WO2023130662A1 - 基于心脏超声视频筛选舒张期和收缩期图像的方法和装置 - Google Patents

基于心脏超声视频筛选舒张期和收缩期图像的方法和装置 Download PDF

Info

Publication number
WO2023130662A1
WO2023130662A1 PCT/CN2022/097244 CN2022097244W WO2023130662A1 WO 2023130662 A1 WO2023130662 A1 WO 2023130662A1 CN 2022097244 W CN2022097244 W CN 2022097244W WO 2023130662 A1 WO2023130662 A1 WO 2023130662A1
Authority
WO
WIPO (PCT)
Prior art keywords
area
point
valley
time curve
peak
Prior art date
Application number
PCT/CN2022/097244
Other languages
English (en)
French (fr)
Inventor
李喆
王思瀚
张碧莹
曹君
Original Assignee
乐普（北京）医疗器械股份有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 乐普（北京）医疗器械股份有限公司 filed Critical 乐普（北京）医疗器械股份有限公司
Publication of WO2023130662A1 publication Critical patent/WO2023130662A1/zh

Links

Images

Classifications

    • 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
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0012—Biomedical image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/40—Analysis of texture
    • G06T7/41—Analysis of texture based on statistical description of texture
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • 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
    • G06T2207/10016—Video; Image sequence
    • 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
    • G06T2207/30048—Heart; Cardiac

Definitions

  • the invention relates to the technical field of data processing, in particular to a method and device for screening diastolic and systolic images based on cardiac ultrasound video.
  • Ejection fraction refers to stroke volume, that is, the percentage of blood ejected from one side of the ventricle in one heartbeat to the end-diastolic volume of the ventricle, and is one of the important indicators for judging the health of the heart.
  • the calculation of ejection fraction is related to end-systole and end-diastole ventricular volume.
  • the real-time output of conventional echocardiography equipment is a set of video data, that is, echocardiographic video. When the operator plays this group of echocardiographic video on the device, he often screens multiple frames of end-systolic and end-diastolic images through personal experience for analysis. A reference image for the score.
  • the purpose of the present invention is to address the defects of the prior art, to provide a method, device, electronic equipment and computer-readable storage medium for screening diastolic and systolic images based on cardiac ultrasound video, and to use image semantic segmentation models to improve left ventricular Recognition accuracy, by identifying the significant peak points and valley points of the left ventricular area-time curve to ensure the recognition accuracy of the left ventricular end-systole and end-diastole extreme points, by further identifying the significant peak points and valley points
  • the time interval is used to improve the recognition accuracy of the extreme points at the end of left ventricular systole and end diastole.
  • the present invention can get rid of the manual intervention link in the traditional method, can improve the screening precision and accuracy of the end-systolic and diastolic images, and can provide stability guarantee for the image screening quality.
  • the first aspect of the embodiment of the present invention provides a method for screening diastolic and systolic images based on cardiac ultrasound video, the method comprising:
  • the framed image corresponding to the effective valley point on the first area-time curve is used as the end-systolic screening image, and the framed image corresponding to the effective peak point is used as the end-diastolic screening image, and according to the preset cardiac cycle
  • the sorting rule sorts all filtered images to generate a corresponding sequence of filtered images.
  • performing image area statistics on each of the left ventricle segmentation images to generate corresponding left ventricle area parameters specifically includes:
  • the total number of pixels of each left ventricle segmentation image is counted as the corresponding left ventricle area parameter.
  • the area-time curve conversion process is performed according to all the left ventricular area parameters to generate a corresponding first area-time curve, which specifically includes:
  • performing the significance threshold confirmation process according to all the left ventricular area parameters to generate a corresponding first significance threshold specifically includes:
  • the left ventricular area parameters According to the order of the left ventricular area parameters from small to large, sort all the left ventricular area parameters to obtain a second area parameter sequence; and in the second area parameter sequence, the first preset sampling The left ventricle area parameter at the position is used as a smaller boundary value, and the left ventricle area parameter at the second preset sampling position that is sorted behind is used as a larger boundary value; and the larger boundary value is combined with the Half of the difference of the smaller boundary value is used as the corresponding first significance threshold.
  • the process of identifying significant peak points and significant valley points on the existing peak points and valley points of the first area-time curve according to the first significance threshold includes:
  • each existing peak point on the first area-time curve as a peak point to be evaluated; and performing a significant feature evaluation process on each peak point to be evaluated to generate a corresponding significance evaluation parameter;
  • the peak point to be evaluated whose evaluation parameter exceeds the first significance threshold is used as the first peak point;
  • each of the first peak points on the first area-time curve as the significance peak point
  • each of the first valley points as the significance valley point
  • performing the significance feature evaluation process on each of the peak points to be evaluated to generate corresponding significance evaluation parameters specifically includes:
  • the parallel line on the horizontal axis of time passing through the peak point to be evaluated is recorded as the first parallel line;
  • the first intersection point between the first parallel line and the falling edge of the left curve of the peak point to be evaluated is recorded as the first left intersection point
  • the curve segment from the first left intersection point to the peak point to be evaluated is Recorded as the first left curve segment
  • the first intersection point of the first parallel line and the rising edge of the right curve of the peak point to be evaluated is recorded as the first right intersection point
  • the peak point to be evaluated to The curve segment of the first right intersection point is recorded as the first right curve segment
  • the valley point with the lowest curve amplitude as the first reference valley point In the first left curve segment, the valley point with the lowest curve amplitude as the first reference valley point; and in the first right curve segment, the valley point with the lowest curve amplitude as the second Reference valley point; from the first and second reference valley points, select a relatively high curve amplitude as the third reference valley point;
  • a curve amplitude difference between the peak point to be evaluated and the third reference valley point is used as the significance evaluation parameter corresponding to the peak point to be evaluated.
  • the screening process of effective peak points and effective valley points on the significant peak points and significant valley points on the first area-time curve specifically includes:
  • the significant peak point on the first area-time curve is recorded as the peak point to be screened; the peak point filtering process is performed on the peak point to be screened on the first area-time curve, specifically: When the time interval between two adjacent peak points to be screened is lower than the preset effective extreme point interval threshold, the peak point to be screened with a lower curve amplitude is filtered from all current peak points to be screened; The remaining peak points to be screened on the first area-time curve continue to perform the peak point filtering process until the time interval between any pair of adjacent peak points to be screened on the first area-time curve Not lower than the effective extreme point interval threshold;
  • valley point filtering on the valley point to be screened on the first area-time curve , specifically: when the time interval between two adjacent valley value points to be screened is lower than the effective extreme point interval threshold, the valley point to be screened with a higher curve amplitude is removed from all current valley value points to be screened Filter out the points; continue to perform the valley point filtering process on the remaining valley points to be screened on the first area-time curve until any pair of adjacent ones on the first area-time curve The time interval of the valley points to be screened is not lower than the effective extreme point interval threshold;
  • the second aspect of the embodiment of the present invention provides a device for implementing the method described in the first aspect above, including: an acquisition module, an image preprocessing module, a saliency processing module, and an image screening module;
  • the acquisition module is used to acquire cardiac ultrasound video
  • the image preprocessing module is used to perform video image frame processing on the cardiac ultrasound video to generate a plurality of framed images; and perform left ventricle semantic segmentation processing on each of the framed images based on the image semantic segmentation model, so that in each A corresponding left ventricle segmented image is obtained from the framed image;
  • the significance processing module is used to perform image area statistics on each of the left ventricle segmentation images to generate corresponding left ventricle area parameters; and perform area-time curve conversion processing according to all the left ventricle area parameters to generate the corresponding first area - a time curve; and performing a significance threshold confirmation process according to all the left ventricular area parameters to generate a corresponding first significance threshold;
  • the image screening module is used to identify significant peak points and significant valley points on the existing peak points and valley points of the first area-time curve according to the first significance threshold; Significant peak points and significant valley points on the first area-time curve are screened for effective peak points and effective valley points; and the corresponding effective valley points on the first area-time curve are
  • the framed images are used as end-systolic screening images, and the framed images corresponding to effective peak points are used as end-diastolic screening images, and all screening images are sorted according to preset cardiac cycle sorting rules to generate corresponding screening image sequences.
