as to obtain ultrasonic echo data.
- the transmitting/receiving circuit 112 sends the electrical signal of the ultrasonic echo into the beam forming circuit, and the beam forming circuit performs processing such as focusing delay, weighting and channel summation on the electrical signal, and then sends the processed ultrasonic echo data to the beam forming circuit. into the processor 114.
- the processor 114 may be implemented by software, hardware, firmware or any combination thereof, and may use circuits, single or multiple application specific integrated circuits (ASICs), single or multiple general-purpose integrated circuits, single or multiple microprocessors, single or multiple programmable logic devices, or any combination of the foregoing circuits or devices, or other suitable circuits or devices, thereby enabling the processor 114 to perform the methods in the various embodiments in this specification corresponding steps. Also, the processor 114 may control other components in the ultrasound imaging system 100 to perform desired functions.
- ASICs application specific integrated circuits
- microprocessors single or multiple programmable logic devices
- the processor 114 processes the received ultrasound echo data to obtain ultrasound data of the ovarian tissue of the measured object, where the ultrasound data may be two-dimensional ultrasound data or three-dimensional ultrasound data.
- the ultrasound probe 110 transmits/receives ultrasound in a series of scanning planes, and is integrated by the processor 114 according to its three-dimensional spatial relationship to realize the three-dimensional scanning and three-dimensional scanning of the ovarian tissue of the measured object. Image reconstruction.
- the processor 114 After part or all of the image post-processing steps such as denoising, smoothing, and enhancement are performed by the processor 114, the three-dimensional ultrasound data of the ovarian tissue of the measured object is acquired.
- the processor 114 may acquire ultrasound images of at least three different examination times of the ovarian tissue of the subject, wherein the ovarian tissue includes the target follicle.
- the processor 114 may also determine corresponding regions of the same target follicle on at least three ultrasound images at different examination times, respectively.
- the processor 114 may also determine growth parameters based on the follicular area corresponding to the target follicle.
- the processor 114 obtains a growth trend graph of the target follicle according to the growth parameters of the target follicle.
- the trend graph obtained by processor 114 may be stored in memory or displayed on display 116 .
- the display 116 is connected to the processor 114, and the display 116 may be a touch display screen, a liquid crystal display screen, etc.; or the display 116 may be an independent display device such as a liquid crystal display, a television set, etc. independent of the ultrasound imaging system 100; or the display 116 may be Displays of electronic devices such as smartphones, tablets, etc.
- the number of displays 116 may be one or more.
- the display 116 may include a main screen and a touch screen, where the main screen is mainly used for displaying ultrasound images, and the touch screen is mainly used for human-computer interaction.
- Display 116 may display ultrasound images obtained by processor 114 .
- the display 116 can also provide a graphical interface for the operator to perform human-computer interaction while displaying the ultrasonic image.
- One or more controlled objects can be set on the graphical interface, and the operator can use the human-computer interaction device to input operation instructions to Control these controlled objects to perform corresponding control operations.
- an icon is displayed on a graphical interface, and the icon can be operated by using a human-computer interaction device to perform a specific function.
- the ultrasound imaging system 100 may further include other human-computer interaction devices other than the display 116, which are connected to the processor 114.
- the processor 114 may be connected to the human-computer interaction device through an external input/output port, and the external input/output port may be connected to the human-computer interaction device.
- the output port can be a wireless communication module, a wired communication module, or a combination of the two.
- External input/output ports may also be implemented based on USB, bus protocols such as CAN, and/or wired network protocols, and the like.
- the human-computer interaction device may include an input device for detecting the input information of the operator, for example, the input information may be a control instruction for the ultrasonic transmission/reception sequence, or may be a point, line or frame drawn on the ultrasonic image, etc.
- the operation input instruction or may also include other instruction types.
- the input device may include one or a combination of a keyboard, a mouse, a scroll wheel, a trackball, a mobile input device (eg, a mobile device with a touch display screen, a cell phone, etc.), a multi-function knob, and the like.
- the human-computer interaction apparatus may also include an output device such as a printer.
- the ultrasound imaging system 100 may also include memory for storing instructions executed by the processor, storing received ultrasound echoes, storing ultrasound images, and the like.
- the memory may be a flash memory card, solid state memory, hard disk, or the like. It may be volatile memory and/or non-volatile memory, removable memory and/or non-removable memory, and the like.
- the components included in the follicle tracking system 100 shown in FIG. 1 are merely illustrative, and may include more or fewer components. This application is not limited to this.
- FIG. 2 is a schematic flowchart of a follicle tracking method 200 according to an embodiment of the present application.
- This embodiment tracks the same target follicle of the tested object, obtains ultrasonic images of the same target follicle at different inspection times, and determines the development and growth of the same target follicle according to the acquired ultrasonic images, as shown in FIG. 3 , The operator can judge the egg retrieval time according to the growth and development of the same follicle. Compared with the existing tracking and monitoring of the overall follicle development, the tracking and monitoring of a single follicle in this implementation will provide more accurate clinical information, and the operator can evaluate the more accurate egg retrieval time accordingly.
- a follicle tracking method 200 includes the following steps:
- Step 201 Acquire at least three ultrasound images of the ovarian tissue of the test subject at different inspection times, where the ovarian tissue includes the target follicle.
- the processor may automatically acquire an ultrasound image including a target follicle from the plurality of ultrasound images, as an ultrasound image for tracking monitoring of the target follicle acquired at the current examination time; or The operator manually selects the ultrasound image including the target follicle from the acquired ultrasound images as the ultrasound image for the tracking monitoring of the target follicle at the current examination time.
- ultrasound images of the ovarian tissue of the same subject at different examination times can be acquired for multiple times, wherein the ovarian tissue includes the target follicle.
- the ultrasound image may be a two-dimensional ultrasound image, or may also be a three-dimensional ultrasound image.
- the subject will usually do four to six ultrasound examinations, at least no less than three ultrasound examinations, and obtain at least three growth parameters for the same target follicle.
- the growth trend chart based on at least three growth parameters will provide the operator with more accurate follicle growth trend information, and the operator will also be more accurate in evaluating the time of egg retrieval.
- Step 202 Determine the follicle regions corresponding to the target follicles on the ultrasound images of the at least three different inspection times respectively, and obtain at least three follicle regions, wherein the at least three follicle regions are follicles corresponding to the same target follicle area.
- the processor automatically determines the region corresponding to the target follicle on the ultrasound images of at least three different examination times; or the operator manually determines the region corresponding to the target follicle on the ultrasound images of at least three different examination times. In this way, the growth parameters of the ultrasound images obtained at different times of the same target follicle can be obtained, that is, the same target follicle can be tracked to evaluate the accurate egg retrieval time.
- the region corresponding to the same target follicle is determined on different ultrasound images, and the process can be determined automatically by the processor, or manually determined by the operator, or confirmed by the processor automatically and manually by the operator.
- the at least three ultrasonic images at different inspection times are respectively a first ultrasonic image, a second ultrasonic image and a third ultrasonic image, wherein the first ultrasonic image and the second ultrasonic image and the third ultrasound image are only for distinguishing the ultrasound images obtained at at least three different examination times, and “first”, “second” and “third” do not indicate the order of examination times.
- the follicle area corresponding to the target follicle is determined in the first ultrasonic image of the at least three ultrasonic images of different inspection times, and the first follicular area is obtained.
- the target follicle may be one target follicle, or may be two or more target follicles.
- the target follicle may grow until the egg is retrieved, or it may disappear before the egg is formed.
- the target follicles may disappear, or new target follicles may be added.
- several more target follicles are selected at the beginning of clinical practice to prevent the loss of target follicles in the follow-up process.
- the follicles with better growth will be selected as target follicles.
- the obtained ultrasound image is a two-dimensional ultrasound image
- a follicle with a larger diameter on the two-dimensional ultrasound image can be selected as the target follicle; or if the obtained ultrasound image is a three-dimensional ultrasound image, a larger follicle can be selected as the target follicle .
- the operator can manually select the target follicle, or the processor can automatically identify the target follicle.
- a plurality of first candidate follicle regions are determined based on the first ultrasound image, growth parameters of the plurality of first candidate follicle regions are determined based on the plurality of first candidate follicle regions, and the growth parameters satisfying the first preset condition are determined.
- the first candidate follicle region is determined as the first region of the target follicle.
- the growth parameter is the volume or diameter of the follicle. Taking the volume as an example, the first preset condition may be that the volume of the follicle is greater than a preset threshold, or the volume of the follicle is relatively large in the volume of the first candidate follicle, or the like.
- the first preset condition can be automatically set by the processor, or manually set by the operator.
- the first candidate follicle area with the largest growth parameter may also be determined as the first area of the target follicle.
- the first candidate follicle region with the largest volume may be determined as the first region of the target follicle, or the first candidate follicle region with the largest diameter and length may be determined as the target follicle region.
- the first region of the target follicle is determined based on the first ultrasound image of the ovarian tissue, which will be exemplified as a three-dimensional ultrasound image by using the ovarian tissue below.
- the processor performs segmentation on the follicle area corresponding to the target follicle on the first ultrasonic image in the ultrasonic images of the at least three different examination times based on the image feature of the follicle, to obtain the first follicular area; or
- the processor detects an operator's instruction to trace the corresponding region of the target follicle of the first ultrasound image to obtain the first follicle region.
- the processor automatically segments the corresponding area of the target follicle on the first ultrasound image based on the image features of the follicle, which can be roughly divided into two methods.
- One is a two-dimensional image segmentation method, which divides the first ultrasound image into multiple two-dimensional images. Sectioning, segmenting the corresponding area of the target follicle from multiple two-dimensional sections, and synthesizing the corresponding areas of the target follicle in the plurality of two-dimensional sections to obtain the first follicle area.
- the plurality of two-dimensional slice images of the first ultrasound image may be all the two-dimensional slice images in the first ultrasound image.
- the plurality of two-dimensional slices of the first ultrasound image may be sampled images obtained by sampling the first ultrasound image according to a preset rule. For example, a point is selected, and the sampled image is cut out around the point in an emission shape, as shown in FIG. 4 ; or the sampled image is cut in parallel, as shown in FIG. 5 . Three-dimensional interpolation is performed on the segmentation result of the sampled image to obtain the first follicle region.
- the traditional image segmentation algorithm is to detect the corresponding area of the target follicle through the target detection method (such as point detection, line detection), so as to segment the area to obtain the first follicle area.
- target detection method such as point detection, line detection
- Commonly used segmentation algorithms are based on level set (Level Set) segmentation algorithm, random walk (Random Walk), graph cut (Graph Cut), Snake and so on.
- Machine learning-based methods segment target follicles by learning the characteristics of the corresponding regions of the target follicles in the database.
- the main steps are: 1) Build a database, which should contain a large number of data sets and their corresponding labeling results.
- the labeling information is the mask information that accurately segmented the target follicle, and the mask information includes the boundary. Area information formed by information and boundary information.
- Segmentation step there are two main segmentation algorithms, one is a segmentation algorithm based on traditional machine learning, and the other is a segmentation algorithm based on deep learning, as follows:
- the common steps are to divide the first ultrasound image into many image blocks, and then extract features from the image blocks.
- the feature extraction methods include traditional PCA, LDA, Harr features, and texture features.
- the image block is the corresponding area of the target follicle
- the classification result is used as the center point marking result of the current image block, and finally the segmentation result of the entire image is obtained, that is, the first follicle area is obtained.
- the structure of the end-to-end semantic segmentation algorithm based on deep learning is to obtain an output image with the same size as the input first ultrasound image through stacking of convolutional layers, pooling layers, upsampling or deconvolutional layers, etc.
- the output image directly segments the corresponding area of the desired target follicle.
- This method is a supervised learning method, so the input supervision information is the mask information of the corresponding area of the target follicle.
- Data preparation is time-consuming, and the common two-dimensional Networks include FCN, U-Net, Mask R-CNN, etc.
- the other is a three-dimensional image segmentation method, that is, the processor directly performs three-dimensional segmentation in the first ultrasound image based on the image features of the follicle to determine the first follicle area, refer to shown in Figure 6.
- the segmentation based on the three-dimensional data can detect the corresponding area of the target follicle through the target detection method (such as point detection, line detection), so as to segment the area to obtain the first follicle area.
- This method does not require a large amount of labeled data.
- Commonly used segmentation algorithms include Level Set segmentation algorithm, Random Walk, Graph Cut, Snake, 3D Great Law, Clustering, Mal Koff random fields, etc.
- the region can also be segmented using machine learning methods by learning the characteristics of the corresponding region of the target follicle in the database.
- the main steps are: 1) Build a database, which should contain a large number of data sets and their corresponding labeling results.
- the labeling information is the mask (MASK) information that accurately segmented the corresponding area of the target follicle. 2) Segmentation step, there are two main segmentation algorithms, one is a segmentation algorithm based on traditional machine learning, and the other is a segmentation algorithm based on deep learning, as follows:
- the common steps are to divide the first ultrasound image into many image blocks, and then extract the features of the image blocks, such as edge detection based on the sobel operator to extract the edge features and gradients of the three-dimensional data, Texture features, etc., or use neural networks, such as Medicalnet, to extract features. Then, the extracted features are classified by cascaded classifiers, wherein the classifiers such as KNN, SVM, random forest and other discriminators are used to determine whether the current image block is the corresponding area of the target follicle, and the classification result is used as the current image block. Finally, the segmentation result of the entire image is obtained, that is, the first follicle area is obtained.
- the features of the image blocks such as edge detection based on the sobel operator to extract the edge features and gradients of the three-dimensional data, Texture features, etc., or use neural networks, such as Medicalnet, to extract features. Then, the extracted features are classified by cascaded classifiers, wherein the classifiers such as KNN, SVM, random
- the structure of the end-to-end semantic segmentation algorithm based on deep learning is to stack convolutional layers, pooling layers, upsampling or deconvolutional layers in three dimensions, so as to obtain an output with the same size as the input first ultrasound image.
- image the output image directly segments the required first follicle area.
- This method is a supervised learning method, so the input supervision information is the mask information of the target area, and data preparation is time-consuming.
- Common 3D segmentation networks include 3D U-Net , 3D FCN, Medical-Net, etc.
- the first ultrasound image is a two-dimensional ultrasound image
- the traditional segmentation algorithm and the image segmentation algorithm based on machine learning are as described above, and will not be repeated here.
- the follicle area corresponding to the target follicle is determined in the second ultrasonic images of at least three ultrasonic images of different inspection times, and the second follicular area is obtained;
- the follicle area corresponding to the target follicle is determined in the third ultrasonic image of the at least three ultrasonic images of different inspection times, and the third follicle area is obtained.
- the process is described by taking the obtaining of the second follicle region in the second ultrasound image based on the first follicle region as an example.
- the processor may automatically determine the follicle area corresponding to the target follicle in the second ultrasound image according to the first follicle area to obtain the second follicle area; or the operator may manually identify the follicle area corresponding to the target follicle in the second ultrasound image.
- the processor detects the operator's manual marking operation, and automatically determines the marked follicle area as the second follicle area, wherein the operator's manual identification includes the operator manually determining the second follicle area, such as The second follicular area is traced, or may also include highlighting of the identified second follicular area.
- the processor automatically determines the second follicle region in the second ultrasound image according to the first follicle region mainly includes two methods, one is the traditional image matching method, that is, based on the feature information of the first follicle region, the second ultrasound image Searching for sub-images similar to the characteristic information of the first follicle region in the middle of the process requires iteration to find the optimal solution, which is slow and does not require a large amount of data.
- the processor automatically determines, from the second ultrasound image, a second region whose feature information similarity with the first follicle region satisfies a first threshold, and determines the second region that satisfies the first threshold as the second follicle region.
- the first threshold may be a percentage, such as 90%.
- the processor may automatically set the first threshold; or the operator may manually set the first threshold according to clinical needs.
- the characteristic information of the target follicle may be the image signal information obtained by transforming the image signal from the time domain to the frequency domain by some transformation methods (such as Fourier transform, inverse transforming the image to the frequency domain, transformation); it can also be the original data information directly from the image, the original data information refers to the data information obtained directly from the image without any processing, such as gray value, contrast, etc.;
- image feature information extracted after processing image features can be mainly divided into points, lines, regions and other features, and can also be divided into local features and global features, among which point features and line features are used more, and point features mainly include Harris, Moravec , KLT, Harr-like, HOG, LBP, SIFT, SURF, BRIEF, SUSAN, FAST, CENSUS, FREAK, BRISK, ORB, optical flow method and A-KAZE, etc.