  • the third aspect of the embodiment of the present invention provides an electronic device, including: a memory, a processor, and a transceiver;
  • the processor is configured to be coupled with the memory, read and execute instructions in the memory, so as to implement the method steps described in the first aspect above;
  • the transceiver is coupled to the processor, and the processor controls the transceiver to send and receive messages.
  • the fourth aspect of the embodiments of the present invention provides a computer-readable storage medium, the computer-readable storage medium stores computer instructions, and when the computer instructions are executed by a computer, the computer executes the above-mentioned first aspect. method directive.
  • An embodiment of the present invention provides a method, device, electronic device, and computer-readable storage medium for screening diastolic and systolic images based on cardiac ultrasound video.
  • the image semantic segmentation model is used to perform left ventricular Segment the left ventricle segmentation images at different time points; then construct the corresponding area-time curve from the corresponding relationship between the left ventricle image area and time, and take the mean value of the size boundary value from the left ventricle image area as the significance threshold; then Use the significance threshold as the basis to identify the significant peak points and valley points on the area-time curve; then use the effective extreme point interval threshold to screen the significant peak points and valley points; finally, combine the effective points with the effective
  • the framed images corresponding to valley points and peak points are used as the screening results of end-systolic and end-diastolic images.
  • FIG. 1 is a schematic diagram of a method for screening diastolic and systolic images based on cardiac ultrasound video provided by Embodiment 1 of the present invention
  • Embodiment 1 of the present invention is a schematic diagram of the area-time curve provided by Embodiment 1 of the present invention.
  • FIG. 3 is a block diagram of a device for screening diastolic and systolic images based on cardiac ultrasound video provided by Embodiment 2 of the present invention
  • FIG. 4 is a schematic structural diagram of an electronic device provided by Embodiment 3 of the present invention.
  • Embodiment 1 of the present invention provides a method for screening diastolic and systolic images based on cardiac ultrasound video, as shown in Figure 1, which is a schematic diagram of a method for screening diastolic and systolic images based on cardiac ultrasound video provided in Embodiment 1 of the present invention As shown, the method mainly includes the following steps:
  • Step 1 obtain cardiac ultrasound video.
  • the cardiac ultrasound video is the ultrasound video data output by the cardiac ultrasound examination equipment.
  • Step 2 performing video image frame processing on the cardiac ultrasound video to generate a plurality of framed images.
  • the image frame extraction operation is performed on the echocardiography video according to the set video frame sampling rate, and the sub-frame image is the extracted single-frame image, and each sub-frame image corresponds to a time point parameter.
  • Step 3 perform left ventricle semantic segmentation processing on each sub-frame image based on the image semantic segmentation model, so as to obtain a corresponding left ventricle segmentation image in each sub-frame image;
  • the network structure of the image semantic segmentation model can adopt a variety of neural network structures for image semantic segmentation, including the DeepLabV3 network structure; the DeepLabV3 network structure can refer to the authors Liang-Chieh Chen, George Papandreou, Florian Schroff and Hartwig Adam The published article "Rethinking Atrous Convolution for Semantic Image Segmentation" will not be further described here; the pixel values of all pixels of the left ventricle segmentation image are the unified first pixel value.
  • the first pixel value is usually selected from other color pixel values that have obvious color contrast effect with the framed image.
  • the left ventricle semantic segmentation process is performed on each framed image to output an output image with the same size as the framed image.
  • the pixel values of the output image only include two values: the preset foreground pixel value and background pixel value; in the output image, the foreground image composed of all pixel values of the foreground pixel value is the model left ventricle segmentation image; in the framed image, each pixel corresponding to the model left ventricle segmentation image
  • the pixel value of the point as the first pixel value
  • the original cardiac ultrasound images are extracted from the training data set that stores a large number of cardiac ultrasound images as training images, and then the left ventricle shape segmentation corresponding to each original cardiac ultrasound image is extracted manually or by machine
  • the segmented image of is used as the comparison image; and the training data pair consisting of multiple sets of training images-comparison images is used to train the image semantic segmentation model until the model error converges.
  • Step 4 performing image area statistics on each left ventricle segmented image to generate corresponding left ventricle area parameters
  • the number of pixels in the segmented image of the left ventricle is used as a parameter representing the size of the area, that is, the area parameter of the left ventricle.
  • Step 5 performing area-time curve conversion processing according to all left ventricular area parameters to generate a corresponding first area-time curve
  • the time curve conversion process generates a first area-time curve.
  • a two-dimensional area-time coordinate space is constructed with the area parameter as the vertical axis and time as the horizontal axis, and then the left ventricular area parameters of the first area parameter sequence are correspondingly plotted in the area-time coordinate space, and then sequentially Connect all tracer coordinate points to obtain the first area-time curve.
  • Step 6 performing significance threshold confirmation processing according to all left ventricular area parameters to generate a corresponding first significance threshold
  • the left ventricle area parameter is used as the smaller boundary value, and the left ventricle area parameter at the second preset sampling position that is sorted behind is used as the larger boundary value; half of the difference between the larger boundary value and the smaller boundary value is used as a corresponding The first significance threshold of .
  • the ratio of the first preset sampling position to the total number of left ventricular area parameters in the second area parameter sequence is a set ratio, that is, the first ratio
  • the second preset sampling position and the left ventricular area parameter in the second area parameter sequence The ratio of the total is also a set ratio, that is, the second ratio.
  • the second area parameter sequence has 100 left ventricle area parameters, the first ratio is a%, and the second ratio is b%; then, the first preset sampling position is the ath left ventricle area in the second area parameter sequence
  • Step 7 according to the first significance threshold, perform significant peak point and significant valley point identification processing on the existing peak point and valley point of the first area-time curve;
  • the first area-time curve is not a smooth curve, and there are many local maxima and minima on the curve, and these local maxima and minima are mostly noise points, not the actual end-systolic and diastolic extremum points , in order to improve the recognition accuracy of extreme points at the end of systole and end of diastole, the current step analyzes the significance of extreme points on the first area-time curve, and uses the first significance threshold as the basis for the significance judgment of peak points to identify each Significant peak point and significant valley point;
  • step 71 using each existing peak point on the first area-time curve as a peak point to be evaluated; and performing a significant feature evaluation process on each peak point to be evaluated to generate a corresponding significance evaluation parameter;
  • the peak point to be evaluated whose evaluation parameter exceeds the first significance threshold is taken as the first peak point;
  • the first significance threshold is used to judge whether the peak point to be evaluated is a significant peak point; before using the first significance threshold to judge the significance peak point, it is necessary to evaluate the significance of each peak point to be evaluated parameter;
  • significance feature evaluation process is performed on each peak point to be evaluated to generate corresponding significance evaluation parameters, specifically including:
  • Step A1 record the area-time curve where the peak point to be evaluated is located as the current area-time curve
  • the area-time curve where the peak point to be evaluated is located that is, the current area-time curve is actually the first area-time curve
  • Step A2 on the current area-time curve, the parallel line on the time horizontal axis passing through the peak point to be evaluated is recorded as the first parallel line;
  • the current area-time curve is shown in Figure 2 as a schematic diagram of the area-time curve provided by Embodiment 1 of the present invention.
  • the peak point P to be evaluated as an example, the first parallel line obtained by making a parallel line on the time axis through point P is as follows: As shown in Figure 2;
  • Step A3 record the first intersection point between the first parallel line and the falling edge of the left curve of the peak point to be evaluated as the first left intersection point, and record the curve segment from the first left intersection point to the peak point to be evaluated as the first left side curve segment; and the first intersection point between the first parallel line and the rising edge of the curve on the right side of the peak point to be evaluated is recorded as the first right side point of intersection, and the curve segment from the peak point to be evaluated to the first right side point of intersection point is recorded as first right curve segment;
  • the first left intersection point PL is the intersection point between the first parallel line and the falling edge of the first curve on the left side of the peak point to be evaluated, and the first left curve segment is PL The curve segment of -P;
  • the first right intersection point PR is the first intersection point of the first parallel line and the rising edge of the curve on the right side of the peak point to be evaluated, and the first left curve segment is the curve segment of P-PR;
  • the first left curve segment is from the starting position of the curve to the peak value to be evaluated The curve segment between points; if there is no intersection with the rising edge of the curve on the right side of the peak point to be evaluated, that is, when the first right intersection point is empty, then the first right curve segment is from the peak point to be evaluated to the end position of the curve the curve segment between;
  • Step A4 taking the valley point with the lowest curve amplitude in the first left curve segment as the first reference valley point; and taking the valley point with the lowest curve amplitude in the first right curve segment as the second reference Valley point; from the first and second reference valley points, select the one with a relatively higher curve amplitude as the third reference valley point;
  • the first left curve segment that is, the PL-P curve segment
  • the valley point with the lowest curve amplitude is V1
  • the first right curve segment is the P-PR curve segment, including two valley points V3 and V4, among which the valley point with a relatively high amplitude of the curve is V4 , then the second reference valley point is V4; among the first reference valley point V1 and the second reference valley point V4, the curve amplitude is relatively higher than the first reference valley point V1, so the third reference valley point The point should be V1;
  • Step A5 taking the curve amplitude difference between the peak point to be evaluated and the third reference valley point as the significance evaluation parameter corresponding to the peak point to be evaluated;
  • the significance evaluation parameter corresponding to the peak point P to be evaluated should be the amplitude of point P - the amplitude of point V1;
  • the conventional evaluation of the significance of the curve or waveform is based on the amplitude difference from the peak point to the baseline as the significance evaluation parameter, but in reality, the area-time curve is an irregular curve, and there is a high possibility of baseline drift , and the baseline drift is also difficult to be an ideal linear drift structure.
  • the obtained peak point significance evaluation parameter error will be very large; the embodiment of the present invention solves this problem through the above steps A1-A4 for each to-be To evaluate the peak point, select a relative baseline point, that is, the third reference valley point. By selecting the relative baseline point, the error impact of baseline drift on the evaluation of the salient features can be reduced, and the accuracy of the salient feature evaluation can be improved.
  • step A5 calculate the remaining By evaluating the amplitude difference between the peak point and the third reference valley point, a more accurate significance evaluation parameter can be obtained;
  • Step 72 performing curve inversion processing on the first area-time curve to generate a corresponding second area-time curve; and using each existing peak point on the second area-time curve as a peak point to be evaluated; and for each peak value to be evaluated Significant feature evaluation processing is performed on the point to generate corresponding significance evaluation parameters; and the peak point to be evaluated whose significance evaluation parameter exceeds the first significance threshold is used as the second peak point; and according to the second area-time curve and the first area - the curve inversion correspondence of the time curve, the existing valley point corresponding to each second peak point inversion in the first area-time curve is used as the first valley point;
  • the current step is actually to use the first significance threshold to judge the significance of each existing valley point of the first area-time curve.
  • the embodiment of the present invention first inverts the first area-time curve to obtain the second area-time curve, and the second The peak point in the area-time curve is also the valley point in the first area-time curve; then use the above-mentioned steps A1-A5 to carry out the significant feature evaluation process on each peak point of the second area-time curve, and the obtained significant
  • the evaluation parameter is actually the significance evaluation parameter of each valley point in the first area-time curve;
  • Step 73 taking each first peak point on the first area-time curve as a significant peak point, and each first valley point as a significant valley point.
  • Step 8 performing effective peak point and effective valley point screening processing on the significant peak point and significant valley point on the first area-time curve
  • step 81 recording the significant peak point on the first area-time curve as the peak point to be screened; performing peak point filtering processing on the peak point to be screened on the first area-time curve;
  • the peak point filtering process is performed on the peak points to be screened on the first area-time curve, specifically including: when the time interval between two adjacent peak points to be screened is lower than the preset effective extreme point interval threshold , to filter out the peak points to be sieved with lower curve amplitudes from all current peak points to be sieved; continue to filter peak points on the remaining peak points to be sieved on the first area-time curve until the first area- The time interval between any pair of adjacent peak points to be screened on the time curve is not lower than the effective extreme point interval threshold;
  • the peak-to-peak distance between adjacent effective peak points is greater than the effective extreme point interval threshold, and the effective extreme point interval threshold is a preset interval threshold ;
  • the embodiment of the present invention also supports another implementation method when performing peak point filtering processing on the peak points to be screened on the first area-time curve, and the steps include:
  • Step B1 on the first area-time curve, use the first peak point to be screened as the first reference point;
  • Step B2 taking the next peak point to be screened of the first reference point as the second reference point;
  • Step B3 calculating the time interval of the first and second reference points to generate the first time interval
  • Step B4 judging whether the first time interval is greater than the effective extreme point interval threshold; if the first time interval is greater than the effective extreme point interval threshold, then use the second reference point as the new first reference point; if the first time interval is less than or equal to the effective extreme point interval threshold, then in the first and second reference points, the peak points to be screened with smaller amplitudes will be filtered out from all current peak points to be screened, and the peaks to be screened with larger amplitudes will be filtered out point as the new first reference point;
  • Step B5 judging whether the new first reference point is the last peak point to be screened on the first area-time curve, if so, stop the current peak point filtering process, if not, go to step B2;
  • the last peak point to be screened on the first area-time curve can also be used as the first reference point, and the previous peak point to be screened before the first reference point can be used as the second reference point, and from the first reference point Points start to traverse forward;
  • the peak point to be screened with the largest amplitude on the first area-time curve can also be used as the first reference point, and traverse forward and backward simultaneously from the first reference point ;
  • Step 82 the significant valley point on the first area-time curve is recorded as the valley point to be sieved;
  • the valley point to be screened on the first area-time curve is carried out valley point filtering process;
  • the valley point filtering process is performed on the valley points to be screened on the first area-time curve, specifically including: when the time interval between two adjacent valley points to be screened is lower than the effective extreme point interval threshold , filter the valley point to be sieved with a higher curve amplitude from all current valley points to be sieved; continue to filter valley points on the remaining valley points to be sieved on the first area-time curve until The time interval between any pair of adjacent valley points to be screened on the first area-time curve is not lower than the effective extreme point interval threshold;
  • the valley distance between adjacent effective valley points is greater than the effective extreme point interval threshold
  • the embodiment of the present invention also supports another implementation mode when performing valley point filtering processing on the valley point to be screened on the first area-time curve, and the steps include:
  • Step C1 on the first area-time curve, use the first valley point to be screened as the third reference point;
  • Step C2 taking the next peak point to be screened of the third reference point as the fourth reference point;
  • Step C3 calculating the time intervals of the third and fourth reference points to generate a second time interval
  • Step C4 judging whether the second time interval is greater than the effective extreme point interval threshold; if the second time interval is greater than the effective extreme point interval threshold, then use the fourth reference point as the new third reference point; if the second time interval is less than or equal to the effective extreme point interval threshold, then in the third and fourth reference points, the valley points to be screened with larger amplitudes will be filtered out from all current valley points to be screened, and the valley points to be screened with smaller amplitudes will be filtered out.