- the line features mainly include LoG operator, Robert operator, Sobel operator, Prewitt operator, Canny operator, etc.
- the traditional image matching methods are divided into three types corresponding to the feature information: 1) The method based on domain transformation, this method is based on the frequency domain signal of the first follicle area for matching, mainly Using phase correlation, Walsh transform, wavelet transform and other methods. 2) A method based on template matching, which is based on the original data information (such as gray value, contrast, etc.) of the first follicle area for matching, and searches for the second ultrasound image according to the original data information of the first follicle area. A sub-image whose similarity of the follicular area satisfies the first threshold is determined as the second follicular area. In this process, there is no need to extract the characteristic information of the ultrasound image.
- matching can be performed based on grayscale, and a grayscale-based matching algorithm is also called a correlation matching algorithm, and a spatial two-dimensional sliding template is used for matching.
- the commonly used algorithms in the method based on template matching include mean absolute difference algorithm (MAD), sum of absolute error algorithm (SAD), error sum of squares algorithm (SSD), mean error sum of squares algorithm (MSD), normalized product correlation algorithm ( NCC), Sequential Similarity Detection Algorithm (SSDA), Hadamard Transform Algorithm (SATD), Local Gray Value Coding Algorithm and PIU, etc.
- a matching method based on image features the method first extracts the image feature information of the first follicle region, then generates a feature description operator according to the extracted image feature information, and finally determines the similarity with the description operator in the second ultrasound image.
- the second follicular area that meets the first threshold is the first follicular area that meets the first threshold.
- an image registration method based on deep learning refers to determining the correspondence between two image areas through a trained learning model, and making the two image areas in the spatial position through at least one of the three operations of rotation, translation and scaling. Relational alignment of orientation and size. In this process, a large number of labeled samples are needed to establish a database, and image registration can be achieved in one step.
- the processor automatically or the operator manually acquires the second candidate follicle area corresponding to the target follicle in the second ultrasound image, and the processor uses the first learning model to determine the correspondence between the first follicle area and the second candidate follicle area, and according to the correspondence The relationship adjusts the first candidate follicle area, and determines the adjusted second candidate follicle area as the second follicle area.
- the main steps are: 1) Build a database.
- the database contains a large number of datasets and their corresponding tag information.
- the tag information is the corresponding relationship of alignment, such as a rotation relationship, a translation relationship, a scaling relationship, or a Any combination of the above three relationships; 2) the registration step, the first follicle area and the second candidate follicle area are input into the database, the corresponding relationship is determined, and the second candidate follicle area of the target follicle is adjusted according to the determined relationship. , wherein the adjustment may be through at least one operation mode of rotation, translation or zooming, and the adjusted second candidate follicle area is output and determined as the second follicle area.
- image registration methods based on deep learning There are two types of image registration methods based on deep learning, one is a two-stage image registration method; the other is an end-to-end image registration method.
- the two-stage registration method based on deep learning is similar to the feature-based matching method, but here, traditional features such as SIFT are replaced with deep learning networks for feature extraction.
- the end-to-end image registration method based on deep learning can be divided into two types.
- One is supervised learning.
- the first follicle region and the second candidate follicle region are input into the trained first learning model, usually through Stacking algorithm of three-dimensional network structures such as convolutional layer, pooling layer, upsampling or deconvolutional layer, so as to output the correspondence between the first follicle area and the second candidate follicle area, and the input supervision information is the correspondence, common
- the network structure includes FCN, U-Net, etc.
- One is unsupervised learning, that is, the input is the first follicle area and the second candidate follicle area, and the output is the adjusted second candidate follicle area.
- the corresponding regions of the same target follicle in the ultrasonic images obtained at different inspection times are sequentially acquired, and the continuous tracking and monitoring of the same target follicle in the ultrasonic images obtained at different inspection times is completed.
- the ovarian tissue regions in different ultrasound images can also be registered or matched, or the whole image can be registered, and then the tracking and monitoring of the target follicle can be performed.
- the target follicle grows faster, and it is precisely because the size of the target follicle is constantly changing, so the size of the same target follicle on the ultrasound images obtained at different inspection times.
- the diameter or volume of the target follicle will be different, and it is precisely because of this that it is difficult to continuously track and monitor the same target follicle.
- the relative position of the same target follicle in the ovarian tissue will not change much, so the ovarian tissue area can be determined in the second ultrasound image first, and then based on the first ultrasound image in the first ultrasound image.
- the relative positional relationship between the target follicle and the first ultrasound image determines the second follicle area on the second ultrasound image, which can reduce the influence of the structure outside the ovary similar to the target follicle, and can also reduce the computational load of the processor, making the result more accurate. Accurate and more efficient to process.
- the first ultrasound image and the second ultrasound image are images obtained from the same part or approximately the same part
- the first ultrasound image and the second ultrasound image can also be adjusted as a whole by at least one operation of rotation, translation and zooming. , to track and monitor the target follicle after the two are registered.
- This method has relatively high requirements on the manual operation of the operator to check the time before and after.
- the acquired ultrasound slices may not include the entire ovarian tissue due to operator manipulation or other factors, which brings inconvenience to the tracking and monitoring of the target follicles, and may even be acquired during the acquisition.
- the target follicles are not included in the ultrasound images of the 2D ultrasound images, so the tracking and monitoring of the same target follicles through two-dimensional ultrasound images requires relatively high operator skills.
- the follicle tracking is performed by three-dimensional ultrasound, the operator can easily obtain the image of the entire ovarian tissue, so that the tracking of the target follicle will be easier to obtain, and the operator's manual requirements can be reduced.
- the registration or matching of the ovarian tissue region or the registration of the whole image is not a necessary step, which may or may not be used.
- the first ovarian region of the ovarian tissue of the test subject is determined in the first ultrasound image
- the second ovarian region of the ovarian tissue of the test subject is determined in the second ultrasound image
- the first ovarian region and the second ovarian region are determined. Two ovarian regions were registered.
- the processor may automatically register the first ovarian region and the second ovarian region, or the operator may perform the registration manually.
- the processor uses a machine learning method based on target segmentation or a traditional segmentation algorithm to determine the first ovarian region in the first ultrasound image, or the operator manually determines the first ovarian region.
- the orientation of the second ovarian region in the second ultrasound image is adjusted according to the orientation of the first ovarian region in the first ultrasound image, so that the adjusted orientation of the second ovarian region in the second ultrasound image
- the orientation of the first ovarian region is the same as that of the first ovarian region in the first ultrasound image; or according to the size of the first ovarian region, the length of the diameter of the cut surface or the size of the volume can be used to adjust the size of the second ovarian region relative to the first ovarian region. size so that the adjusted second ovarian region is the same size as the first ovarian region.
- the adjustment of the orientation and the adjustment of the size can be adjusted automatically by the processor or manually by the operator.
- the two ovarian data are not aligned, for example, the position of the first ovarian region in the first ultrasound image is to the left in the first ultrasound image.
- the position of the second ovarian region in the second ultrasound image is to the right or in the center.
- the orientation of the second ovarian region needs to be adjusted so that the orientation of the second ovarian region in the second ultrasound image is the same as that of the first ovarian region.
- the ovarian regions are in the same orientation in the first ultrasound image.
- the size here can be the size of the volume, or it can be the size of the diameter.
- the second ovarian area needs to be enlarged. so that the adjusted second ovarian area is the same size as the first ovarian area.
- the same orientation and the same size include the same absolute orientation and the same absolute size, and also include the same approximate orientation and the same approximate size.
- the registration of the ovarian tissue area is not an essential step but only an optimization step, so the ovarian tissue area
- the ability to achieve approximate registration can basically meet the clinical needs.
- the processor may use a machine learning method to automatically adjust the second ovarian region, so that the adjusted second ovarian region is registered with the first ovarian region, as shown in FIG. 8 .
- the processor uses the second learning model to determine the corresponding relationship between the first ovarian region and the second ovarian region, such as a rotation relationship, or a translation relationship, or a scaling relationship, or any combination of the above three relationships, and according to The determined correspondence adjusts the second ovarian region such that the second follicular region is registered with the first follicular region.
- the main steps are: 1) building a database, the database contains a large number of data sets and their corresponding marking information, and the marking information is a corresponding relationship; 2) a registration step, the first ovarian region and the second ovarian region are input into the database, The corresponding relationship between the first ovarian region and the second ovarian region is determined, the second ovarian region is adjusted according to the determined corresponding relationship, and the adjusted second ovarian region is output.
- the machine learning method also includes a deep learning method.
- the registration of ovarian tissue based on deep learning is similar to the above-mentioned registration method of target follicles based on deep learning, and will not be repeated here.
- the registration of the ovarian region can also be performed in a manner that does not require first determination of the second ovarian region, such as a traditional image matching method.
- the traditional image matching method is to search the second ultrasound image for sub-images whose similarity satisfies a preset threshold based on the feature information of the first ovarian region, and determine the sub-image whose similarity satisfies the preset threshold as the second follicular region.
- the traditional image registration methods include: methods based on domain transformation; methods based on template matching; matching methods based on image features. Determining the first ovarian region using the traditional image matching method is similar to determining the first follicle region using the traditional image matching method described above, and will not be repeated here.
- the operator can manually achieve the registration of the ovarian region based on the method of overall image registration.
- An operator's instruction to register the second ultrasound image with the first ultrasound image is received, and the second ultrasound image is rotated, translated or scaled according to the instruction to register the second ultrasound image with the first ultrasound image.
- new follicles may appear, or the target follicle may disappear.
- the target follicles that disappear in the image will be deleted; for the target follicles that appear in the next ultrasound image, if the growth parameters of the newly added follicles meet the preset conditions, the newly added follicles whose growth parameters meet the preset conditions will be regarded as new follicles.
- the target follicles are added to the tracking ranks, otherwise removed.
- the preset condition is that the volume satisfies a certain threshold condition or the path length satisfies a certain threshold condition, and the specific threshold size can be automatically set by the processor, or manually set by the operator.
- Step 203 Determine the growth parameters of the target follicle according to the at least three follicle regions, respectively, and obtain at least three growth parameters of the target follicle.
- the growth parameters include at least one of the following: volume, diameter and growth rate.
- the processor can automatically determine the growth parameters of the target follicle according to at least three follicle regions or the operator can manually determine the growth parameters of the target follicle to obtain at least three growth parameters of the target follicle, as shown in FIG. Obtain the long and short diameters of the target follicles.
- the acquired ultrasound image is a three-dimensional ultrasound image.
- the volume of the target follicle can be measured based on the voxels of the segmentation result and the distance between the voxels; the diameter of the target follicle can be measured by fitting the segmentation result of the target follicle area into an ellipsoid, and the target follicle can be determined from the ellipsoid.
- the section with the largest area or the area of the target follicle area meets the preset requirements.
- the preset requirements can be customized by the user or set by the machine.
- the diameter with the largest length is determined on the determined section as the long diameter of the target follicle.
- the longest short diameter perpendicular to the long diameter is determined as the short diameter of the target follicle, and the vertical relationship includes absolute vertical and approximately vertical, for example, clinically between 85 degrees and 95 degrees can basically be considered vertical relation.
- the acquired ultrasound image is a two-dimensional ultrasound image.
- an ellipse is fitted according to the target follicle area, and the long and short diameters of the target follicle are determined based on the ellipse.
- the growth rate of the target follicle can be determined based on the obtained volume or the ratio of the diameter and the growth time.
- Step 204 obtaining a growth trend graph of the target follicle according to at least three growth parameters of the target follicle.
- Step 205 displaying a growth trend graph.
- the growth trend graph of the target follicle may be a growth parameter graph, wherein the growth parameter graph takes the inspection time as the first coordinate and the growth parameter as the second coordinate, for example, as shown in FIG. 10 .
- the growth trend graph of the target follicle may be a list of growth parameters corresponding to different inspection times, as shown in FIG. 11 .
- a single target follicle can be tracked and monitored, and finally a growth trend graph of a single target follicle can be obtained.
- the optimal egg retrieval time can effectively improve work efficiency and accuracy.
- FIG. 12 is a schematic flowchart of a follicle tracking method 300 according to an embodiment of the present application.
- a follicle tracking method 300 includes the following steps:
- Step 301 acquiring a first ultrasound image of the ovarian tissue of the test object, wherein the ovarian tissue includes the target follicle.
- the processor automatically or the operator manually acquires a first ultrasound image of the ovarian tissue of the subject, wherein the ovarian tissue includes the target follicle.
- the ultrasound image may be a two-dimensional ultrasound image, or may also be a three-dimensional ultrasound image.
- the test subject usually performs multiple ultrasound examinations on the test subject during the process from the injection of the ovulation induction drug to the egg retrieval. Based on clinical needs, at least two ultrasound examinations are generally performed, and the same target follicle is tracked based on the acquisition of at least two ultrasound images at different examination times. Of course, if the same target follicle is tracked by obtaining three to six ultrasound images at different examination times, the obtained egg retrieval time will be more accurate.
- Step 302 determining a follicle area corresponding to the target follicle in the first ultrasound image to obtain a first follicle area.
- the target follicle may be one target follicle, or may be two or more target follicles.
- the target follicle may continue to grow and eventually become an egg to be retrieved, or it may disappear without forming an egg.
- the target follicles may disappear, or new target follicles may be added.
- several more target follicles are selected at the beginning of clinical practice to prevent the loss of target follicles in the follow-up process.
- the follicles with better growth will be selected as target follicles.
- the follicles that grow better may be follicles with larger diameters on the cut surface, or may be larger follicles, or may also be follicles with longer diameters.
- the doctor can manually select or the processor can automatically identify better growing follicles and list them as target follicles.
- a plurality of first candidate follicle regions are determined based on the first ultrasound image, growth parameters of the plurality of first candidate follicle regions are determined based on the plurality of first candidate follicle regions, and the growth parameters satisfying the first preset condition are determined.
- the first candidate follicle region is determined as the first region of the target follicle.
- the growth parameter is the volume or diameter of the follicle. Taking the volume as an example, the first preset condition may be that the volume of the follicle is greater than a certain threshold, or the volume of the follicle is relatively large in the volume of the first candidate follicle, or the like.
- the first preset condition can be automatically set by the processor, or manually set by the operator.
- the first candidate follicle region with the largest growth parameter may also be determined as the first region of the target follicle.
- the first candidate follicle region with the largest volume may be determined as the first region of the target follicle, or the first candidate follicle region with the largest diameter and length may be determined as the target follicle region.
- the first ultrasound image of the ovarian tissue may be a two-dimensional ultrasound image, or may also be a three-dimensional ultrasound image.
- the processor may automatically segment the follicle area corresponding to the target follicle on the first ultrasonic image to obtain the first follicle area; or the processor may detect the operator's instruction to trace the corresponding area of the target follicle in the first ultrasonic image to obtain the first follicle area.
- First follicular area may be automatically segment the follicle area corresponding to the target follicle on the first ultrasonic image to obtain the first follicle area; or the processor may detect the operator's instruction to trace the corresponding area of the target follicle in the first ultrasonic image to obtain the first follicle area.
- the processor automatically segments the target follicle on the first ultrasound image based on the image features of the follicle.
- the multiple two-dimensional slice images of the first ultrasound image are all the two-dimensional slice images in the first ultrasound image; or the multiple two-dimensional slice images of the first ultrasound image are images of the first ultrasound image with Preset rules are used to obtain sampled images, and three-dimensional interpolation is performed on the segmentation results of the sampled images to obtain the first follicle region.
- the other is a three-dimensional image segmentation method, in which the processor directly performs three-dimensional segmentation in the first ultrasound image based on the image features of the follicle, so as to determine the boundary of the first follicle area.
- the segmentation based on 3D data can detect the corresponding area of the target follicle through target detection methods (such as point detection, line detection), so as to segment the area. This method does not require a large amount of labeled data.
- target detection methods such as point detection, line detection
- the commonly used segmentation algorithms are based on level set ( Level Set) segmentation algorithm, random walk (Random Walk), graph cut (Graph Cut), Snake, three-dimensional big law, clustering, Markov random field, etc.