  • the sieve valley point is used as the new third reference point;
  • Step C5 judging whether the new third reference point is the last valley point to be screened on the first area-time curve, if so, stop the current valley point filtering process, otherwise go to step C2;
  • the last valley point to be sieved on the first area-time curve can also be used as the third reference point, and the valley point to be screened before the third reference point can be used as the fourth reference point, and from the first Three reference points start to traverse forward;
  • the valley point to be screened with the minimum amplitude on the first area-time curve can also be used as the third reference point, starting from the third reference point forward and backward simultaneously After traversal;
  • Step 83 taking the last remaining peak point to be screened on the first area-time curve as an effective peak point, and the last remaining valley point to be screened as an effective valley point.
  • the effective peak point obtained on the first area-time curve should correspond to the end-diastole time point of each heartbeat
  • the effective valley point should correspond to the end-systole time point of each heartbeat
  • Step 9 Use the framed image corresponding to the effective valley point on the first area-time curve as the end-systolic screening image, and the framed image corresponding to the effective peak point as the end-diastolic screening image, and sort the images according to the preset cardiac cycle sorting rules. All screened images are sorted to generate a corresponding screened image sequence.
  • the framed image corresponding to the effective valley point that is, the screened image at the end of systole
  • the framed image corresponding to the effective peak point that is, the screened image at the end of diastole is the framed image of each heartbeat.
  • the framed images with the most obvious characteristics of the diastolic period; the cardiac cycle sorting rules can be defined according to actual needs, including at least the following rules: sort all end-systolic images in chronological order, or sort all diastolic images in chronological order. Sorting the end-of-systole images, or sorting each group of end-systole images + end-diastole images in chronological order and in the order that each systole first and diastole last.
  • the real-time ejection fraction of the current heartbeat can be counted according to the end-systolic screening image + end-diastolic screening image of each group;
  • the end-diastolic mean ventricular volume was calculated from the end-systolic image sequence, and then the statistical ejection fraction was calculated based on the end-systolic mean ventricular volume and the end-diastolic mean ventricular volume.
  • FIG. 3 is a block diagram of a device for screening diastolic and systolic images based on cardiac ultrasound video provided in Embodiment 2 of the present invention.
  • the device can be a terminal device or a server that implements the method of the embodiment of the present invention, or it can be connected with
  • the device connected to the above-mentioned terminal device or server to implement the method in the embodiment of the present invention may be, for example, a device or a chip system of the above-mentioned terminal device or server.
  • the device includes: an acquisition module 201 , an image preprocessing module 202 , a saliency processing module 203 and an image screening module 204 .
  • the acquiring module 201 is used for acquiring cardiac ultrasound video.
  • the image preprocessing module 202 is used to perform video image frame processing on the echocardiography video to generate a plurality of frame images; and perform left ventricle semantic segmentation processing on each frame image based on the image semantic segmentation model, thereby obtaining in each frame image A corresponding segmented image of the left ventricle.
  • the significance processing module 203 is used to perform image area statistics on each left ventricle segmented image to generate corresponding left ventricle area parameters; and perform area-time curve conversion processing according to all left ventricle area parameters to generate a corresponding first area-time curve; and A significance threshold confirmation process is performed according to all left ventricle area parameters to generate a corresponding first significance threshold.
  • the image screening module 204 is used to identify the peak points and valley points of the first area-time curve according to the first significance threshold; and to identify the peak points and valley points of the first area-time curve Significant peak points and significant valley points above are screened for effective peak points and effective valley points; and the framed images corresponding to effective valley points on the first area-time curve are used as end-systolic screening images, effective peak The framed images corresponding to the points are used as end-diastolic screening images, and all screening images are sorted according to the preset cardiac cycle sorting rules to generate corresponding screening image sequences.
  • the embodiment of the present invention provides a device for screening diastolic and systolic images based on echocardiographic video, which can execute the method steps in the above-mentioned method embodiments, and its implementation principle and technical effect are similar, and will not be repeated here.
  • each module of the above device is only a division of logical functions, and may be fully or partially integrated into one physical entity or physically separated during actual implementation.
  • these modules can all be implemented in the form of calling software through processing elements; they can also be implemented in the form of hardware; some modules can also be implemented in the form of calling software through processing elements, and some modules can be implemented in the form of hardware.
  • the acquisition module can be a separate processing element, or it can be integrated into a chip of the above-mentioned device.
  • it can also be stored in the memory of the above-mentioned device in the form of program code, and a certain processing element of the above-mentioned device can Call and execute the functions of the modules identified above.
  • each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or an instruction in the form of software.
  • the above modules may be one or more integrated circuits configured to implement the above method, for example: one or more specific integrated circuits (Application Specific Integrated Circuit, ASIC), or, one or more digital signal processors ( Digital Signal Processor, DSP), or, one or more Field Programmable Gate Arrays (Field Programmable Gate Array, FPGA), etc.
  • ASIC Application Specific Integrated Circuit
  • DSP Digital Signal Processor
  • FPGA Field Programmable Gate Array
  • the processing element may be a general-purpose processor, such as a central processing unit (Central Processing Unit, CPU) or other processors that can call program codes.
  • these modules can be integrated together and implemented in the form of a System-on-a-chip (SOC).
  • SOC System-on-a-chip
  • all or part of them may be implemented by software, hardware, firmware or any combination thereof.
  • software When implemented using software, it may be implemented in whole or in part in the form of a computer program product.
  • the computer program product includes one or more computer instructions.
  • the above-mentioned computers may be general-purpose computers, special-purpose computers, computer networks, or other programmable devices.
  • the above-mentioned computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.
  • the above-mentioned computer-readable storage medium may be any available medium that can be accessed by a computer, or a data storage device such as a server or a data center integrated with one or more available media.
  • the above-mentioned usable medium may be a magnetic medium (for example, a floppy disk, a hard disk, or a magnetic tape), an optical medium (for example, DVD), or a semiconductor medium (for example, a solid state disk (solid state disk, SSD)) and the like.
  • FIG. 4 is a schematic structural diagram of an electronic device provided by Embodiment 3 of the present invention.
  • the electronic device may be the aforementioned terminal device or server, or may be a terminal device or server connected to the aforementioned terminal device or server to implement the method of the embodiment of the present invention.
  • the electronic device may include: a processor 301 (such as a CPU), a memory 302, and a transceiver 303; Various instructions may be stored in the memory 302 for completing various processing functions and realizing the methods and processing procedures provided in the above-mentioned embodiments of the present invention.
  • the electronic device involved in this embodiment of the present invention further includes: a power supply 304 , a system bus 305 and a communication port 306 .
  • the system bus 305 is used to realize the communication connection among the components.
  • the above-mentioned communication port 306 is used for connection and communication between the electronic device and other peripheral devices.
  • the system bus mentioned in FIG. 4 may be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (Extended Industry Standard Architecture, EISA) bus or the like.
  • PCI Peripheral Component Interconnect
  • EISA Extended Industry Standard Architecture
  • the system bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.
  • the communication interface is used to realize the communication between the database access device and other devices (such as client, read-write library and read-only library).
  • the memory may include random access memory (Random Access Memory, RAM), and may also include non-volatile memory (Non-Volatile Memory), such as at least one disk memory.
  • processor can be general-purpose processor, comprises central processing unit CPU, network processor (Network Processor, NP) etc.; Can also be digital signal processor DSP, application-specific integrated circuit ASIC, field programmable gate array FPGA or other available Program logic devices, discrete gate or transistor logic devices, discrete hardware components.
  • CPU central processing unit
  • NP Network Processor
  • DSP digital signal processor
  • ASIC application-specific integrated circuit
  • FPGA field programmable gate array
  • embodiments of the present invention also provide a computer-readable storage medium, and instructions are stored in the storage medium, and when the storage medium is run on a computer, the computer executes the methods and processing procedures provided in the above-mentioned embodiments.
  • the embodiment of the present invention also provides a chip for running instructions, and the chip is used for executing the method and the processing procedure provided in the foregoing embodiments.
  • Embodiments of the present invention provide a method, device, electronic device, and computer-readable storage medium for screening diastolic and systolic images based on echocardiographic video.
  • the image semantic segmentation model is used to improve the recognition accuracy of the left ventricle.
  • Significant peak points and valley points of the area-time curve ensure the recognition accuracy of left ventricular end-systole and end-diastole extreme points, and further identify the effective time interval of significant peak points and valley points to improve the accuracy of Accuracy of identification of left ventricular end-systole and end-diastole extreme points.
  • the manual intervention link in the traditional method is eliminated, the precision and accuracy of screening images at the end of systole and end of diastole are improved, and the stability of image screening quality is guaranteed.
  • RAM random access memory
  • ROM read-only memory
  • EEPROM electrically programmable ROM
  • EEPROM electrically erasable programmable ROM
  • registers hard disk, removable disk, CD-ROM, or any other Any other known storage medium.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Probability & Statistics with Applications (AREA)
  • Biophysics (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Quality & Reliability (AREA)
  • Geometry (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Biomedical Technology (AREA)
  • Medical Informatics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Ultra Sonic Daignosis Equipment (AREA)

Abstract

本发明实施例涉及一种基于心脏超声视频筛选舒张期和收缩期图像的方法和装置，所述方法包括：获取心脏超声视频；进行视频图像分帧生成多个分帧图像；对各个分帧图像进行左心室语义分割得到对应的左心室分割图像；进行图像面积统计生成对应的左心室面积参数；进行面积-时间曲线转换生成第一面积-时间曲线；进行显著性阈值确认生成第一显著性阈值；对第一面积-时间曲线现有的峰、谷值点进行显著性峰、谷值点识别；对显著性峰、谷值点进行有效峰、谷值点筛选；将有效谷值点对应的分帧图像作为收缩期末筛选图像，有效峰值点对应的分帧图像作为舒张期末筛选图像，并对所有筛选图像进行排序生成筛选图像序列。通过本发明可以保障图像筛选质量稳定性。

Description

基于心脏超声视频筛选舒张期和收缩期图像的方法和装置
本申请要求于2022年1月7日提交中国专利局、申请号为202210018424.5、发明名称为“基于心脏超声视频筛选舒张期和收缩期图像的方法和装置”的中国专利申请的优先权。
技术领域
本发明涉及数据处理技术领域，特别涉及一种基于心脏超声视频筛选舒张期和收缩期图像的方法和装置。
背景技术
射血分数指每搏输出量也就是一次心搏中一侧心室射出的血量占心室舒张末期容积量的百分比，是判断心脏健康状态的重要指征之一。射血分数的计算与收缩期末、舒张期末心室容积有关。常规心脏超声检查设备实时输出的是一组视频数据也就是心脏超声视频，操作人员在设备播放该组心脏超声视频时，往往是通过个人经验从中筛选多帧收缩期末、舒张期末图像作为分析射血分数的参考图像。这种操作方式过于依赖个人的人工经验以及对设备的人工操作熟练程度，无法对收缩期末、舒张期末图像的筛选质量稳定性进行有效保障。而被筛选出的图像将会作为进一步计算射血分数的计算依据，在图像筛选质量较差的情况下还可能对射血分数的计算精度造成影响。
发明内容
本发明的目的，就是针对现有技术的缺陷，提供一种基于心脏超声视频筛选舒张期和收缩期图像的方法、装置、电子设备及计算机可读存储介质，使用 图像语义分割模型来提高左心室识别精度，通过识别左心室面积-时间曲线的显著性峰值点、谷值点来确保对左心室收缩期末、舒张期末极值点的识别精度，通过进一步识别显著性峰值点、谷值点的有效时间间隔来提高对左心室收缩期末、舒张期末极值点的识别准确度。通过本发明，可以摆脱传统做法中的人工干预环节，可以提高对收缩期末、舒张期末图像的筛选精度与准确度，可以对图像筛选质量给出稳定性保障。
为实现上述目的，本发明实施例第一方面提供了一种基于心脏超声视频筛选舒张期和收缩期图像的方法，所述方法包括：
获取心脏超声视频；
对所述心脏超声视频进行视频图像分帧处理生成多个分帧图像；
基于图像语义分割模型对各个所述分帧图像进行左心室语义分割处理，从而在各个所述分帧图像中得到一个对应的左心室分割图像；
对各个所述左心室分割图像进行图像面积统计生成对应的左心室面积参数；
根据所有所述左心室面积参数进行面积-时间曲线转换处理生成对应的第一面积-时间曲线；
根据所有所述左心室面积参数进行显著性阈值确认处理生成对应的第一显著性阈值；
根据所述第一显著性阈值对所述第一面积-时间曲线现有的峰值点与谷值点进行显著性峰值点与显著性谷值点识别处理；
对所述第一面积-时间曲线上的显著性峰值点与显著性谷值点进行有效峰值点和有效谷值点筛选处理；