- the region can also be segmented using machine learning methods by learning the characteristics of the corresponding region of the target follicle in the database.
- the main steps are: 1) Build a database.
- the database should contain a large number of datasets and their corresponding labeling results.
- the labeling information is the boundary information that accurately segmented the target.
- Segmentation step there are two main segmentation algorithms, one is a segmentation algorithm based on traditional machine learning, and the other is a segmentation algorithm based on deep learning. The segmentation method based on machine learning has been described above and will not be repeated here.
- Step 303 acquiring a second ultrasound image of the ovarian tissue of the tested object.
- the processor automatically acquires the second ultrasound image of the ovarian tissue of the subject, or the operator manually determines the second ultrasound image of the ovarian tissue of the subject.
- Step 304 based on the first follicle area, determine a follicle area corresponding to the target follicle in the second ultrasound image, and obtain a second follicle area.
- the processor may automatically determine the follicle area corresponding to the target follicle in the second ultrasound image according to the first follicle area to obtain the second follicle area; or the operator may manually identify the follicle area corresponding to the target follicle in the second ultrasound image.
- the processor detects the operator's manual marking operation, and automatically determines the marked follicle area as the second follicle area.
- the processor automatically determines the second follicle region in the second ultrasound image according to the first follicle region mainly includes two methods, one is the traditional image matching method, that is, based on the feature information of the first follicle region, the second ultrasound image Searching for sub-images similar to the characteristic information of the first follicle region in the middle of the process requires iteration to find the optimal solution, which is slow and does not require a large amount of data.
- the processor automatically determines, from the second ultrasound image, a second region whose similarity information to the feature information of the first follicular region satisfies a first threshold, and determines the second region that satisfies the first threshold as the second follicular region.
- the first threshold may be a percentage, such as 90%.
- the processor may automatically set the first threshold; or the operator may manually set the first threshold according to clinical needs.
- the characteristic information of the target follicle may be the image signal information obtained by transforming the image signal from the time domain to the frequency domain by some transformation methods (such as Fourier transform, inverse transforming the image to the frequency domain, transformation); it can also be the original data information directly from the image, the original data information refers to the data information obtained directly from the image without any processing, such as gray value, contrast, etc.; or it can be extracted from the original data information
- Image features can be mainly divided into points, lines, regions and other features, and can also be divided into local features and global features.
- point features and line features are widely used, and point features mainly include Harris, Moravec, KLT, Harr-like, HOG, LBP, SIFT, SURF, BRIEF, SUSAN, FAST, CENSUS, FREAK, BRISK, ORB, optical flow method and A-KAZE, etc.
- Line features mainly include LoG operator, Robert operator, Sobel operator , Prewitt operator, Canny operator, etc.
- the traditional image matching methods are specifically divided into three types corresponding to the feature information: 1) The method based on domain transformation, this method is based on the frequency domain information of the first follicle area for matching, mainly Using phase correlation, Walsh transform, wavelet transform and other methods. 2) A method based on template matching, which is based on the original data information of the first follicle area, such as gray value, etc., according to the original data information of the first follicle area to search for the second ultrasound image and the first follicle area. For the sub-images whose similarity satisfies the first threshold, the sub-images whose similarity meets the first threshold are determined as the second follicle region.
- a matching method based on image features the method first extracts the image feature information of the first follicle region, then generates a feature description operator according to the extracted image feature information, and finally determines the similarity with the description operator in the second ultrasound image. A second region that satisfies the first threshold.
- an image registration method based on deep learning In addition to the traditional image matching method, there is another method for determining the second follicle area in the second ultrasound image based on the first follicle area, that is, an image registration method based on deep learning.
- the image registration method based on deep learning requires a large number of labeled samples to establish a database, which can realize image registration in one step.
- the processor automatically or the operator manually acquires the second candidate follicle area corresponding to the target follicle in the second ultrasound image, and the processor uses the first learning model to determine the correspondence between the first follicle area and the second candidate follicle area, and according to the correspondence The relationship adjusts the first candidate follicle area, and determines the adjusted second candidate follicle area as the second follicle area.
- the main steps are: 1) Build a database.
- the database contains a large number of datasets and their corresponding tag information.
- the tag information is the corresponding relationship of alignment, such as a rotation relationship, a translation relationship, a scaling relationship, or a Any combination of the above three relationships; 2) the registration step, the first follicle area and the second candidate follicle area are input into the database, the corresponding relationship is determined, and the second candidate follicle area of the target follicle is adjusted according to the determined relationship. , wherein the adjustment may be through at least one operation mode of rotation, translation or zooming, and the adjusted second candidate follicle area is output and determined as the second follicle area.
- image registration methods based on deep learning There are two types of image registration methods based on deep learning, one is a two-stage image registration method; the other is an end-to-end image registration method.
- the specific deep learning-based image registration method is as described above, and will not be repeated here.
- the corresponding regions of the same target follicle in the ultrasonic images obtained at different inspection times are sequentially acquired, and the continuous tracking and monitoring of the same target follicle in the ultrasonic images obtained at different inspection times is completed.
- the ovarian tissue regions in different ultrasound images can also be registered or matched, or the whole image can be registered, and then the tracking and monitoring of the target follicle can be performed.
- the first ovarian region of the ovarian tissue of the test subject is determined in the first ultrasound image
- the second ovarian region of the ovarian tissue of the test subject is determined in the second ultrasound image
- the first ovarian region and the second ovarian region are determined. Two ovarian regions were registered.
- the processor may automatically register the first ovarian region and the second ovarian region, or the operator may perform the registration manually.
- the processor uses a machine learning method based on target segmentation or a traditional segmentation algorithm to determine the first ovarian region in the first ultrasound image, or the operator manually determines the first ovarian region.
- the orientation of the second ovarian region in the second ultrasound image is adjusted according to the orientation of the first ovarian region in the first ultrasound image, so that the adjusted orientation of the second ovarian region in the second ultrasound image