将所述第一面积-时间曲线上有效谷值点对应的所述分帧图像作为收缩期末筛选图像，有效峰值点对应的所述分帧图像作为舒张期末筛选图像，并按预设的心动周期排序规则对所有筛选图像进行排序生成对应的筛选图像序列。
优选的，所述对各个所述左心室分割图像进行图像面积统计生成对应的 左心室面积参数，具体包括：
统计各个所述左心室分割图像的像素点总数，作为对应的所述左心室面积参数。
优选的，所述根据所有所述左心室面积参数进行面积-时间曲线转换处理生成对应的第一面积-时间曲线，具体包括：
按对应的所述分帧图像的时间先后顺序，对所有所述左心室面积参数进行排序得到第一面积参数序列；并以面积参数为纵轴、时间为横轴，对所述第一面积参数序列进行面积-时间曲线转换处理生成所述第一面积-时间曲线。
优选的，所述根据所有所述左心室面积参数进行显著性阈值确认处理生成对应的第一显著性阈值，具体包括：
按所述左心室面积参数从小到大的顺序，对所有所述左心室面积参数进行排序得到第二面积参数序列；并将所述第二面积参数序列中，排序靠前的第一预设采样位置处的所述左心室面积参数作为较小边界值，排序靠后的第二预设采样位置处的所述左心室面积参数作为较大边界值；并将所述较大边界值与所述较小边界值的差值的一半作为对应的所述第一显著性阈值。
优选的，所述根据所述第一显著性阈值对所述第一面积-时间曲线现有的峰值点与谷值点进行显著性峰值点与显著性谷值点识别处理，具体包括：
将所述第一面积-时间曲线上各个现有峰值点作为待评估峰值点；并对各个所述待评估峰值点进行显著性特征评估处理生成对应的显著性评估参数；并将所述显著性评估参数超过所述第一显著性阈值的所述待评估峰值点作为第一峰值点；
对所述第一面积-时间曲线进行曲线倒置处理生成对应的第二面积-时间曲线；并将所述第二面积-时间曲线上各个现有峰值点作为待评估峰值点；并对各个所述待评估峰值点进行显著性特征评估处理生成对应的显著性评估参数；并将所述显著性评估参数超过所述第一显著性阈值的所述待评估峰值点作为第二峰值点；并根据所述第二面积-时间曲线与所述第一面积-时间曲线 的曲线倒置对应关系，将所述第一面积-时间曲线中与各个所述第二峰值点倒置对应的现有谷值点作为第一谷值点；
将所述第一面积-时间曲线上各个所述第一峰值点作为所述显著性峰值点，各个所述第一谷值点作为所述显著性谷值点。
进一步的，所述对各个所述待评估峰值点进行显著性特征评估处理生成对应的显著性评估参数，具体包括：
将所述待评估峰值点所在的面积-时间曲线记为当前面积-时间曲线；
在所述当前面积-时间曲线上，过所述待评估峰值点做时间横轴的平行线记为第一平行线；
将所述第一平行线与所述待评估峰值点左侧曲线下降沿的第一个交点记为第一左侧交点，将所述第一左侧交点到所述待评估峰值点的曲线片段记为第一左侧曲线片段；并将所述第一平行线与所述待评估峰值点右侧曲线上升沿的第一个交点记为第一右侧交点，将所述待评估峰值点到所述第一右侧交点的曲线片段，记为第一右侧曲线片段；
将所述第一左侧曲线片段中，曲线幅值最低的谷值点作为第一参考谷值点；并将所述第一右侧曲线片段中，曲线幅值最低的谷值点作为第二参考谷值点；从所述第一、第二参考谷值点中，选择曲线幅值相对较高的作为第三参考谷值点；
将所述待评估峰值点与所述第三参考谷值点的曲线幅值差，作为与所述待评估峰值点对应的所述显著性评估参数。
优选的，所述对所述第一面积-时间曲线上的显著性峰值点与显著性谷值点进行有效峰值点和有效谷值点筛选处理，具体包括：
将所述第一面积-时间曲线上的所述显著性峰值点记为待筛峰值点；对所述第一面积-时间曲线上的所述待筛峰值点进行峰值点过滤处理，具体为：当相邻的两个待筛峰值点的时间间隔低于预设的有效极值点间隔阈值时，将其中曲线幅值较低的待筛峰值点从当前所有待筛峰值点中滤除；对所述第一面 积-时间曲线上剩余的所述待筛峰值点持续进行所述峰值点过滤处理，直到所述第一面积-时间曲线上任意一对相邻的待筛峰值点的时间间隔都不低于所述有效极值点间隔阈值为止；
将所述第一面积-时间曲线上的所述显著性谷值点记为待筛谷值点；对所述第一面积-时间曲线上的所述待筛谷值点进行谷值点过滤处理，具体为：当相邻的两个待筛谷值点的时间间隔低于所述有效极值点间隔阈值时，将其中曲线幅值较高的待筛谷值点从当前所有待筛谷值点中滤除；对所述第一面积-时间曲线上剩余的所述待筛谷值点持续进行所述谷值点过滤处理，直到所述第一面积-时间曲线上任意一对相邻的待筛谷值点的时间间隔都不低于所述有效极值点间隔阈值为止；
将所述第一面积-时间曲线上最后剩余的所述待筛峰值点作为所述有效峰值点，最后剩余的所述待筛谷值点作为所述有效谷值点。
本发明实施例第二方面提供了一种实现上述第一方面所述的方法的装置，包括：获取模块、图像预处理模块、显著性处理模块和图像筛选模块；
所述获取模块用于获取心脏超声视频；
所述图像预处理模块用于对所述心脏超声视频进行视频图像分帧处理生成多个分帧图像；并基于图像语义分割模型对各个所述分帧图像进行左心室语义分割处理，从而在各个所述分帧图像中得到一个对应的左心室分割图像；
所述显著性处理模块用于对各个所述左心室分割图像进行图像面积统计生成对应的左心室面积参数；并根据所有所述左心室面积参数进行面积-时间曲线转换处理生成对应的第一面积-时间曲线；并根据所有所述左心室面积参数进行显著性阈值确认处理生成对应的第一显著性阈值；
所述图像筛选模块用于根据所述第一显著性阈值对所述第一面积-时间曲线现有的峰值点与谷值点进行显著性峰值点与显著性谷值点识别处理；并对所述第一面积-时间曲线上的显著性峰值点与显著性谷值点进行有效峰值点和有效谷值点筛选处理；并将所述第一面积-时间曲线上有效谷值点对应的 所述分帧图像作为收缩期末筛选图像，有效峰值点对应的所述分帧图像作为舒张期末筛选图像，并按预设的心动周期排序规则对所有筛选图像进行排序生成对应的筛选图像序列。
本发明实施例第三方面提供了一种电子设备，包括：存储器、处理器和收发器；
所述处理器用于与所述存储器耦合，读取并执行所述存储器中的指令，以实现上述第一方面所述的方法步骤；
所述收发器与所述处理器耦合，由所述处理器控制所述收发器进行消息收发。
本发明实施例第四方面提供了一种计算机可读存储介质，所述计算机可读存储介质存储有计算机指令，当所述计算机指令被计算机执行时，使得所述计算机执行上述第一方面所述的方法的指令。
本发明实施例提供了一种基于心脏超声视频筛选舒张期和收缩期图像的方法、装置、电子设备及计算机可读存储介质，首先使用图像语义分割模型对心脏超声视频的分帧图像进行左心室分割得到不同时间点上的左心室分割图像；再由左心室图像面积与时间的对应关系构建对应的面积-时间曲线，并从左心室图像面积中取大小边界值的均值作为显著性阈值；再以显著性阈值为判断依据对面积-时间曲线进行显著性峰值点、谷值点识别；再以有效极值点间隔阈值对显著性峰值点、谷值点进行有效点筛选；最后，将与有效谷值点、峰值点对应的分帧图像作为收缩期末、舒张期末图像筛选结果。通过本发明，摆脱了传统做法中的人工干预环节，提高了对收缩期末、舒张期末图像的筛选精度与准确度，对图像筛选质量给出了稳定性保障。
附图说明
图1为本发明实施例一提供的一种基于心脏超声视频筛选舒张期和收缩期图像的方法示意图；
图2为本发明实施例一提供的面积-时间曲线示意图；
图3为本发明实施例二提供的一种基于心脏超声视频筛选舒张期和收缩期图像的装置的模块结构图；
图4为本发明实施例三提供的一种电子设备的结构示意图。
具体实施方式
为了使本发明的目的、技术方案和优点更加清楚，下面将结合附图对本发明作进一步地详细描述，显然，所描述的实施例仅仅是本发明一部份实施例，而不是全部的实施例。基于本发明中的实施例，本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其它实施例，都属于本发明保护的范围。
本发明实施例一提供的一种基于心脏超声视频筛选舒张期和收缩期图像的方法，如图1为本发明实施例一提供的一种基于心脏超声视频筛选舒张期和收缩期图像的方法示意图所示，本方法主要包括如下步骤：
步骤1，获取心脏超声视频。
这里，心脏超声视频即是心脏超声检查设备输出的超声视频数据。
步骤2，对心脏超声视频进行视频图像分帧处理生成多个分帧图像。
这里，按设定的视频帧采样率对心脏超声视频进行图像帧提取操作，分帧图像也就是提取出的单帧图像，每个分帧图像对应一个时间点参数。
步骤3，基于图像语义分割模型对各个分帧图像进行左心室语义分割处理，从而在各个分帧图像中得到一个对应的左心室分割图像；
其中，图像语义分割模型的网络结构可采用多种用于实现图像语义分割的神经网络结构，其中包括DeepLabV3网络结构；DeepLabV3网络结构可参考由作者Liang-Chieh Chen、George Papandreou、Florian Schroff和Hartwig Adam发表的文章《Rethinking Atrous Convolution for Semantic Image Segmentation》，在此不做进一步赘述；左心室分割图像的所有像素点的像素值为统一的第一像素值。
此处，第一像素值常规选用与分帧图像有明显颜色对比效果的其他颜色像素值。
这里，基于图像语义分割模型对各个分帧图像进行左心室语义分割处理输出一个尺寸与分帧图像尺寸一致的输出图像，该输出图像的像素点像素值只包括两种取值：预设的前景像素值和背景像素值；该输出图像中，所有像素值为前景像素值的像素点构成的前景图像即为模型左心室分割图像；在分帧图像中将与模型左心室分割图像对应的各个像素点的像素值设为第一像素值，即可在各个分帧图像中得到一个对应的左心室分割图像。
基于图像语义分割模型对各个分帧图像进行左心室语义分割处理之前，需要对图像语义分割模型进行模型训练。对图像语义分割模型进行模型训练时，从存储了海量心脏超声图像的训练数据集中提取原始的心脏超声图像作为训练图像，再提取与各个原始心脏超声图像对应的经由人工或机器完成左心室形状分割的分割图像作为比对图像；并由多组训练图像-比对图像组成的训练数据对，对图像语义分割模型进行训练，直到模型误差收敛为止。
步骤4，对各个左心室分割图像进行图像面积统计生成对应的左心室面积参数；
具体包括：统计各个左心室分割图像的像素点总数，作为对应的左心室面积参数。
这里，以左心室分割图像的像素点数量作为表征面积大小的参数也就是左心室面积参数。
步骤5，根据所有左心室面积参数进行面积-时间曲线转换处理生成对应的第一面积-时间曲线；
具体包括：按对应的分帧图像的时间先后顺序，对所有左心室面积参数进行排序得到第一面积参数序列；并以面积参数为纵轴、时间为横轴，对第一面积参数序列进行面积-时间曲线转换处理生成第一面积-时间曲线。
这里，以面积参数为纵轴、时间为横轴构建二维的面积-时间坐标空间， 再在面积-时间坐标空间上对第一面积参数序列的左心室面积参数进行对应坐标点描计，再依次连接所有描计坐标点得到第一面积-时间曲线。
步骤6，根据所有左心室面积参数进行显著性阈值确认处理生成对应的第一显著性阈值；
具体包括：按左心室面积参数从小到大的顺序，对所有左心室面积参数进行排序得到第二面积参数序列；并将第二面积参数序列中，排序靠前的第一预设采样位置处的左心室面积参数作为较小边界值，排序靠后的第二预设采样位置处的左心室面积参数作为较大边界值；并将较大边界值与较小边界值的差值的一半作为对应的第一显著性阈值。
这里，第一预设采样位置与第二面积参数序列的左心室面积参数总数的比值为一个设定比值也就是第一比值，第二预设采样位置与第二面积参数序列的左心室面积参数总数的比值也为一个设定比值也就是第二比值。当前步骤之所以不直接选择最小值和最大值作为边界值，是为了避免由突变极值点产生的误差，第一显著性阈值在后续步骤会被用作判断峰值点的显著性特征。
例如，第二面积参数序列有100个左心室面积参数，第一比值为a％，第二比值为b％；那么,第一预设采样位置就是第二面积参数序列中第a个左心室面积参数的所在位置，第二预设采样位置就是第b个左心室面积参数的所在位置；第一显著性阈值＝(较大边界值-较小边界值)/2＝(第b个左心室面积参数-第a个左心室面积参数)/2。
步骤7，根据第一显著性阈值对第一面积-时间曲线现有的峰值点与谷值点进行显著性峰值点与显著性谷值点识别处理；
这里，第一面积-时间曲线并不是光滑曲线，曲线上存在许多局部极大、极小值，而这些局部极大、极小值大多是一些噪点，并非实际的收缩期末、舒张期末极值点，为了能够提高收缩期末、舒张期末极值点的识别精度，当前步骤对第一面积-时间曲线进行极值点显著性分析，并以第一显著性阈值作为峰值点显著性判断依据识别出各个显著性峰值点与显著性谷值点；
具体包括：步骤71，将第一面积-时间曲线上各个现有峰值点作为待评估峰值点；并对各个待评估峰值点进行显著性特征评估处理生成对应的显著性评估参数；并将显著性评估参数超过第一显著性阈值的待评估峰值点作为第一峰值点；
这里，第一显著性阈值用于对待评估峰值点是否为显著性峰值点进行判断；在使用第一显著性阈值对显著性峰值点进行判断之前，需要评估出各个待评估峰值点的显著性评估参数；
进一步的，对各个待评估峰值点进行显著性特征评估处理生成对应的显著性评估参数，具体包括：
步骤A1，将待评估峰值点所在的面积-时间曲线记为当前面积-时间曲线；
这里，待评估峰值点所在的面积-时间曲线也就是当前面积-时间曲线实际就是第一面积-时间曲线；
步骤A2，在当前面积-时间曲线上，过待评估峰值点做时间横轴的平行线记为第一平行线；
例如，当前面积-时间曲线如图2为本发明实施例一提供的面积-时间曲线示意图所示，以待评估峰值点P为例，过P点做时间轴平行线得到的第一平行线如图2所示；
步骤A3，将第一平行线与待评估峰值点左侧曲线下降沿的第一个交点记为第一左侧交点，将第一左侧交点到待评估峰值点的曲线片段记为第一左侧曲线片段；并将第一平行线与待评估峰值点右侧曲线上升沿的第一个交点记为第一右侧交点，将待评估峰值点到第一右侧交点的曲线片段，记为第一右侧曲线片段；
例如图2所示，以待评估峰值点P为例，第一左侧交点PL为第一平行线与待评估峰值点左侧第一个曲线下降沿的交点，第一左侧曲线片段为PL-P的曲线片段；第一右侧交点PR为第一平行线与待评估峰值点右侧曲线上升沿的第一个交点，第一左侧曲线片段为P-PR的曲线片段；
需要说明的是，倘若第一平行线与待评估峰值点左侧曲线下降沿没有交点也就是第一左侧交点为空的时候，第一左侧曲线片段就是从曲线起始位置到待评估峰值点之间的曲线段；倘若与待评估峰值点右侧曲线上升沿没有交点也就是第一右侧交点为空的时候，那么第一右侧曲线片段就是从待评估峰值点到曲线结束位置之间的曲线段；
步骤A4，将第一左侧曲线片段中，曲线幅值最低的谷值点作为第一参考谷值点；并将第一右侧曲线片段中，曲线幅值最低的谷值点作为第二参考谷值点；从第一、第二参考谷值点中，选择曲线幅值相对较高的作为第三参考谷值点；
例如图2所示，以待评估峰值点P为例，第一左侧曲线片段也就是PL-P曲线片段中，包括两个谷值点V1和V2，其中曲线幅值最低的谷值点为V1，那么第一参考谷值点就是V1；第一右侧曲线片段也就是P-PR曲线片段中，包括两个谷值点V3和V4，其中曲线幅值相对较高的谷值点为V4，那么第二参考谷值点就是V4；第一参考谷值点V1和第二参考谷值点V4中，曲线幅值相对较高的是第一参考谷值点V1，因此第三参考谷值点就应为V1；