- the orientation of the first ovarian region is the same as that of the first ovarian region in the first ultrasound image; or according to the size of the first ovarian region, the length of the diameter of the cut surface or the size of the volume can be used to adjust the size of the second ovarian region relative to the first ovarian region. size so that the adjusted second ovarian region is the same size as the first ovarian region.
- the adjustment of the orientation and the adjustment of the size can be adjusted automatically by the processor or manually by the operator.
- the processor may use a machine learning method to automatically adjust the second ovarian region so that the adjusted second ovarian region is registered with the first ovarian region.
- the processor uses the second learning model to determine the corresponding relationship between the first ovarian region and the second ovarian region, such as a rotation relationship, or a translation relationship, or a scaling relationship, or any combination of the above three relationships, and according to The determined correspondence adjusts the second ovarian region such that the second follicular region is registered with the first follicular region.
- the main steps are: 1) building a database, the database contains a large number of data sets and their corresponding marking information, and the marking information is a corresponding relationship; 2) a registration step, the first ovarian region and the second ovarian region are input into the database, The corresponding relationship between the first ovarian region and the second ovarian region is determined, the second ovarian region is adjusted according to the determined corresponding relationship, and the adjusted second ovarian region is output.
- the machine learning method also includes a deep learning method.
- the registration of ovarian tissue based on deep learning is similar to the above-mentioned registration method of target follicles based on deep learning, and will not be repeated here.
- the registration of the ovarian region can also be performed in a manner that does not require first determination of the second ovarian region, such as a traditional image registration method.
- the traditional image registration method is to search the second ultrasound image for sub-images whose similarity satisfies a preset threshold based on the feature information of the first ovarian region, and determine the sub-image whose similarity satisfies the preset threshold as the second follicular region.
- the traditional image registration methods include: methods based on domain transformation; methods based on template matching; matching methods based on image features. Using the traditional image matching method to determine the second ovarian region is similar to the above-mentioned determination of the second follicle region using the traditional image matching method, and will not be repeated here.
- the operator can manually achieve the registration of the ovarian region based on the method of overall image registration.
- An operator's instruction to register the second ultrasound image with the first ultrasound image is received, and the second ultrasound image is rotated, translated or scaled according to the instruction to register the second ultrasound image with the first ultrasound image.
- the target follicles In the process of registering the target follicle, there will be new follicles or disappearance of the target follicle.
- the target follicles that disappear in the image will be deleted; for the target follicles that appear in the next ultrasound image, if the growth parameters of the newly added follicles meet the preset conditions, the follicle will be added as a new target follicle to the track, otherwise it will be deleted.
- the preset condition is that the volume satisfies a certain threshold condition or the path length satisfies a certain threshold condition, and the specific threshold size can be automatically set by the processor, or manually set by the operator.
- the same target follicle can be tracked and monitored on multiple ultrasound images, and the growth and development of the same target follicle can be clinically achieved. Operators can accurately assess the optimal egg retrieval time, effectively improving work efficiency and accuracy.
- the disclosed apparatus and method may be implemented in other manners.
- the device embodiments described above are only illustrative.
- the division of the units is only a logical function division. In actual implementation, there may be other division methods.
- multiple units or components may be combined or May be integrated into another device, or some features may be omitted, or not implemented.
- Various component embodiments of the present application may be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof.
- a microprocessor or a digital signal processor (DSP) may be used in practice to implement some or all functions of some modules according to the embodiments of the present application.
- DSP digital signal processor
- the present application can also be implemented as a program of apparatus (eg, computer programs and computer program products) for performing part or all of the methods described herein.
- Such a program implementing the present application may be stored on a computer-readable medium, or may be in the form of one or more signals. Such signals may be downloaded from Internet sites, or provided on carrier signals, or in any other form.
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Ultra Sonic Daignosis Equipment (AREA)
Abstract
一种卵泡跟踪方法和系统,该方法包括:获取被测对象的卵巢组织的至少三个不同检查时间的超声图像,其中卵巢组织包括目标卵泡(201);分别在至少三个不同检查时间的超声图像上确定目标卵泡对应的卵泡区域,获得至少三个卵泡区域,其中至少三个卵泡区域为同一个目标卵泡对应的卵泡区域(202);根据至少三个卵泡区域分别确定目标卵泡的生长参数,获得目标卵泡的至少三个生长参数(203);根据目标卵泡的至少三个生长参数获得目标卵泡的生长趋势图(204),显示生长趋势图(205)。该方法在不同检查时间获得超声图像中针对同一个目标卵泡进行跟踪检测并获取同一个目标卵泡的生长参数趋势图,操作者可以根据目标卵泡的生长趋势图准确评估最佳的取卵时间,有效提高评估取卵时间的准确性工作效率。
Description
本申请涉及卵泡跟踪技术领域,更具体地涉及一种卵泡跟踪方法和系统。
目前很多家庭存在不孕不育的问题,试管婴儿解决不孕不育的问题的主要方式之一。试管婴儿的关键步骤就是卵子的获取。患者的自然周期通常只有一个优势卵泡得到一个胚胎,所以为了提升移植的成功率,需要采用控制性超排卵来增强与改善卵巢功能,以达到不受自然周期的限制、获得多个健康卵子的目的。为了获得健康卵子,取卵时间至关重要。
一般患者在注射促排卵药物进行卵巢刺激后,在排卵周期内需要进行4-6次跟踪检查来监测卵泡的发育情况。因为无法定位跟踪单个卵泡的生长发育情况,目前的监测基本都集中在判断卵泡总体的发育情况,导致获取的信息不准确,无法针对单个卵泡获取准确的取卵时间。
发明内容
在发明内容部分中引入了一系列简化形式的概念,这将在具体实施方式部分中进一步详细说明。本发明的发明内容部分并不意味着要试图限定出所要求保护的技术方案的关键特征和必要技术特征,更不意味着试图确定所要求保护的技术方案的保护范围。
一个实施例中,提供了一种卵泡跟踪方法,包括:
获取被测对象的卵巢组织的至少三个不同检查时间的超声图像,其中所述卵巢组织包括目标卵泡;
分别在所述至少三个不同检查时间的超声图像上确定所述目标卵泡对应的卵泡区域,获得至少三个卵泡区域,其中所述至少三个卵泡区域为同一个目标卵泡对应的卵泡区域;
根据所述至少三个卵泡区域分别确定所述目标卵泡的生长参数,获得所述目标卵泡的至少三个生长参数;
根据所述目标卵泡的至少三个生长参数获得所述目标卵泡的生长趋势图;
显示所述生长趋势图。
一个实施例中,所述生长趋势图为生长参数曲线图,其中所述生长参数曲线图以检查时间为第一坐标,以生长参数为第二坐标;或者,
所述生长趋势图为所述不同检查时间对应的所述生长参数的列表。
一个实施例中,分别在所述至少三个不同检查时间的超声图像上确定所述目标卵泡对应的卵泡区域包括:
在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域;
根据所述第一卵泡区域,在所述至少三个不同检查时间的超声图像中的第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域;
根据所述第一卵泡区域或第二卵泡区域,在所述至少三个不同检查时间的超声图像中 的第三超声图像中确定所述目标卵泡对应的卵泡区域,获得第三卵泡区域。
一个实施例中,在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:
基于卵泡的图像特征,在所述至少三个不同检查时间的超声图像中的第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域;或者,
检测操作者对所述至少三个不同检查时间的超声图像中的第一超声图像中的目标卵泡的对应区域描迹的操作,以获得第一卵泡区域。
一个实施例中,所述超声图像为三维超声图像,所述基于卵泡的图像特征,在所述至少三个不同检查时间的超声图像中的第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域,包括:
基于卵泡的图像特征,在所述至少三个不同检查时间的三维超声图像中的第一超声图像的多个二维切面图像中对所述目标卵泡的对应区域进行分割;
综合所述多个二维切面图像上所述目标卵泡的对应区域,以得到所述第一卵泡区域。
一个实施例中,所述第一超声图像的多个二维切面图像为所述第一超声图像中的所有二维切面图像,或者,
所述第一超声图像的多个二维切面为对所述第一超声图像中以第一预设规则进行采样得到的采样图像,所述综合所述多个二维切面上的所述目标卵泡的对应区域包括:对所述采样图像的分割结果进行三维插值,以得到所述第一卵泡区域。
一个实施例中,所述超声图像为三维超声图像,所述基于卵泡的图像特征,在所述至少三个不同检查时间的超声图像中的第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域,包括:
基于所述卵泡的图像特征,在所述至少三个不同检查时间的三维超声图像中的第一超声图像中对所述目标卵泡对应的卵泡区域进行三维分割,以获得所述第一卵泡区域。
一个实施例中,在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:
在所述至少三个不同检查时间的超声图像中的第一超声图像中确定所述卵巢组织的区域,获得第一卵巢区域;
基于所述第一卵巢区域确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域。
一个实施例中,在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:
在所述至少三个不同检查时间的超声图像中的第一超声图像中确定多个第一候选卵泡区域;
获取每个第一候选卵泡区域的生长参数,将生长参数满足第一预设条件的第一候选卵泡区域确定为所述第一卵泡区域。
一个实施例中,在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:
在所述至少三个不同检查时间的超声图像中的第一超声图像中确定多个第一候选卵泡区域;
获取每个第一候选卵泡区域的生长参数,将生长参数为最大值的第一候选卵泡区域确定为所述第一卵泡区域。
一个实施例中,所述根据所述第一卵泡区域,在所述至少三个不同检查时间的超声图像中的第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域,包括:
在所述至少三个不同检查时间的超声图像中的第二超声图像中获取与所述第一卵泡区域的特征信息相似度满足第一阈值的卵泡区域,将获取的卵泡区域确定为第二卵泡区域;或者,
获取所述第二超声图像中的目标卵泡对应的第二候选卵泡区域,采用第一学习模型确定所述第一卵泡区域与所述第二候选卵泡区域的对应关系,并根据所述对应关系调整所述第二候选卵泡区域,并将调整后的所述第二候选卵泡区域确定为第二卵泡区域。
一个实施例中,所述根据所述第一卵泡区域,在所述至少三个不同检查时间的超声图像中的第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域,包括:
检测操作者在所述至少三个不同检查时间的超声图像中的第二超声图像中标识与所述第一卵泡区域相对应的卵泡区域的操作,将标识的卵泡区域确定为第二卵泡区域。
一个实施例中,在上述实施例的基础上,还包括:
在所述第一超声图像中确定所述被测对象的卵巢组织的第一卵巢区域;
在所述第二超声图像中确定所述被测对象的卵巢组织的第二卵巢区域;
将所述第一卵巢区域与所述第二卵巢区域进行配准。
一个实施例中,将所述第一卵巢区域与所述第二卵巢区域进行配准,包括:
根据所述第一卵巢区域在所述第一超声图像中的方位,调整所述第二卵巢区域在所述第二超声图像中的方位,以使得所述调整后的第二卵巢区域在所述第二超声图像中的方位与所述第一卵巢区域在所述第一超声图像中的方位相同;或者,
根据所述第一卵巢区域的大小,调整所述第二卵巢区域相对于所述第一卵巢区域的大小,以使得所述调整后的第二卵巢区域与所述第一卵巢区域的大小相同。
一个实施例中,将所述第一卵巢区域与第二卵巢区域进行配准,包括:
采用第二学习模型确定所述第二卵巢区域与所述第一卵巢区域的对应关系,并根据所述对应关系调整所述第二卵巢区域;或者,
接收操作者根据所述第一卵巢区域调整所述第二卵巢区域的指令,根据所述指令调整所述第二卵巢区域。
一个实施例中,在上述实施例的基础上,还包括:
在所述第一超声图像中确定所述被测对象的卵巢组织的第一卵巢区域;
将所述第二超声图像与所述第一卵巢区域的模板进行匹配,并根据匹配结果确定所述第二超声图像中所述卵巢组织的第二卵巢区域。
一个实施例中,在上述实施例的基础上,还包括:
接收操作者对所述第二超声图像和所述第一超声图像配准的指令,根据所述指令旋转、平移或缩放第二超声图像,以使得所述第二超声图像和所述第一超声图像配准。
一个实施例中,所述生长参数包括以下至少之一:体积、径长和生长速度。
一个实施例中,提供了一种卵泡跟踪系统,包括:
超声探头;
发射电路,用于激励所述超声探头向被测对象的卵巢组织发射超声波;
接收电路,用于激励所述超声探头接收所述超声波的回波,以获得所述超声波的回波信号;
处理器,用于执行如上述实施例中任一项所述的卵泡跟踪方法。
显示器,用于显示所述生长趋势图。
一个实施例中,提供了一种卵泡跟踪方法,包括:
获取被测对象的卵巢组织的至少两个不同检查时间的超声图像,其中所述卵巢组织包括目标卵泡;
分别在所述至少两个不同检查时间的超声图像上确定所述目标卵泡对应的卵泡区域,获得至少两个卵泡区域,其中所述至少两个卵泡区域为同一个目标卵泡对应的卵泡区域;
根据所述至少两个卵泡区域分别确定所述目标卵泡的生长参数,获得所述目标卵泡的至少两个生长参数;
根据所述目标卵泡的至少两个生长参数获得所述目标卵泡的生长趋势图;
显示所述生长趋势图。
一个实施例中,提供了一种卵泡跟踪系统,包括:
超声探头;
发射电路,用于激励所述超声探头向被测对象的卵巢组织发射超声波;
接收电路,用于激励所述超声探头接收所述超声波的回波,以获得所述超声波的回波信号;
处理器,用于执行如上述实施例中所述的卵泡跟踪方法。
根据本申请实施例的卵泡跟踪方法和系统,能够对单个目标卵泡进行跟踪监测,并最终获取单个目标卵泡的生长趋势图,操作者可以根据目标卵泡的生长趋势图准确评估最佳的取卵时间,有效地提升工作效率和准确性。
一个实施例中,提供了一种卵泡跟踪方法,包括:
获取被测对象的卵巢组织的第一超声图像,其中所述卵巢组织包括目标卵泡;
在所述第一超声图像中确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域;
获取所述被测对象的卵巢组织的第二超声图像;
基于所述第一卵泡区域,在所述第二超声图像中确定与所述目标卵泡对应的卵泡区域,获得第二卵泡区域。
一个实施例中,在所述第一超声图像中确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:
基于卵泡的图像特征,在所述第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域;或者,
检测操作者对所述第一超声图像中的目标卵泡的对应区域描迹的操作,以获得第一卵泡区域。
一个实施例中,所述超声图像为三维超声图像,所述基于卵泡的图像特征,在所述第 一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域,包括:
基于卵泡的图像特征,在所述第一超声图像的多个二维切面图像中对所述目标卵泡的对应区域进行分割;
综合所述多个所述二维切面图像上所述目标卵泡的对应区域,以得到所述第一卵泡区域。
一个实施例中,所述第一超声图像的多个二维切面图像为所述第一超声图像中的所有二维切面图像,或者,
所述第一超声图像的多个二维切面为对所述第一超声图像中以第一预设规则进行采样得到的采样图像,所述综合所述多个二维切面上的所述目标卵泡的对应区域包括:对所述采样图像的分割结果进行三维插值,以得到所述第一卵泡区域。
一个实施例中,所述超声图像为三维超声图像,所述基于卵泡的图像特征,在所述第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域,包括:
基于卵泡的图像特征,在所述第一超声图像中对所述目标卵泡对应的卵泡区域进行三维分割,以获得所述第一卵泡区域。
一个实施例中,在所述第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:
在所述第一超声图像中确定所述卵巢组织的区域,获得第一卵巢区域;
基于所述第一卵巢区域确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域。
一个实施例中,在所述第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:
在所述第一超声图像中确定多个第一候选卵泡区域;
获取每个第一候选卵泡区域的生长参数,在生长参数满足第一预设条件的第一候选卵泡区域确定为第一卵泡区域。
一个实施例中,在所述第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:
在所述第一超声图像中确定多个第一候选卵泡区域;
获取每个第一候选卵泡区域的生长参数,在生长参数为最大值的第一候选卵泡区域确定为第一卵泡区域。
一个实施例中,所述根据所述第一卵泡区域,在所述第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域,包括:
在所述第二超声图像中获取与所述第一卵泡区域的特征信息相似度满足第一阈值的卵泡区域,将获取的卵泡区域确定为第二卵泡区域;或者,
获取所述第二超声图像中的目标卵泡对应的第二候选卵泡区域,采用第一学习模型确定所述第一卵泡区域与所述第二候选卵泡区域的对应关系,并根据所述对应关系调整所述第二候选卵泡区域,并将调整后的所述第二候选卵泡区域确定为第二卵泡区域。
一个实施例中,所述根据所述第一卵泡区域,在所述第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域,包括:
检测操作者在所述第二超声图像中标识与所述第一卵泡区域相对应的卵泡区域的操 作,将标识的卵泡区域确定为第二卵泡区域。
一个实施例中,在上述实施例的基础上,还包括:
在所述第一超声图像中确定所述被测对象的卵巢组织的第一卵巢区域;
在所述第二超声图像中确定所述被测对象的卵巢组织的第二卵巢区域;
将所述第一卵巢区域与所述第二卵巢区域进行配准。
一个实施例中,将所述第一卵巢区域与所述第二卵巢区域进行配准,包括:
根据所述第一卵巢区域在所述第一超声图像中的方位,调整所述第二卵巢区域在所述第二超声图像中的方位,以使得所述调整后的第二卵巢区域在所述第二超声图像中的方位与所述第一卵巢区域在所述第一超声图像中的方位相同;或者,
根据所述第一卵巢区域的大小,调整所述第二卵巢区域相对于所述第一卵巢区域的大小,以使得所述调整后的第二卵巢区域与所述第一卵巢区域的大小相同。
一个实施例中,将所述第一卵巢区域与所述第二卵巢区域进行配准,包括:
采用第二学习模型确定所述第二卵巢区域与所述第一卵巢区域的对应关系,并根据所述对应关系调整所述第二卵巢区域;或者,
接收操作者根据所述第一卵巢区域调整所述待配准第二卵巢区域的指令,根据所述指令调整所述待配准第二卵巢区域。
一个实施例中,在上述实施例的基础上,还包括:
在所述第一超声图像中确定所述卵巢组织的第一卵巢区域;
将所述第二超声图像与所述第一卵巢区域的模板进行匹配,并根据匹配结果确定所述第二超声图像中所述卵巢组织的第二卵巢区域。
一个实施例中,在上述实施例的基础上,还包括:
接收操作者对所述第二超声图像和所述第一超声图像配准的指令,根据所述指令旋转、平移或缩放第二超声图像,以使得所述第二超声图像和所述第一超声图像配准。
一个实施例中,所述生长参数包括以下至少之一:体积、径长和生长速度。
一个实施例中,提供了一种卵泡跟踪系统,包括:
超声探头;
发射电路,用于激励所述超声探头向被测对象的卵巢组织发射超声波;
接收电路,用于激励所述超声探头接收所述超声波的回波,以获得所述超声波的回波信号;
处理器,用于执行如上述实施例中任一项所述的卵泡跟踪方法。
根据本申请实施例的卵泡跟踪方法和系统,能够在多个超声图像上对同一个目标卵泡进行跟踪监测,在临床上便能够实现对同一个目标卵泡生长发育情况,据此操作者可以准确评估最佳的取卵时间,有效地提升工作效率和准确性。
为了更清楚地说明本申请实施例中的技术方案,下面将对实施例描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本发明的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动性的前提下,还可以根据这些附 图获得其他的附图。
在附图中:
图1示出根据本申请实施例的超声成像系统的示意性框图;
图2示出根据本发明一实施例的卵泡跟踪方法的示意性流程图;
图3示出根据本发明另一实施例的卵泡跟踪方法的示意性图;
图4示出根据本发明另一实施例的卵泡跟踪方法的示意性图;
图5示出根据本发明另一实施例的卵泡跟踪方法的示意性图;
图6示出根据本发明另一实施例的卵泡跟踪方法的示意性图;
图7示出根据本发明另一实施例的卵泡跟踪方法的示意性图;
图8示出根据本发明另一实施例的卵泡跟踪方法的示意性图;
图9示出根据本发明另一实施例的卵泡跟踪方法的示意性图;
图10示出根据本发明另一实施例的卵泡跟踪方法的示意性图;
图11示出根据本发明另一实施例的卵泡跟踪方法的示意性图;
图12示出根据本发明又一实施例的卵泡跟踪方法的示意性流程图。
为了使得本申请的目的、技术方案和优点更为明显,下面将参照附图详细描述根据本申请的示例实施例。显然,所描述的实施例仅仅是本申请的一部分实施例,而不是本申请的全部实施例,应理解,本申请不受这里描述的示例实施例的限制。基于本申请中描述的本申请实施例,本领域技术人员在没有付出创造性劳动的情况下所得到的所有其它实施例都应落入本申请的保护范围之内。
在下文的描述中,给出了大量具体的细节以便提供对本申请更为彻底的理解。然而,对于本领域技术人员而言显而易见的是,本申请可以无需一个或多个这些细节而得以实施。在其他的例子中,为了避免与本申请发生混淆,对于本领域公知的一些技术特征未进行描述。
应当理解的是,本申请能够以不同形式实施,而不应当解释为局限于这里提出的实施例。相反地,提供这些实施例将使公开彻底和完全,并且将本申请的范围完全地传递给本领域技术人员。
在此使用的术语的目的仅在于描述具体实施例并且不作为本申请的限制。在此使用时,单数形式的“一”、“一个”和“所述/该”也意图包括复数形式,除非上下文清楚指出另外的方式。还应明白术语“组成”和/或“包括”,当在该说明书中使用时,确定所述特征、整数、步骤、操作、元件和/或部件的存在,但不排除一个或更多其它的特征、整数、步骤、操作、元件、部件和/或组的存在或添加。在此使用时,术语“和/或”包括相关所列项目的任何及所有组合。
为了彻底理解本申请,将在下列的描述中提出详细的结构,以便阐释本申请提出的技术方案。本申请的可选实施例详细描述如下,然而除了这些详细描述外,本申请还可以具有其他实施方式。
下面,首先参考图1描述根据本申请一个实施例的超声成像系统,图1示出了根 据本申请实施例的超声成像系统100的示意性结构框图。
如图1所示,超声成像系统100包括超声探头110、发射/接收电路112、处理器114以及显示器116。进一步地,超声成像系统100还可以包括波束合成电路和发射/接收选择开关等。
超声探头110通常包括多个阵元的阵列。在每次发射超声波时,超声探头110的所有阵元或者部分阵元参与超声波的发射。此时,这些参与超声波发射的阵元中的每个阵元或者每部分阵元分别受到发射脉冲的激励并分别发射超声波,这些阵元分别发射的超声波在传播过程中发生叠加,形成被发射到被测对象的卵巢组织所在区域的合成超声波束。
发射/接收电路112可以通过发射/接收选择开关与超声探头110连接。发射/接收选择开关也可以被称为发送/接收控制器,其可以包括发送控制器和接收控制器,发送控制器用于激励超声探头110经由发射电路向被测对象的卵巢组织所在区域发射超声波;接收控制器用于通过超声探头110经由接收电路接收从被测对象的卵巢组织所在区域返回的超声回波,从而获得超声回波数据。之后,发射/接收电路112将超声回波的电信号送入波束合成电路,波束合成电路对该电信号进行聚焦延时、加权和通道求和等处理,然后将处理后的超声回波数据送入处理器114。
可选地,处理器114可以通过软件、硬件、固件或其任意组合来实现,可以使用电路、单个或多个专用集成电路(Application Specific Integrated Circuit,ASIC)、单个或多个通用集成电路、单个或多个微处理器、单个或多个可编程逻辑器件、或者前述电路或器件的任意组合、或者其他适合的电路或器件,从而使得处理器114可以执行本说明书中的各个实施例中的方法的相应步骤。并且,处理器114可以控制所述超声成像系统100中的其它组件以执行期望的功能。
处理器114对其接收到的超声回波数据进行处理,得到被测对象的卵巢组织的超声数据,其中超声数据可以是二维超声数据,或者可以是三维超声数据。以超声数据为三维超声数据作为示例,超声探头110在一系列扫描平面内发射/接收超声波,由处理器114根据其三维空间关系进行整合,实现被测对象的卵巢组织在三维空间的扫描以及三维图像的重建。最后,由处理器114对其进行去噪、平滑、增强等部分或全部图像后处理步骤后,获取被测对象的卵巢组织的三维超声数据。处理器114可以获取被测对象的卵巢组织的至少三个不同检查时间的超声图像,其中卵巢组织包括目标卵泡。处理器114还可以分别在至少三个不同检查时间的超声图像上确定同一目标卵泡的对应区域。处理器114还可以基于目标卵泡对应的卵泡区域确定生长参数。处理器114根据目标卵泡的生长参数获得目标卵泡的生长趋势图。处理器114得到的趋势图可以存储于存储器中或在显示器116上显示。
显示器116与处理器114连接,显示器116可以为触摸显示屏、液晶显示屏等;或者显示器116可以为独立于超声成像系统100之外的液晶显示器、电视机等独立显示设备;或者显示器116可以是智能手机、平板电脑等电子设备的显示屏,等等。其中,显示器116的数量可以为一个或多个。例如,显示器116可以包括主屏和触摸屏,主屏主要用于显示超声图像,触摸屏主要用于人机交互。
显示器116可以显示处理器114得到的超声图像。此外,显示器116在显示超声图像的同时还可以提供给操作者进行人机交互的图形界面,在图形界面上设置一个或多个被控对象,提供给操作者利用人机交互装置输入操作指令来控制这些被控对象,从而执行相应的控制操作。例如,在图形界面上显示图标,利用人机交互装置可以对该图标进行操作,用来执行特定的功能。
可选地,超声成像系统100还可以包括显示器116之外的其他人机交互装置,其与处理器114连接,例如,处理器114可以通过外部输入/输出端口与人机交互装置连接,外部输入/输出端口可以是无线通信模块,也可以是有线通信模块,或者两者的组合。外部输入/输出端口也可基于USB、如CAN等总线协议、和/或有线网络协议等来实现。
其中,人机交互装置可以包括输入设备,用于检测操作者的输入信息,该输入信息例如可以是对超声波发射/接收时序的控制指令,可以是在超声图像上绘制出点、线或框等的操作输入指令,或者还可以包括其他指令类型。输入设备可以包括键盘、鼠标、滚轮、轨迹球、移动式输入设备(比如带触摸显示屏的移动设备、手机等等)、多功能旋钮等等其中之一或者多个的结合。人机交互装置还可以包括诸如打印机之类的输出设备。
超声成像系统100还可以包括存储器,用于存储处理器执行的指令、存储接收到的超声回波、存储超声图像,等等。存储器可以为闪存卡、固态存储器、硬盘等。其可以为易失性存储器和/或非易失性存储器,为可移除存储器和/或不可移除存储器等。
应理解,图1所示的卵泡跟踪系统100所包括的部件只是示意性的,其可以包括更多或更少的部件。本申请对此不限定。
下面,将参考图2描述根据本申请实施例的卵泡跟踪方法。图2是本申请实施例的卵泡跟踪方法200的一个示意性流程图。
本实施例是针对被测对象的同一个目标卵泡进行跟踪,获取同一目标卵泡在不同的检查时间中的超声图像,根据获取的超声图像确定同一目标卵泡的发育生长情况,参考图3所示,操作者能够根据同一卵泡生长发育情况判断取卵时间。相对于现有针对整体卵泡发育情况进行跟踪监测,本实施中针对单个卵泡的跟踪监测将会提供更加准确的临床信息,操作者据此可以评估更为准确的取卵时间。
参考图2所示,本申请一个实施例的卵泡跟踪方法200包括如下步骤:
步骤201,获取被测对象的卵巢组织的至少三个不同检查时间的超声图像,所述卵巢组织包括目标卵泡。
一个实施例中,针对每次检查获取的多个超声图像,处理器可以自动从多个超声图像中获取包括目标卵泡的超声图像,以作为当前检查时间获取的目标卵泡跟踪监测的超声图像;或者从获取的多个超声图像中操作者手动选择包括目标卵泡的超声图像,以作为当前检查时间的目标卵泡跟踪监测的超声图像。同样的方法,可以获取多次不同检查时间同一个受测者的卵巢组织的超声图像,其中卵巢组织包括目标卵泡。
其中超声图像可以为二维超声图像,或者也可以为三维超声图像。在实际临床中, 受测者在注射完促排卵药物到取卵的过程中,通常会做四到六次的超声检查,最少不少于三次超声检查,对同一个目标卵泡获取至少三次生长参数,基于至少三次生长参数绘制的生长趋势图将会提供操作者更加准确的卵泡生长趋势信息,操作者以此评估取卵时间也会更加精准。
步骤202,分别在所述至少三个不同检查时间的超声图像上确定所述目标卵泡对应的卵泡区域,获得至少三个卵泡区域,其中所述至少三个卵区域为同一个目标卵泡对应的卵泡区域。
针对同一个目标卵泡,处理器在至少三个不同检查时间的超声图像上自动确定目标卵泡对应的区域;或者操作者手动在至少三个不同检查时间的超声图像上确定目标卵泡对应区域。以此获得同一个目标卵泡在不同时间获取的超声图像的生长参数,即能够针对同样的目标卵泡进行跟踪,以此评估准确的取卵时间。
在不同的超声图像上确定同一个目标卵泡对应的区域,该过程可以处理器自动确定,或者也可以操作者手动确定,或者处理器自动和操作者手动结合操作进行确认。
在不同的超声图像上确定同一个目标卵泡对应的区域过程中,首先需要确定一个超声图像中目标卵泡对应的区域,然后在其它不同检查时间获取的多个超声图像中确定同一个目标卵泡区域对应的卵泡区域。为了更好的对该过程进行说明,在此所述至少三个不同检查时间的超声图像分别为第一超声图像、第二超声图像和第三超声图像,其中第一超声图像、第二超声图像和第三超声图像仅是为了区分是至少三个不同检查时间获取的超声图像,其中的“第一”、“第二”和“第三”并不表示检查时间的先后顺序。
首先,在所述至少三个不同检查时间的超声图像中的第一超声图像中确定目标卵泡对应的卵泡区域,获得第一卵泡区域。
一个实施例中,目标卵泡可以是一个目标卵泡,或者可以是两个或者更多的目标卵泡。目标卵泡在其生长过程中,可能一直生长最后成为卵子被取出,或者可能在未形成卵子之前便消失了。当然在整个跟踪目标卵泡的过程中,目标卵泡可能会消失,也可能会增加新的目标卵泡。一般临床上会在开始多选几个目标卵泡,防止在后续跟踪过程中会损失目标卵泡。
其中目标卵泡在第一次选取的过程中,一般来说会选取生长较好的卵泡作为目标卵泡。例如如果获得的超声图像是二维超声图像,可以选择二维超声图像上径长较大的卵泡作为目标卵泡;或者如果获得的超声图像是三维超声图像,可以是体积较大的卵泡作为目标卵泡。其中,操作者可以手动选择目标卵泡,或者处理器自动识别目标卵泡。
一个实施例中,基于第一超声图像确定多个第一候选卵泡区域,基于多个第一候选卵泡区域确定多个第一候选卵泡区域的生长参数,并将生长参数满足第一预设条件的第一候选卵泡区域确定为目标卵泡的第一区域。其中生长参数为卵泡的体积或径长等。以体积为例,第一预设条件可以为卵泡的体积大于预设阈值,或者卵泡的体积在第一候选卵泡体积中属于相对较大等。第一预设条件为可处理器自动设定,或者操作者手动设定。
一个实施例中,也可以将生长参数最大的第一候选卵泡区域确定为目标卵泡的第 一区域。其中可以将体积最大的第一候选卵泡区域确定为目标卵泡的第一区域,或者将径长最大的第一候选卵泡区域确定为目标卵泡区域。
基于卵巢组织的第一超声图像确定目标卵泡的第一区域,以下将通过卵巢组织为三维超声图像进行举例说明。
一个实施例中,处理器基于卵泡的图像特征,在所述至少三个不同检查时间的超声图像中的第一超声图像上对目标卵泡对应的卵泡区域进行分割,以获取第一卵泡区域;或者处理器检测操作者对第一超声图像的目标卵泡的对应区域描迹的指令,以获得第一卵泡区域。
其中处理器基于卵泡的图像特征在第一超声图像上自动分割目标卵泡的对应区域,大体分为两种方法,一种是二维图像分割法,即将第一超声图像切分为多个二维切面,从多个二维切面中分割出目标卵泡的对应区域,综合多个二维切面中目标卵泡的对应区域,以得到第一卵泡区域。
在上述实施例的基础上,其中第一超声图像的多个二维切面图像可以为所述第一超声图像中的所有二维切面图像。
在上述实施例的基础上,所述第一超声图像的多个二维切面可以为对所述第一超声图像中以预设规则进行采以得到的采样图像。例如选取一个点,围绕该点呈发射状切取采样图像,参考图4所示;或者平行切取采样图像,参考图5所示。对所述采样图像的分割结果进行三维插值,以得到所述第一卵泡区域。
二维图像分割法主要有两种方式,一种是传统的分割算法,不需要通过大量数据进行建模;一种是基于机器学习的图像分割法,需要大量数据建立学习模型。
传统的图像分割传统的图像分割算法是通过目标检测方法(如点检测、线检测)检测目标路卵泡的对应区域,从而对该区域进行分割,获得第一卵泡区域。常用的分割算法有基于水平集(Level Set)分割算法,随机游走(Random Walk),图割(Graph Cut),Snake等。
基于机器学习的方法通过学习数据库中的目标卵泡的对应区域的特点来对目标卵泡进行分割。其主要步骤为:1)构建数据库,数据库中应包含了大量的数据集及其对应的标记结果,标记信息是对目标卵泡进行了精确分割的掩膜(Mask)信息,其中掩膜信息包括边界信息和边界信息形成的区域信息。2)分割步骤,分割算法主要有两种,一种是基于传统机器学习的分割算法,一种是基于深度学习的分割算法,具体如下:
基于传统机器学习的语义分割算法,常见的步骤为将第一超声图像分为很多个图像块,然后对图像块进行特征的提取,特征的提取方式有传统的PCA、LDA、Harr特征、纹理特征等,也可以使用深度神经网络,如Overfeat网络等来进行特征的提取,然后对提取的特征使用级联的分类器进行分类,其中分类器如KNN、SVM、随机森林等判别器,从而确定当前图像块是否为目标卵泡的对应区域,将该分类结果作为当前图像块的中心点标记结果,最后得到整个图像的分割结果,即获得第一卵泡区域。
基于深度学习的端到端的语义分割算法,其结构就是通过卷积层、池化层、上采样或者反卷积层等的堆叠,从而得到一个和输入的第一超声图像尺寸一致的输出图像,该输出图像直接分割出需要的目标卵泡的对应区域,该方法是一个监督学习,所以输 入的监督信息是目标卵泡的对应区域的掩膜(Mask)信息,数据准备比较耗时,常见的二维网络有FCN、U-Net、Mask R-CNN等。
除了上述一种是二维图像分割法之外,另一种是三维图像分割法,即处理器基于卵泡的图像特征,在第一超声图像中直接进行三维分割,从而确定第一卵泡区域,参考图6所示。
基于三维数据的分割可以通过目标检测方法(如点检测、线检测)检测目标卵泡的对应区域,从而对该区域进行分割,获得第一卵泡区域。该方法不需要大量的标记数据,常用的分割算法有基于水平集(Level Set)分割算法,随机游走(Random Walk),图割(Graph Cut),Snake,三维大律法,聚类,马尔可夫随机场等。
也可以利用机器学习方法通过学习数据库中目标卵泡的对应区域的特征来对该区域进行分割。其主要步骤为:1)构建数据库,数据库中应包含了大量的数据集及其对应的标记结果,标记信息是对目标卵泡的对应区域进行了精确分割的掩膜(MASK)信息。2)分割步骤,分割算法主要有两种,一种是基于传统机器学习的分割算法,一种是基于深度学习的分割算法,具体如下:
基于传统机器学习的语义分割算法,常见的步骤为将第一超声图像分为很多个图像块,然后对图像块进行特征的提取,如基于sobel算子的边缘检测提取三维数据边缘特征、梯度,纹理特征等,或者是利用神经网络,如Medicalnet进行特征的提取。然后对提取的特征使用级联的分类器进行分类,其中分类器如如KNN、SVM、随机森林等判别器,从而确定当前图像块是否为目标卵泡的对应区域,将该分类结果作为当前图像块的中心点标记结果,最后得到整个图像的分割结果,即获得第一卵泡区域。
基于深度学习的端到端语义分割算法,其结构就是通过三维下的卷积层、池化层、上采样或者反卷积层等的堆叠,从而得到一个和输入第一超声图像尺寸一致的输出图像,该输出图像直接分割出需要的第一卵泡区域,该方法是一个监督学习,所以输入的监督信息是目标区域的掩膜信息,数据准备比较耗时,常见三维分割网络有3D U-Net、3D FCN、Medical-Net等。
当第一超声图像为二维超声图像时,在二维图像上获取目标卵泡的第一卵泡区域分割法主要有两种方式,一种是传统的分割算法,不需要通过大量数据进行建模;一种是基于机器学习的图像分割法,需要大量数据建立学习模型。传统的分割算法和基于机器学习的图像分割算法如前所述,在此不再赘述。
基于上述实施例中获得的第一卵泡区域,在至少三个不同检查时间的超声图像中的第二超声图像中确定目标卵泡对应的卵泡区域,获得第二卵泡区域;根据第一卵泡区域或第二卵泡区域,在所述至少三个不同检查时间的超声图像中的第三超声图像中确定所述目标卵泡对应的卵泡区域,获得第三卵泡区域。其中以基于第一卵泡区域在第二超声图像中获得第二卵泡区域为例进行过程说明。
一个实施例中,处理器可以根据第一卵泡区域自动在第二超声图像中确定与目标卵泡相对应的卵泡区域,获得第二卵泡区域;或者操作者手动在第二超声图像中标识与所述第一卵泡区域相对应的卵泡区域,处理器检测到操作者的手动标识操作,自动将标识的卵泡区域确定为第二卵泡区域,其中操作者手动标识包括操作者手动确定第 二卵泡区域,比如对第二卵泡区域进行描迹,或者还可以包括突出显示确定的第二卵泡区域。
其中处理器根据第一卵泡区域在自动在第二超声图像中确定第二卵泡区域主要包括两种方法,一种是传统的图像匹配方法,即基于第一卵泡区域的特征信息在第二超声图像中搜索与第一卵泡区域的特征信息相似的子图像,需要迭代以寻求最优解,速度较慢,不需要大量数据。
一个实施例中,处理器从第二超声图像中自动确定与第一卵泡区域的特征信息相似度满足第一阈值的第二区域,将满足第一阈值的第二区域确定为第二卵泡区域。其中,第一阈值可以百分数比,比如90%。处理器可以自动设定第一阈值;或者也可以操作者根据临床需求手动设定第一阈值。
一个实施例中,目标卵泡的特征信息可以是通过某些变换方式,将图像信号从时域变换为频域得到的图像信号信息(如傅里叶变换,将图像反变换到频域上,域变换);也可以是直接来自于图像的原始数据信息,原始数据信息是指直接从图像中获取的未经任何处理的数据信息,如灰度值,对比度等;或者可以是经过原始数据信息经处理后提取的图像特征信息,图像特征主要可分为点、线、区域等特征,也可以分为局部特征和全局特征,其中以点特征和线特征应用比较多,点特征主要包括Harris、Moravec、KLT、Harr-like、HOG、LBP、SIFT、SURF、BRIEF、SUSAN、FAST、CENSUS、FREAK、BRISK、ORB、光流法和A-KAZE等,线特征主要包括LoG算子、Robert算子、Sobel算子、Prewitt算子、Canny算子等。
基于上述特征信息的不同,传统的图像匹配方法具体又分为与特征信息相对应的三种类型:1)基于域变换的方法,该方法是基于第一卵泡区域的频域信号进行匹配,主要采用相位相关、沃尔什变换、小波变换等方式。2)基于模板匹配的方法,该方法是基于第一卵泡区域的原始数据信息(如灰度值、对比度等)进行匹配,根据第一卵泡区域的原始数据信息到第二超声图像中搜索与第一卵泡区域相似相似度满足第一阈值的子图像,将所述满足第一阈值的子图像确定为第二卵泡区域。在这个过程中是不需要提取超声图像的特征信息。例如,可以基于灰度进行匹配,基于灰度的匹配算法也称为相关匹配算法,用空间二维滑动模板进行匹配。基于模板匹配的方法中常用的算法有平均绝对差算法(MAD)、绝对误差和算法(SAD)、误差平方和算法(SSD)、平均误差平方和算法(MSD)、归一化积相关算法(NCC)、序贯相似性检测算法(SSDA)、hadamard变换算法(SATD)、局部灰度值编码算法和PIU等。3)基于图像特征的匹配方法,该方法首先提取第一卵泡区域的图像特征信息,再根据提取的图像特征信息生成特征描述算子,最后在第二超声图像中确定跟描述算子的相似度满足第一阈值的第二卵泡区域。
基于第一卵泡区域在第二超声图像中确定第二卵泡区域的方法除了传统的图像匹配方法之外,还有另一种方式即基于深度学习的图像配准方式,参考图7所示。基基于深度学习的图像配准方式是指通过已经训练好的学习模型确定两个图像区域之间的对应关系,通过旋转、平移和缩放三个操作中至少一个操作使得两个图像区域在空间位置关系上达到方位和大小的对齐。在该过程中需要大量的标注样本来建立数据库, 可以一步到位实现图像的配准。
处理器自动或操作者手动获取第二超声图像中目标卵泡对应的第二候选卵泡区域,处理器采用第一学习模型确定第一卵泡区域与第二候选卵泡区域的对应关系,并根据所述对应关系调整第一候选卵泡区域,并将调整后的第二候选卵泡区域确定为第二卵泡区域。主要步骤为:1)构建数据库,数据库中包含大量的数据集及其对应的标记信息,标记信息是对齐的对应关系,比如可以是旋转关系,或者可以是平移关系,或者可以缩放关系,或者是以上三种关系中任意组合关系;2)配准步骤,将第一卵泡区域与第二候选卵泡区域输入至数据库中,判断出对应关系,根据判断出的关系调整目标卵泡的第二候选卵泡区域,其中调整可以是通过旋转、平移或缩放中的至少一种操作方式,输出调整后的第二候选卵泡区域确定为第二卵泡区域。
基于深度学习的图像配准方法又包括两种类型,一种是two-stage的图像配准方式;一种是端到端的图像配准方式。
基于深度学习的two-stage配准方式,该种方式类似于基于特征的匹配方法,但是此处将SIFT等传统特征置换为利用深度学习网络进行特征的提取。
基于深度学习的端到端的图像配准方法,可以分为两种类型,一种是监督性学习,将第一卵泡区域和第二候选卵泡区域输入都已训练好第一学习模型中,通常通过卷积层、池化层、上采样或者反卷积层等三维网络结构的堆叠算法,从而输出第一卵泡区域和第二候选卵泡区域之间的对应关系,输入的监督信息是对应关系,常见的网络结构有FCN、U-Net等。一种是无监督学习,即输入为第一卵泡区域和第二候选卵泡区域,输出为已调整的第二候选卵泡区域。
通过上述同样的方法,依次获取同一目标卵泡在其它不同检查时间的超声图像中的对应区域,完成同一目标卵泡在不同检查时间获取的超声图像中持续跟踪监测。
在上述实施例的基础上,还可以先对不同超声图像中的卵巢组织区域进行配准或者匹配,或者也可以对整体图像进行配准,在此基础上再进行目标卵泡的跟踪监测。由于在对同一个目标卵泡跟踪检测的过程中,目标卵泡生长速度较快,也正是因为目标卵泡的大小是处于不断变化中的,所以同一个目标卵泡在不同检查时间获取的超声图像上大小是会有差别的,比如目标卵泡的径长或者体积会有差别,也正是因为此,所以针对同一个目标卵泡进行持续跟踪监测会存在难度。但是因为检查间隔时间比较短,同一个目标卵泡在卵巢组织中的相对位置不会有太大的变化,所以可以先在第二超声图像中确定卵巢组织区域,再根据第一超声图像中第一目标卵泡和第一超声图像的相对位置关系在第二超声图像上确定第二卵泡区域,这样可以减小卵巢外与目标卵泡相似结构组织的影响,还可以减少处理器的计算量,使得结果更加准确并且处理效率更高。
当然如果第一超声图像和第二超图像是相同部位或大致相同部位所获取图像,也可以先通过旋转、平移和缩放中至少之一的操作对第一超声图像和第二超声图像整体进行调整,使两者配准之后在对目标卵泡进行跟踪监测。这种方式对操作者前后检查时间的手法操作要求比较高。
值得说明的是,通常在二维超声图像中,由于操作者手法或其它因素导致获取的 超声切面可能不能包括整个卵巢组织,这样就给目标卵泡的跟踪监测带来了不便,甚至有可能在获取的超声图像中并没有包括目标卵泡,所以通过二维超声图像进行同一个目标卵泡的跟踪监测对操作者手法要求比较高。而如果通过三维超声的方法进行卵泡跟踪,操作者比较容易就能获取整个卵巢组织的图像,这样再进行目标卵泡的跟踪就会比较容易获取,也能够降低对操作者的手法要求。
需要说明的是,其中卵巢组织区域的配准或匹配或整体图像的配准并不是一个必要的步骤,可以采用也可以不采用。
一个实施例中,在第一超声图像中确定被测对象的卵巢组织的第一卵巢区域,在第二超声图像中确定被测对象的卵巢组织的第二卵巢区域,将第一卵巢区域与第二卵巢区域进行配准。其中处理器可以对第一卵巢区域和第二卵巢区域进行自动配准,或者操作者手动进行配准。
一个实施例中,处理器在第一超声图像中采用基于目标分割的机器学习方法或者传统的分割算法确定第一卵巢区域,或者操作者手动在确定第一卵巢区域。
一个实施例中,根据第一卵巢区域在第一超声图像中的方位,调整第二卵巢区域在第二超声图像中的方位,以使得调整后的第二卵巢区域在第二超声图像中的方位与第一卵巢区域在第一超声图像中的方位相同;或者根据第一卵巢区域的大小,可以切面上径长的大小,或者是体积的大小,调整第二卵巢区域相对于第一卵巢区域的大小,以使得所述调整后的第二卵巢区域与所述第一卵巢区域的大小相同。其中方位的调整和大小的调整可以处理器自动调整,也可以操作者手动调整。
由于两次采集时间存在间隔,医生手法、卵泡生长导致卵巢形状微变等因素造成两次卵巢数据没有对齐,比如是第一超声图像中第一卵巢区域的位置在第一超声图像中偏左的位置,而第二卵巢区域在第二超声图像中的位置偏右或者居中,这时就需要对第二卵巢区域的方位进行调整,使得第二卵巢区域在第二超声图像中的方位与第一卵巢区域在第一超声图像中的方位相同。或者可能存在第一卵巢区域比较大,而第二卵巢区域比较小的情况,此处的大小可以为体积的大小,或者可以为径的大小,这时就需要对第二卵巢区域进行放大处理,以使得调整后的第二卵巢区域与第一卵巢区域的大小相同。在本实施例中方位的相同和大小的相同包括了绝对的方位相同和绝对的大小相同,还包括大致的方位相同和大致的大小相同。特别是操作者手动进行调整时,不可避免的存在一定的误差无法做绝对的相同,但也因为卵巢组织区域的配准并不是一个必不可少的步骤而仅仅是一个优化步骤,所以卵巢组织区域能够达到大致配准基本上也能满足临床需求。
一个实施例中,处理器可以采用机器学习方法自动调整第二卵巢区域,使得调整后的第二卵巢区域和第一卵巢区域配准,参考图8所示。
一个实施例中,处理器采用第二学习模型确定第一卵巢区域和第二卵巢区域的对应关系,比如旋转关系,或者平移关系,或者缩放关系,或者以上三种关系的任意组合关系,并根据确定的对应关系调整第二卵巢区域,使得第二卵泡区域与第一卵泡区域配准。主要步骤为:1)构建数据库,数据库中包含大量的数据集及其对应的标记信息,标记信息是对应关系;2)配准步骤,将第一卵巢区域与第二卵巢区域输入至数据 库中,判断出对应第一卵巢区域和第二卵巢区域的对应关系,根据判断出的对应关系调整第二卵巢区域,输出调整后的第二卵巢区域。
机器学习方法中又包括深度学习方法,基于深度学习的卵巢组织的配准与前述中基于深度学习的目标卵泡的配准方法类似,在此不再赘述。
上述实施例中,需要在第二超声图像中先确定第二卵巢区域,然后基于获取的第一卵巢区域和第二卵巢区域进行配准。除此之外,还可以采用不需要先确定第二卵巢区域的方式进行卵巢区域的配准,例如传统的图像匹配方法。传统的图像匹配方法是基于第一卵巢区域的特征信息在第二超声图像中搜索与其相似度满足预设阈值子图像,将相似度满足预设阈值的子图像确定为第二卵泡区域。
其中传统的图像配准方法包括:基于域变换的方法;基于模板匹配的方法;基于图像特征的匹配方法。利用传统的图像匹配方法确定第一卵巢区域与前述中利用传统的图像匹配方法确定第一卵泡区域类似,在此不再赘述。
一个实施例中,操作者可以手动基于图像整体配准的方法实现卵巢区域的配准。接收操作者对第二超声图像和第一超声图像配准的指令,根据所述指令旋转、平移或缩放第二超声图像,以使得第二超声图像和第一超声图像配准。
在对同一个目标卵泡进行跟踪监测的过程中,会有新增卵泡的情况出现,或者目标卵泡消失的情况。对于在后一次超声图像中找到与前一次超声图像中目标卵泡相对应的卵泡,进行保留,记录其相关位置已经对应的体积径长等信息;对于在前一次超声图像中出现,在后一次超声图像中消失的目标卵泡,则进行删除;对于在后一次超声图像中新出现的目标卵泡,如果新增卵泡的生长参数满足预设条件,则将生长参数满足预设条件的新增卵泡作为新的目标卵泡加入追踪行列,否则删除。其中预设条件为体积满足一定的阈值条件或者径长满足一定的阈值条件,而这个具体的阈值大小可以由处理器自动设定,或者操作者手动设定。
步骤203,根据所述至少三个卵泡区域分别确定所述目标卵泡的生长参数,获得所述目标卵泡的至少三个生长参数。
其中生长参数包括以下至少之一:体积、径长和生长速度。
一个实施例中,根据至少三个卵泡区域处理器可以自动确定或者操作者可以手动确定目标卵泡的生长参数,获得目标卵泡的至少三个生长参数,参考图9所示为手动在二维图像中获取目标卵泡的长径和短径。
示例性地,获取的超声图像为三维超声图像。目标卵泡的体积可以基于分割结果的体素以及体素之间的距离进行测量;目标卵泡的径长则可以先将目标卵泡区域分割结果拟合成一个椭球,从椭球中确定出目标卵泡区域面积最大或目标卵泡区域面积满足预设要求的切面,其中预设要求可以用户自定义,也可以机器设定,在确定的切面上确定长度最大的径长为目标卵泡的长径,在确定的切面上确定与长径垂直的最长的短径为目标卵泡的短径,其中垂直关系包括绝对垂直和大致垂直,例如在临床上处于85度到95度之间基本上都可以认为是垂直关系。
示例性地,获取的超声图像为二维超声图像。在获得的目标卵泡的二维图像中根据目标卵泡区域拟合成一个椭圆,基于椭圆从而确定目标卵泡的长径和短径。在基于 目标卵泡的二维图像的垂直方向获取多帧切面图像,从获取的多帧切面图像中手动或自动选取目标卵泡径长最大或径长满足预设要求的切面图像,其中预设要求可以用户自定义,也可以机器设定;基于选取的切面图像中确定目标卵泡的高长;基于确定的高长,长径和短径计算目标卵泡的体积。
在上述实施例的基础上,目标卵泡的生长速度则可以基于获取的体积或径长与生长时间的比值确定。
步骤204,根据所述目标卵泡的至少三个生长参数获得所述目标卵泡的生长趋势图。
步骤205,显示生长趋势图。
一个实施例中,目标卵泡的生长趋势图可以为生长参数曲线图,其中生长参数曲线图以检查时间为第一坐标,以生长参数为第二坐标,例如参考图10所示。
一个实施例中,目标卵泡的生长趋势图可以为不同检查时间对应的生长参数的列表,参考图11所示。
基于上面的描述,根据本申请实施例的卵泡跟踪方法和系统,能够对单个目标卵泡进行跟踪监测,并最终获取单个目标卵泡的生长趋势图,操作者可以根据目标卵泡的生长趋势图准确评估最佳的取卵时间,有效地提升工作效率和准确性。
一个实施例中,提供了一种卵泡跟踪方法。下面,将参考图12描述根据本申请实施例的卵泡跟踪方法。图12是本申请实施例的卵泡跟踪方法300的一个示意性流程图。
如图12所示,本申请一个实施例的卵泡跟踪方法300包括如下步骤:
步骤301,获取被测对象的卵巢组织的第一超声图像,其中所述卵巢组织包括目标卵泡。
处理器自动或者操作者手动获取被测对象的卵巢组织的第一超声图像,其中卵巢组织包括目标卵泡。其中超声图像可以为二维超声图像,或者也可以为三维超声图像。在实际临床中,一般被测对象在注射完促排卵药物到取卵的过程中,通常会对被测者做多次的超声检查。基于临床需求,一般来说至少会做两次超声检查,基于获取至少两次不同检查时间的超声图像跟踪同一目标卵泡。当然,如果获取三次到六次不同检查时间的超声图像对同一目标卵泡进行跟踪,获取的取卵时间会更加准确。
步骤302,在所述第一超声图像中确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域。
一个实施例中,目标卵泡可以是一个目标卵泡,或者可以是两个或者更多的目标卵泡。目标卵泡在其生长过程中,可能一直生长最后成为卵子被取出,或者可能在未形成卵子过程中消失。当然在整个跟踪目标卵泡的过程中,目标卵泡可能会消失,也可能会增加新的目标卵泡。一般临床上会在开始多选几个目标卵泡,防止在后续跟踪过程中会损失目标卵泡。
其中目标卵泡在第一次选取的过程中,一般来说会选取生长较好的卵泡作为目标卵泡。其中生长较好的卵泡可以是切面上径长较大的卵泡,或者可以是体积较大的卵泡,或者还可以是径长将长的卵泡。医生可以手动选择或者处理器自动识别出生长较 好的卵泡,并将其列为目标卵泡。
一个实施例中,基于第一超声图像确定多个第一候选卵泡区域,基于多个第一候选卵泡区域确定多个第一候选卵泡区域的生长参数,并将生长参数满足第一预设条件的第一候选卵泡区域确定为目标卵泡的第一区域。其中生长参数为卵泡的体积或径长等。以体积为例,第一预设条件可以为卵泡的体积大于一定的阈值,或者卵泡的体积在第一候选卵泡体积中属于相对较大等。第一预设条件为可处理器自动设定,或者操作者手动设定。
一个实施例中,也可以将生长参数最大的第一候选卵泡区域确定为目标卵泡的第一区域。其中可以将体积最大的第一候选卵泡区域确定为目标卵泡的第一区域,或者将径长最大的第一候选卵泡区域确定为目标卵泡区域。
一个实施例中,卵巢组织的第一超声图像可以为二维超声图像,或者也可以为三维超声图像。
处理器可以自动对第一超声图像上对目标卵泡对应的卵泡区域进行分割以获取第一卵泡区域;或者处理器检测操作者对第一超声图像的目标卵泡的对应区域描迹的指令,以获得第一卵泡区域。
其中处理器基于卵泡的图像特征在第一超声图像上自动分割目标卵泡大体分为两种方法,一种是二维图像分割法,即将第一超声图像切分为多个二维切面,从多个二维切面中分割出目标卵泡对应的区域,综合多个二维切面中目标卵泡的对应区域,以得到第一卵泡区域。
其中第一超声图像的多个二维切面图像为所述第一超声图像中的所有二维切面图像;或者所述第一超声图像的多个二维切面为对所述第一超声图像中以预设规则进行采以得到的采样图像,对所述采样图像的分割结果进行三维插值,以得到所述第一卵泡区域。
二维图像分割法主要有两种方式,一种是传统的分割算法,不需要通过大量数据进行建模;一种是基于机器学习的图像分割法,需要大量数据建立学习模型。二维图像分割法如前所述,在此不再赘述。
另一种是三维图像分割法,处理器基于卵泡的图像特征,在第一超声图像中直接进行三维分割,从而确定第一卵泡区域的边界。
基于三维数据的分割可以通过目标检测方法(如点检测、线检测)检测目标卵泡的对应区域,从而对该区域进行分割,该方法不需要大量的标记数据,常用的分割算法有基于水平集(Level Set)分割算法,随机游走(Random Walk),图割(Graph Cut),Snake,三维大律法,聚类,马尔可夫随机场等。
也可以利用机器学习方法通过学习数据库中目标卵泡的对应区域的特征来对该区域进行分割。其主要步骤为:1)构建数据库,数据库中应包含了大量的数据集及其对应的标记结果,标记信息是对目标进行了精确分割的边界信息。2)分割步骤,分割算法主要有两种,一种是基于传统机器学习的分割算法,一种是基于深度学习的分割算法。基于机器学习的分割方法如前所述,在此不再赘述。
步骤303,获取所述被测对象的卵巢组织的第二超声图像。
处理器自动获取被测对象的卵巢组织的第二超声图像,或者操作者手动确定被测对象的卵巢组织的第二超声图像。
步骤304,基于所述第一卵泡区域,在所述第二超声图像中确定与所述目标卵泡对应的卵泡区域,获得第二卵泡区域。
一个实施例中,处理器可以根据第一卵泡区域自动在第二超声图像中确定与目标卵泡相对应的卵泡区域,获得第二卵泡区域;或者操作者手动在第二超声图像中标识与所述第一卵泡区域相对应的卵泡区域,处理器检测到操作者的手动标识操作,自动将标识的卵泡区域确定为第二卵泡区域。
其中处理器根据第一卵泡区域在自动在第二超声图像中确定第二卵泡区域主要包括两种方法,一种是传统的图像匹配方法,即基于第一卵泡区域的特征信息在第二超声图像中搜索与第一卵泡区域的特征信息相似的子图像,需要迭代以寻求最优解,速度较慢,不需要大量数据。
一个实施例中,处理器从第二超声图像中自动确定与第一卵泡区域的特征信息相似度信息满足第一阈值的第二区域,将满足第一阈值的第二区域确定为第二卵泡区域。其中,第一阈值可以百分数比,比如90%。处理器可以自动设定第一阈值;或者也可以操作者根据临床需求手动设定第一阈值。
一个实施例中,目标卵泡的特征信息可以是通过某些变换方式,将图像信号从时域变换为频域得到的图像信号信息(如傅里叶变换,将图像反变换到频域上,域变换);也可以是直接来自于图像的原始数据信息,原始数据信息是指直接从图像中获取的未经任何处理的数据信息,如灰度值,对比度等;或者可以是经过原始数据信息提取的图像特征信息,图像特征主要可分为点、线、区域等特征,也可以分为局部特征和全局特征,其中以点特征和线特征应用比较多,点特征主要包括Harris、Moravec、KLT、Harr-like、HOG、LBP、SIFT、SURF、BRIEF、SUSAN、FAST、CENSUS、FREAK、BRISK、ORB、光流法和A-KAZE等,线特征主要包括LoG算子、Robert算子、Sobel算子、Prewitt算子、Canny算子等。
基于上述特征信息的不同,传统的图像匹配方法具体又分为与特征信息相对应的三种类型:1)基于域变换的方法,该方法是基于第一卵泡区域的频域信息进行匹配,主要采用相位相关、沃尔什变换、小波变换等方式。2)基于模板匹配的方法,该方法是基于第一卵泡区域的原始数据信息进行匹配,如灰度值等,根据第一卵泡区域的原始数据信息到第二超声图像中搜索与第一卵泡区域相似相似度满足第一阈值的子图像,将所述满足第一阈值的子图像确定为第二卵泡区域。在这个过程中是不需要提取超声图像的特征信息。3)基于图像特征的匹配方法,该方法首先提取第一卵泡区域的图像特征信息,再根据提取的图像特征信息生成特征描述算子,最后在第二超声图像中确定跟描述算子的相似度满足第一阈值的第二区域。
基于第一卵泡区域在第二超声图像中确定第二卵泡区域的方法除了传统的图像匹配方法之外,还有另一种方式即基于深度学习的图像配准方式。基于深度学习的图像配准方式,需要大量的标注样本来建立数据库,可以一步到位实现图像的配准。
处理器自动或操作者手动获取第二超声图像中目标卵泡对应的第二候选卵泡区 域,处理器采用第一学习模型确定第一卵泡区域与第二候选卵泡区域的对应关系,并根据所述对应关系调整第一候选卵泡区域,并将调整后的第二候选卵泡区域确定为第二卵泡区域。主要步骤为:1)构建数据库,数据库中包含大量的数据集及其对应的标记信息,标记信息是对齐的对应关系,比如可以是旋转关系,或者可以是平移关系,或者可以缩放关系,或者是以上三种关系中任意组合关系;2)配准步骤,将第一卵泡区域与第二候选卵泡区域输入至数据库中,判断出对应关系,根据判断出的关系调整目标卵泡的第二候选卵泡区域,其中调整可以是通过旋转、平移或缩放中的至少一种操作方式,输出调整后的第二候选卵泡区域确定为第二卵泡区域。
基于深度学习的图像配准方法又包括两种类型,一种是two-stage的图像配准方式;一种是端到端的图像配准方式。具体的基于深度学习的图像配准方法如前所述,在此不再赘述。
通过上述同样的方法,依次获取同一目标卵泡在其它不同检查时间的超声图像中的对应区域,完成同一目标卵泡在不同检查时间获取的超声图像中持续跟踪监测。
在上述实施例的基础上,还可以先对不同超声图像中的卵巢组织区域进行配准或者匹配,或者也可以对整体图像进行配准,在此基础上再进行目标卵泡的跟踪监测。
一个实施例中,在第一超声图像中确定被测对象的卵巢组织的第一卵巢区域,在第二超声图像中确定被测对象的卵巢组织的第二卵巢区域,将第一卵巢区域与第二卵巢区域进行配准。其中处理器可以对第一卵巢区域和第二卵巢区域进行自动配准,或者操作者手动进行配准。
一个实施例中,处理器在第一超声图像中采用基于目标分割的机器学习方法或者传统的分割算法确定第一卵巢区域,或者操作者手动在确定第一卵巢区域。
一个实施例中,根据第一卵巢区域在第一超声图像中的方位,调整第二卵巢区域在第二超声图像中的方位,以使得调整后的第二卵巢区域在第二超声图像中的方位与第一卵巢区域在第一超声图像中的方位相同;或者根据第一卵巢区域的大小,可以切面上径长的大小,或者是体积的大小,调整第二卵巢区域相对于第一卵巢区域的大小,以使得所述调整后的第二卵巢区域与所述第一卵巢区域的大小相同。其中方位的调整和大小的调整可以处理器自动调整,也可以操作者手动调整。
一个实施例中,处理器可以采用机器学习方法自动调整第二卵巢区域,使得调整后的第二卵巢区域和第一卵巢区域配准。
一个实施例中,处理器采用第二学习模型确定第一卵巢区域和第二卵巢区域的对应关系,比如旋转关系,或者平移关系,或者缩放关系,或者以上三种关系的任意组合关系,并根据确定的对应关系调整第二卵巢区域,使得第二卵泡区域与第一卵泡区域配准。主要步骤为:1)构建数据库,数据库中包含大量的数据集及其对应的标记信息,标记信息是对应关系;2)配准步骤,将第一卵巢区域与第二卵巢区域输入至数据库中,判断出对应第一卵巢区域和第二卵巢区域的对应关系,根据判断出的对应关系调整第二卵巢区域,输出调整后的第二卵巢区域。
机器学习方法中又包括深度学习方法,基于深度学习的卵巢组织的配准与前述中基于深度学习的目标卵泡的配准方法类似,在此不再赘述。
上述实施例中,需要在第二超声图像中先确定第二卵巢区域,然后基于获取的第一卵巢区域和第二卵巢区域进行配准。除此之外,还可以采用不需要先确定第二卵巢区域的方式进行卵巢区域的配准,例如传统的图像配准方法。传统的图像配准方法是基于第一卵巢区域的特征信息在第二超声图像中搜索与其相似度满足预设阈值子图像,将相似度满足预设阈值的子图像确定为第二卵泡区域。
其中传统的图像配准方法包括:基于域变换的方法;基于模板匹配的方法;基于图像特征的匹配方法。利用传统的图像匹配方法确定第二卵巢区域与前述中利用传统的图像匹配方法确定第二卵泡区域类似,在此不再赘述。
一个实施例中,操作者可以手动基于图像整体配准的方法实现卵巢区域的配准。接收操作者对第二超声图像和第一超声图像配准的指令,根据所述指令旋转、平移或缩放第二超声图像,以使得第二超声图像和第一超声图像配准。
在对目标卵泡进行配准的过程中,会有新增卵泡的情况出现,或者目标卵泡消失的情况。对于在后一次超声图像中找到与前一次超声图像中目标卵泡相对应的卵泡,进行保留,记录其相关位置已经对应的体积径长等信息;对于在前一次超声图像中出现,在后一次超声图像中消失的目标卵泡,则进行删除;对于在后一次超声图像中新出现的目标卵泡,如果新增卵泡的生长参数满足预设条件,则将器作为新的目标卵泡加入追踪行列,否则删除。其中预设条件为体积满足一定的阈值条件或者径长满足一定的阈值条件,而这个具体的阈值大小可以由处理器自动设定,或者操作者手动设定。
基于上面的描述,根据本申请实施例的卵泡跟踪方法和系统,能够在多个超声图像上对同一个目标卵泡进行跟踪监测,在临床上便能够实现对同一个目标卵泡生长发育情况,据此操作者可以准确评估最佳的取卵时间,有效地提升工作效率和准确性。
尽管这里已经参考附图描述了示例实施例,应理解上述示例实施例仅仅是示例性的,并且不意图将本申请的范围限制于此。本领域普通技术人员可以在其中进行各种改变和修改,而不偏离本申请的范围和精神。所有这些改变和修改意在被包括在所附权利要求所要求的本申请的范围之内。
本领域普通技术人员可以意识到,结合本文中所公开的实施例描述的各示例的单元及算法步骤,能够以电子硬件、或者计算机软件和电子硬件的结合来实现。这些功能究竟以硬件还是软件方式来执行,取决于技术方案的特定应用和设计约束条件。专业技术人员可以对每个特定的应用来使用不同方法来实现所描述的功能,但是这种实现不应认为超出本申请的范围。
在本申请所提供的几个实施例中,应该理解到,所揭露的设备和方法,可以通过其它的方式实现。例如,以上所描述的设备实施例仅仅是示意性的,例如,所述单元的划分,仅仅为一种逻辑功能划分,实际实现时可以有另外的划分方式,例如多个单元或组件可以结合或者可以集成到另一个设备,或一些特征可以忽略,或不执行。
在此处所提供的说明书中,说明了大量具体细节。然而,能够理解,本申请的实施例可以在没有这些具体细节的情况下实践。在一些实例中,并未详细示出公知的方法、结构和技术,以便不模糊对本说明书的理解。
类似地,应当理解,为了精简本申请并帮助理解各个发明方面中的一个或多个,在对本申请的示例性实施例的描述中,本申请的各个特征有时被一起分组到单个实施例、图、或者对其的描述中。然而,并不应将该本申请的方法解释成反映如下意图:即所要求保护的本申请要求比在每个权利要求中所明确记载的特征更多的特征。更确切地说,如相应的权利要求书所反映的那样,其发明点在于可以用少于某个公开的单个实施例的所有特征的特征来解决相应的技术问题。因此,遵循具体实施方式的权利要求书由此明确地并入该具体实施方式,其中每个权利要求本身都作为本申请的单独实施例。
本领域的技术人员可以理解,除了特征之间相互排斥之外,可以采用任何组合对本说明书(包括伴随的权利要求、摘要和附图)中公开的所有特征以及如此公开的任何方法或者设备的所有过程或单元进行组合。除非另外明确陈述,本说明书(包括伴随的权利要求、摘要和附图)中公开的每个特征可以由提供相同、等同或相似目的替代特征来代替。
此外,本领域的技术人员能够理解,尽管在此所述的一些实施例包括其它实施例中所包括的某些特征而不是其它特征,但是不同实施例的特征的组合意味着处于本申请的范围之内并且形成不同的实施例。例如,在权利要求书中,所要求保护的实施例的任意之一都可以以任意的组合方式来使用。
本申请的各个部件实施例可以以硬件实现,或者以在一个或者多个处理器上运行的软件模块实现,或者以它们的组合实现。本领域的技术人员应当理解,可以在实践中使用微处理器或者数字信号处理器(DSP)来实现根据本申请实施例的一些模块的一些或者全部功能。本申请还可以实现为用于执行这里所描述的方法的一部分或者全部的装置程序(例如,计算机程序和计算机程序产品)。这样的实现本申请的程序可以存储在计算机可读介质上,或者可以具有一个或者多个信号的形式。这样的信号可以从因特网网站上下载得到,或者在载体信号上提供,或者以任何其他形式提供。
应该注意的是上述实施例对本申请进行说明而不是对本申请进行限制,并且本领域技术人员在不脱离所附权利要求的范围的情况下可设计出替换实施例。在权利要求中,不应将位于括号之间的任何参考符号构造成对权利要求的限制。本申请可以借助于包括有若干不同元件的硬件以及借助于适当编程的计算机来实现。在列举了若干装置的单元权利要求中,这些装置中的若干个可以是通过同一个硬件项来具体体现。单词第一、第二、以及第三等的使用不表示任何顺序。可将这些单词解释为名称。
以上所述,仅为本申请的具体实施方式或对具体实施方式的说明,本申请的保护范围并不局限于此,任何熟悉本技术领域的技术人员在本申请揭露的技术范围内,可轻易想到变化或替换,都应涵盖在本申请的保护范围之内。本申请的保护范围应以权利要求的保护范围为准。
Claims (38)
- 一种卵泡跟踪方法,其特征在于,包括:获取被测对象的卵巢组织的至少三个不同检查时间的超声图像,其中所述卵巢组织包括目标卵泡;分别在所述至少三个不同检查时间的超声图像上确定所述目标卵泡对应的卵泡区域,获得至少三个卵泡区域,其中所述至少三个卵泡区域为同一个目标卵泡对应的卵泡区域;根据所述至少三个卵泡区域分别确定所述目标卵泡的生长参数,获得所述目标卵泡的至少三个生长参数;根据所述目标卵泡的至少三个生长参数获得所述目标卵泡的生长趋势图;显示所述生长趋势图。
- 根据权利要求1所述的方法,其特征在于:所述生长趋势图为生长参数曲线图,其中所述生长参数曲线图以检查时间为第一坐标,以生长参数为第二坐标;或者,所述生长趋势图为所述不同检查时间对应的所述生长参数的列表。
- 根据权利要求1所述的方法,其特征在于,分别在所述至少三个不同检查时间的超声图像上确定所述目标卵泡对应的卵泡区域包括:在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域;根据所述第一卵泡区域,在所述至少三个不同检查时间的超声图像中的第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域;根据所述第一卵泡区域或第二卵泡区域,在所述至少三个不同检查时间的超声图像中的第三超声图像中确定所述目标卵泡对应的卵泡区域,获得第三卵泡区域。
- 根据权利要求3所述的方法,其特征在于,在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:基于卵泡的图像特征,在所述至少三个不同检查时间的超声图像中的第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域;或者,检测操作者对所述至少三个不同检查时间的超声图像中的第一超声图像中的目标卵泡的对应区域描迹的操作,以获得第一卵泡区域。
- 根据权利要求4所述的方法,其特征在于,所述超声图像为三维超声图像,所述基于卵泡的图像特征,在所述至少三个不同检查时间的超声图像中的第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域,包括:基于卵泡的图像特征,在所述至少三个不同检查时间的三维超声图像中的第一超声图像的多个二维切面图像中对所述目标卵泡的对应区域进行分割;综合所述多个二维切面图像上所述目标卵泡的对应区域,以得到所述第一卵泡区域。
- 根据权利要求5所述的方法,其特征在于,所述第一超声图像的多个二维切面图像为所述第一超声图像中的所有二维切面图像,或者,所述第一超声图像的多个二维切面为对所述第一超声图像中以第一预设规则进行采样得到的采样图像,所述综合所述多个二维切面上的所述目标卵泡的对应区域包括:对所述采样图像的分割结果进行三维插值,以得到所述第一卵泡区域。
- 根据权利要求4所述的方法,其特征在于,所述超声图像为三维超声图像,所述基于卵泡的图像特征,在所述至少三个不同检查时间的超声图像中的第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域,包括:基于所述卵泡的图像特征,在所述至少三个不同检查时间的三维超声图像中的第一超声图像中对所述目标卵泡对应的卵泡区域进行三维分割,以获得所述第一卵泡区域。
- 根据权利要求3所述的方法,其特征在于,在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:在所述至少三个不同检查时间的超声图像中的第一超声图像中确定所述卵巢组织的区域,获得第一卵巢区域;基于所述第一卵巢区域确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域。
- 根据权利要求3所述的方法,其特征在于,在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:在所述至少三个不同检查时间的超声图像中的第一超声图像中确定多个第一候选卵泡区域;获取每个第一候选卵泡区域的生长参数,将生长参数满足第一预设条件的第一候选卵泡区域确定为所述第一卵泡区域。
- 根据权利要求3所述的方法,其特征在于,在所述至少三个不同检查时间的超声图像中的第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:在所述至少三个不同检查时间的超声图像中的第一超声图像中确定多个第一候选卵泡区域;获取每个第一候选卵泡区域的生长参数,将生长参数为最大值的第一候选卵泡区域确定为所述第一卵泡区域。
- 根据权利要求3所述的方法,其特征在于,所述根据所述第一卵泡区域,在所述至少三个不同检查时间的超声图像中的第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域,包括:在所述至少三个不同检查时间的超声图像中的第二超声图像中获取与所述第一卵泡区域的特征信息相似度满足第一阈值的卵泡区域,将获取的卵泡区域确定为第二卵泡区域;或者,获取所述第二超声图像中的目标卵泡对应的第二候选卵泡区域,采用第一学习模 型确定所述第一卵泡区域与所述第二候选卵泡区域的对应关系,并根据所述对应关系调整所述第二候选卵泡区域,并将调整后的所述第二候选卵泡区域确定为第二卵泡区域。
- 根据权利要求3所述的方法,其特征在于,所述根据所述第一卵泡区域,在所述至少三个不同检查时间的超声图像中的第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域,包括:检测操作者在所述至少三个不同检查时间的超声图像中的第二超声图像中标识与所述第一卵泡区域相对应的卵泡区域的操作,将标识的卵泡区域确定为第二卵泡区域。
- 根据权利要求11或12所述的方法,其特征在于,还包括:在所述第一超声图像中确定所述被测对象的卵巢组织的第一卵巢区域;在所述第二超声图像中确定所述被测对象的卵巢组织的第二卵巢区域;将所述第一卵巢区域与所述第二卵巢区域进行配准。
- 根据权利要求13所述的方法,其特征在于,将所述第一卵巢区域与所述第二卵巢区域进行配准,包括:根据所述第一卵巢区域在所述第一超声图像中的方位,调整所述第二卵巢区域在所述第二超声图像中的方位,以使得所述调整后的第二卵巢区域在所述第二超声图像中的方位与所述第一卵巢区域在所述第一超声图像中的方位相同;或者,根据所述第一卵巢区域的大小,调整所述第二卵巢区域相对于所述第一卵巢区域的大小,以使得所述调整后的第二卵巢区域与所述第一卵巢区域的大小相同。
- 根据权利要求13所述的方法,其特征在于,将所述第一卵巢区域与第二卵巢区域进行配准,包括:采用第二学习模型确定所述第二卵巢区域与所述第一卵巢区域的对应关系,并根据所述对应关系调整所述第二卵巢区域;或者,接收操作者根据所述第一卵巢区域调整所述第二卵巢区域的指令,根据所述指令调整所述第二卵巢区域。
- 根据权利要求11或12所述的方法,其特征在于,还包括:在所述第一超声图像中确定所述被测对象的卵巢组织的第一卵巢区域;将所述第二超声图像与所述第一卵巢区域的模板进行匹配,并根据匹配结果确定所述第二超声图像中所述卵巢组织的第二卵巢区域。
- 根据权利要求11或12所述的方法,其特征在于,还包括:接收操作者对所述第二超声图像和所述第一超声图像配准的指令,根据所述指令旋转、平移或缩放第二超声图像,以使得所述第二超声图像和所述第一超声图像配准。
- 根据权利要求1至17中任意一项所述的方法,其特征在于,所述生长参数包括以下至少之一:体积、径长和生长速度。
- 一种卵泡跟踪方法,其特征在于,包括:获取被测对象的卵巢组织的第一超声图像,其中所述卵巢组织包括目标卵泡;在所述第一超声图像中确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域;获取所述被测对象的卵巢组织的第二超声图像;基于所述第一卵泡区域,在所述第二超声图像中确定与所述目标卵泡对应的卵泡区域,获得第二卵泡区域。
- 根据权利要求19所述的方法,其特征在于,在所述第一超声图像中确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:基于卵泡的图像特征,在所述第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域;或者,检测操作者对所述第一超声图像中的目标卵泡的对应区域描迹的操作,以获得第一卵泡区域。
- 根据权利要求20所述的方法,其特征在于,所述超声图像为三维超声图像,所述基于卵泡的图像特征,在所述第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域,包括:基于卵泡的图像特征,在所述第一超声图像的多个二维切面图像中对所述目标卵泡的对应区域进行分割;综合所述多个所述二维切面图像上所述目标卵泡的对应区域,以得到所述第一卵泡区域。
- 根据权利要求21所述的方法,其特征在于,所述第一超声图像的多个二维切面图像为所述第一超声图像中的所有二维切面图像,或者,所述第一超声图像的多个二维切面为对所述第一超声图像中以第一预设规则进行采样得到的采样图像,所述综合所述多个二维切面上的所述目标卵泡的对应区域包括:对所述采样图像的分割结果进行三维插值,以得到所述第一卵泡区域。
- 根据权利要求19所述的方法,其特征在于,所述超声图像为三维超声图像,所述基于卵泡的图像特征,在所述第一超声图像中对所述目标卵泡对应的卵泡区域进行分割,以获得第一卵泡区域,包括:基于卵泡的图像特征,在所述第一超声图像中对所述目标卵泡对应的卵泡区域进行三维分割,以获得所述第一卵泡区域。
- 根据权利要求19所述的方法,其特征在于,在所述第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:在所述第一超声图像中确定所述卵巢组织的区域,获得第一卵巢区域;基于所述第一卵巢区域确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域。
- 根据权利要求19所述的方法,其特征在于,在所述第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:在所述第一超声图像中确定多个第一候选卵泡区域;获取每个第一候选卵泡区域的生长参数,在生长参数满足第一预设条件的第一候选卵泡区域确定为第一卵泡区域。
- 根据权利要求19所述的方法,其特征在于,在所述第一超声图像中,确定所述目标卵泡对应的卵泡区域,获得第一卵泡区域,包括:在所述第一超声图像中确定多个第一候选卵泡区域;获取每个第一候选卵泡区域的生长参数,在生长参数为最大值的第一候选卵泡区 域确定为第一卵泡区域。
- 根据权利要求19所述的方法,其特征在于,所述根据所述第一卵泡区域,在所述第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域,包括:在所述第二超声图像中获取与所述第一卵泡区域的特征信息相似度满足第一阈值的卵泡区域,将获取的卵泡区域确定为第二卵泡区域;或者,获取所述第二超声图像中的目标卵泡对应的第二候选卵泡区域,采用第一学习模型确定所述第一卵泡区域与所述第二候选卵泡区域的对应关系,并根据所述对应关系调整所述第二候选卵泡区域,并将调整后的所述第二候选卵泡区域确定为第二卵泡区域。
- 根据权利要求19所述的方法,其特征在于,所述根据所述第一卵泡区域,在所述第二超声图像中确定所述目标卵泡对应的卵泡区域,获得第二卵泡区域,包括:检测操作者在所述第二超声图像中标识与所述第一卵泡区域相对应的卵泡区域的操作,将标识的卵泡区域确定为第二卵泡区域。
- 根据权利要求27或28所述的方法,其特征在于,还包括:在所述第一超声图像中确定所述被测对象的卵巢组织的第一卵巢区域;在所述第二超声图像中确定所述被测对象的卵巢组织的第二卵巢区域;将所述第一卵巢区域与所述第二卵巢区域进行配准。
- 根据权利要求29所述的方法,其特征在于,将所述第一卵巢区域与所述第二卵巢区域进行配准,包括:根据所述第一卵巢区域在所述第一超声图像中的方位,调整所述第二卵巢区域在所述第二超声图像中的方位,以使得所述调整后的第二卵巢区域在所述第二超声图像中的方位与所述第一卵巢区域在所述第一超声图像中的方位相同;或者,根据所述第一卵巢区域的大小,调整所述第二卵巢区域相对于所述第一卵巢区域的大小,以使得所述调整后的第二卵巢区域与所述第一卵巢区域的大小相同。
- 根据权利要求29所述的方法,其特征在于,将所述第一卵巢区域与所述第二卵巢区域进行配准,包括:采用第二学习模型确定所述第二卵巢区域与所述第一卵巢区域的对应关系,并根据所述对应关系调整所述第二卵巢区域;或者,接收操作者根据所述第一卵巢区域调整所述待配准第二卵巢区域的指令,根据所述指令调整所述待配准第二卵巢区域。
- 根据权利要求27或28所述的方法,其特征在于,还包括:在所述第一超声图像中确定所述卵巢组织的第一卵巢区域;将所述第二超声图像与所述第一卵巢区域的模板进行匹配,并根据匹配结果确定所述第二超声图像中所述卵巢组织的第二卵巢区域。
- 根据权利要求27或28所述的方法,其特征在于,还包括:接收操作者对所述第二超声图像和所述第一超声图像配准的指令,根据所述指令旋转、平移或缩放第二超声图像,以使得所述第二超声图像和所述第一超声图像配准。
- 根据权利要求19-33中任一所述的方法,其特征在于,所述生长参数包括以下 至少之一:体积、径长和生长速度。
- 一种卵泡跟踪方法,其特征在于,包括:获取被测对象的卵巢组织的至少两个不同检查时间的超声图像,其中所述卵巢组织包括目标卵泡;分别在所述至少两个不同检查时间的超声图像上确定所述目标卵泡对应的卵泡区域,获得至少两个卵泡区域,其中所述至少两个卵泡区域为同一个目标卵泡对应的卵泡区域;根据所述至少两个卵泡区域分别确定所述目标卵泡的生长参数,获得所述目标卵泡的至少两个生长参数;根据所述目标卵泡的至少两个生长参数获得所述目标卵泡的生长趋势图;显示所述生长趋势图。
- 一种卵泡跟踪系统,其特征在于,包括:超声探头;发射电路,用于激励所述超声探头向被测对象的卵巢组织发射超声波;接收电路,用于激励所述超声探头接收所述超声波的回波,以获得所述超声波的回波信号;处理器,用于执行如权利要求1-18中任一项所述的卵泡跟踪方法;显示器,用于显示所述生长趋势图。
- 一种卵泡跟踪系统,其特征在于,包括:超声探头;发射电路,用于激励所述超声探头向被测对象的卵巢组织发射超声波;接收电路,用于激励所述超声探头接收所述超声波的回波,以获得所述超声波的回波信号;处理器,用于执行如权利要求19-34中任一项所述的卵泡跟踪方法。
- 一种卵泡跟踪系统,其特征在于,包括:超声探头;发射电路,用于激励所述超声探头向被测对象的卵巢组织发射超声波;接收电路,用于激励所述超声探头接收所述超声波的回波,以获得所述超声波的回波信号;处理器,用于执行如权利要求35中所述的卵泡跟踪方法。
Priority Applications (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2020/140317 WO2022140960A1 (zh) | 2020-12-28 | 2020-12-28 | 一种卵泡跟踪方法和系统 |
CN202080047331.4A CN114041166A (zh) | 2020-12-28 | 2020-12-28 | 一种卵泡跟踪方法和系统 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2020/140317 WO2022140960A1 (zh) | 2020-12-28 | 2020-12-28 | 一种卵泡跟踪方法和系统 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2022140960A1 true WO2022140960A1 (zh) | 2022-07-07 |
Family
ID=80140840
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2020/140317 WO2022140960A1 (zh) | 2020-12-28 | 2020-12-28 | 一种卵泡跟踪方法和系统 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN114041166A (zh) |
WO (1) | WO2022140960A1 (zh) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116912255A (zh) * | 2023-09-14 | 2023-10-20 | 济南宝林信息技术有限公司 | 一种用于卵巢组织分析的卵泡区域分割方法 |
CN117982214A (zh) * | 2024-04-07 | 2024-05-07 | 安徽医科大学第一附属医院 | 一种快速捡卵用辅助识别方法及系统 |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105496453A (zh) * | 2015-11-13 | 2016-04-20 | 武汉科技大学 | 一种黄牛卵泡超声监测系统及其监测方法 |
CN106175826A (zh) * | 2014-08-29 | 2016-12-07 | 三星麦迪森株式会社 | 超声图像显示设备和显示超声图像的方法 |
CN110021025A (zh) * | 2019-03-29 | 2019-07-16 | 上海联影智能医疗科技有限公司 | 感兴趣区域的匹配和显示方法、装置、设备及存储介质 |
CN110197713A (zh) * | 2019-05-10 | 2019-09-03 | 上海依智医疗技术有限公司 | 一种医疗影像的处理方法、装置、设备和介质 |
CN110246135A (zh) * | 2019-07-22 | 2019-09-17 | 新名医(北京)科技有限公司 | 卵泡监测方法、装置、系统及存储介质 |
CN110325119A (zh) * | 2017-02-20 | 2019-10-11 | 皇家飞利浦有限公司 | 卵巢卵泡计数和大小确定 |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110693529B (zh) * | 2019-10-12 | 2021-07-13 | 深圳度影医疗科技有限公司 | 一种基于二维图像的卵巢和/或卵泡自动测量方法 |
-
2020
- 2020-12-28 WO PCT/CN2020/140317 patent/WO2022140960A1/zh active Application Filing
- 2020-12-28 CN CN202080047331.4A patent/CN114041166A/zh active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106175826A (zh) * | 2014-08-29 | 2016-12-07 | 三星麦迪森株式会社 | 超声图像显示设备和显示超声图像的方法 |
CN105496453A (zh) * | 2015-11-13 | 2016-04-20 | 武汉科技大学 | 一种黄牛卵泡超声监测系统及其监测方法 |
CN110325119A (zh) * | 2017-02-20 | 2019-10-11 | 皇家飞利浦有限公司 | 卵巢卵泡计数和大小确定 |
CN110021025A (zh) * | 2019-03-29 | 2019-07-16 | 上海联影智能医疗科技有限公司 | 感兴趣区域的匹配和显示方法、装置、设备及存储介质 |
CN110197713A (zh) * | 2019-05-10 | 2019-09-03 | 上海依智医疗技术有限公司 | 一种医疗影像的处理方法、装置、设备和介质 |
CN110246135A (zh) * | 2019-07-22 | 2019-09-17 | 新名医(北京)科技有限公司 | 卵泡监测方法、装置、系统及存储介质 |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116912255A (zh) * | 2023-09-14 | 2023-10-20 | 济南宝林信息技术有限公司 | 一种用于卵巢组织分析的卵泡区域分割方法 |
CN116912255B (zh) * | 2023-09-14 | 2023-12-19 | 济南宝林信息技术有限公司 | 一种用于卵巢组织分析的卵泡区域分割方法 |
CN117982214A (zh) * | 2024-04-07 | 2024-05-07 | 安徽医科大学第一附属医院 | 一种快速捡卵用辅助识别方法及系统 |
Also Published As
Publication number | Publication date |
---|---|
CN114041166A (zh) | 2022-02-11 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
JP6467041B2 (ja) | 超音波診断装置、及び画像処理方法 | |
CN110325119B (zh) | 卵巢卵泡计数和大小确定 | |
EP3655917B1 (en) | Fetal ultrasound image processing | |
WO2022140960A1 (zh) | 一种卵泡跟踪方法和系统 | |
US20210393240A1 (en) | Ultrasonic imaging method and device | |
TWI501754B (zh) | 影像辨識方法及影像辨識系統 | |
RU2526752C1 (ru) | Система и способ для автоматического планирования двухмерных видов в объемных медицинских изображениях | |
EP3742973B1 (en) | Device and method for obtaining anatomical measurements from an ultrasound image | |
CN111374708B (zh) | 一种胎儿心率检测方法及超声成像装置、存储介质 | |
RU2746152C2 (ru) | Обнаружение биологического объекта | |
Nanthagopal et al. | A region-based segmentation of tumour from brain CT images using nonlinear support vector machine classifier | |
Aji et al. | Automatic measurement of fetal head circumference from 2-dimensional ultrasound | |
CN112998755A (zh) | 解剖结构的自动测量方法和超声成像系统 | |
CN115813433A (zh) | 基于二维超声成像的卵泡测量方法和超声成像系统 | |
Cigale et al. | Directional 3D wavelet transform based on gaussian mixtures for the analysis of 3D ultrasound ovarian volumes | |
US11944486B2 (en) | Analysis method for breast image and electronic apparatus using the same | |
CN116211349A (zh) | 一种胎儿颜面部的超声成像方法、超声成像装置和介质 | |
CN113792740A (zh) | 眼底彩照的动静脉分割方法、系统、设备及介质 | |
CN114699106A (zh) | 超声图像处理方法及设备 | |
KR101900683B1 (ko) | 캡슐 내시경의 영상 시퀀스 분할 장치 및 방법 | |
Farsana et al. | Automaticfetal head circumference measurement in 2D ultrasound fetal images using histogram of oriented gradient | |
CN113974688B (zh) | 超声成像方法和超声成像系统 | |
Okishio et al. | Identifying an egg-containing follicle by displacement analysis of the follicle border | |
WO2022134049A1 (zh) | 胎儿颅骨的超声成像方法和超声成像系统 | |
CN117934356A (zh) | 一种超声成像系统和卵巢间质自动定量分析方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20967312 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20967312 Country of ref document: EP Kind code of ref document: A1 |