步骤A5，将待评估峰值点与第三参考谷值点的曲线幅值差，作为与待评估峰值点对应的显著性评估参数；
例如图2所示，以待评估峰值点P为例，待评估峰值点P对应的显著性评估参数就应为P点的幅值-V1点的幅值；
这里，常规对曲线或波形显著性进行评估时都是以峰值点到基线的幅差作为显著性评估参数，但实际情况中面积-时间曲线是一个不规则的曲线，极可能存在基线漂移的情况，并且基线漂移也很难是理想的线性漂移结构，这种情况下得到的峰值点显著性评估参数误差就会很大；本发明实施例为解决这个问题通过上述步骤A1-A4为每个待评估峰值点选定一个相对基线点也就是第三参考谷值点，通过选择相对基线点可以降低基线漂移对显著性特征评估的误差影响，提高显著性特征评估精度，最后通过上述步骤A5计算待评估峰 值点与第三参考谷值点的幅差就可得到精度更高的显著性评估参数；
步骤72，对第一面积-时间曲线进行曲线倒置处理生成对应的第二面积-时间曲线；并将第二面积-时间曲线上各个现有峰值点作为待评估峰值点；并对各个待评估峰值点进行显著性特征评估处理生成对应的显著性评估参数；并将显著性评估参数超过第一显著性阈值的待评估峰值点作为第二峰值点；并根据第二面积-时间曲线与第一面积-时间曲线的曲线倒置对应关系，将第一面积-时间曲线中与各个第二峰值点倒置对应的现有谷值点作为第一谷值点；
这里，当前步骤实际是使用第一显著性阈值对第一面积-时间曲线的各个现有谷值点进行显著性谷值点判断，在对显著性谷值点进行判断之前，需要评估出各个待评估谷值点的显著性评估参数；本发明实施例在评估各个待评估谷值点的显著性评估参数时，先将第一面积-时间曲线进行曲线倒置得到第二面积-时间曲线，第二面积-时间曲线中的峰值点也就是第一面积-时间曲线中的谷值点；再采用上述步骤A1-A5对第二面积-时间曲线的各个峰值点进行显著性特征评估处理，得到的显著性评估参数实际就是第一面积-时间曲线中各个谷值点的显著性评估参数；
步骤73，将第一面积-时间曲线上各个第一峰值点作为显著性峰值点，各个第一谷值点作为显著性谷值点。
步骤8，对第一面积-时间曲线上的显著性峰值点与显著性谷值点进行有效峰值点和有效谷值点筛选处理；
这里，通过前述步骤过滤了大多数显著性较弱的噪点之后，第一面积-时间曲线上还可能存在一些由环境干扰导致的显著性较强的局部极值点，当前步骤再进一步通过时间间隔对这些干扰极值点进行过滤；
具体包括：步骤81，将第一面积-时间曲线上的显著性峰值点记为待筛峰值点；对第一面积-时间曲线上的待筛峰值点进行峰值点过滤处理；
进一步的，对第一面积-时间曲线上的待筛峰值点进行峰值点过滤处理， 具体包括：当相邻的两个待筛峰值点的时间间隔低于预设的有效极值点间隔阈值时，将其中曲线幅值较低的待筛峰值点从当前所有待筛峰值点中滤除；对第一面积-时间曲线上剩余的待筛峰值点持续进行峰值点过滤处理，直到第一面积-时间曲线上任意一对相邻的待筛峰值点的时间间隔都不低于有效极值点间隔阈值为止；
这里，在完成峰值点过滤处理之后的第一面积-时间曲线上，相邻有效峰值点间的峰峰间距大于有效极值点间隔阈值，有效极值点间隔阈值为一个预先设定的间隔阈值；
需要说明的是，本发明实施例在对第一面积-时间曲线上的待筛峰值点进行峰值点过滤处理时还支持另一种实现方式，其步骤包括：
步骤B1，在第一面积-时间曲线上，将第一个待筛峰值点作为第一参考点；
步骤B2，将第一参考点的下一个待筛峰值点作为第二参考点；
步骤B3，计算第一、第二参考点的时间间隔生成第一时间间隔；
步骤B4，判断第一时间间隔是否大于有效极值点间隔阈值；若第一时间间隔大于有效极值点间隔阈值，则将第二参考点作为新的第一参考点；若第一时间间隔小于或等于有效极值点间隔阈值，则在第一、第二参考点中将幅值较小的待筛峰值点从当前所有待筛峰值点中滤除，并将幅值较大的待筛峰值点作为新的第一参考点；
步骤B5，判断新的第一参考点是否为第一面积-时间曲线上的最后一个待筛峰值点，若是则停止当前的峰值点过滤处理，若否则转至步骤B2；
基于上述实现方式，还可以第一面积-时间曲线上的最后一个待筛峰值点作为第一参考点，以第一参考点的前一个待筛峰值点作为第二参考点，并从第一参考点开始向前进行遍历；基于上述实现方式，还可以第一面积-时间曲线上的最大幅值的待筛峰值点作为第一参考点，从第一参考点开始同时向前和向后进行遍历；
步骤82，将第一面积-时间曲线上的显著性谷值点记为待筛谷值点；对第 一面积-时间曲线上的待筛谷值点进行谷值点过滤处理；
进一步的，对第一面积-时间曲线上的待筛谷值点进行谷值点过滤处理，具体包括：当相邻的两个待筛谷值点的时间间隔低于有效极值点间隔阈值时，将其中曲线幅值较高的待筛谷值点从当前所有待筛谷值点中滤除；对第一面积-时间曲线上剩余的待筛谷值点持续进行谷值点过滤处理，直到第一面积-时间曲线上任意一对相邻的待筛谷值点的时间间隔都不低于有效极值点间隔阈值为止；
这里，在完成谷值点过滤处理之后的第一面积-时间曲线上，相邻有效谷值点间的谷谷间距大于有效极值点间隔阈值；
需要说明的是，本发明实施例在对第一面积-时间曲线上的待筛谷值点进行谷值点过滤处理时还支持另一种实现方式，其步骤包括：
步骤C1，在第一面积-时间曲线上，将第一个待筛谷值点作为第三参考点；
步骤C2，将第三参考点的下一个待筛峰值点作为第四参考点；
步骤C3，计算第三、第四参考点的时间间隔生成第二时间间隔；
步骤C4，判断第二时间间隔是否大于有效极值点间隔阈值；若第二时间间隔大于有效极值点间隔阈值，则将第四参考点作为新的第三参考点；若第二时间间隔小于或等于有效极值点间隔阈值，则在第三、第四参考点中将幅值较大的待筛谷值点从当前所有待筛谷值点中滤除，并将幅值较小的待筛谷值点作为新的第三参考点；
步骤C5，判断新的第三参考点是否为第一面积-时间曲线上的最后一个待筛谷值点，若是则停止当前的谷值点过滤处理，若否则转至步骤C2；
基于上述实现方式，还可以第一面积-时间曲线上的最后一个待筛谷值点作为第三参考点，以第三参考点的前一个待筛谷值点作为第四参考点，并从第三参考点开始向前进行遍历；基于上述实现方式，还可以第一面积-时间曲线上的最小幅值的待筛谷值点作为第三参考点，从第三参考点开始同时向前和向后进行遍历；
步骤83，将第一面积-时间曲线上最后剩余的待筛峰值点作为有效峰值点，最后剩余的待筛谷值点作为有效谷值点。
这里，第一面积-时间曲线上获得的有效峰值点对应的应是每次心搏的舒张期末时间点，有效谷值点对应的应是每次心搏的收缩期末时间点。
步骤9，将第一面积-时间曲线上有效谷值点对应的分帧图像作为收缩期末筛选图像，有效峰值点对应的分帧图像作为舒张期末筛选图像，并按预设的心动周期排序规则对所有筛选图像进行排序生成对应的筛选图像序列。
这里，有效谷值点对应的分帧图像也就是收缩期末筛选图像为各次心搏收缩期特征最明显的分帧图像，有效峰值点对应的分帧图像也就是舒张期末筛选图像为各次心搏舒张期特征最明显的分帧图像；心动周期排序规则可根据实际需求进行定义，其中至少包括以下几种规则：按时间先后顺序对所有收缩期末图像进行排序，或者按时间先后顺序对所有舒张期末图像进行排序，或者按时间先后顺序并以每次收缩期在前、舒张期在后的顺序对每组收缩期末图像+舒张期末图像进行排序。
在得到筛选图像序列之后，就可以根据每组收缩期末筛选图像+舒张期末筛选图像，统计当次心搏的实时射血分数；还可以根据收缩期末筛选图像序列计算收缩期末心室容积均值，根据舒张期末筛选图像序列计算舒张期末心室容积均值，再根据收缩期末心室容积均值和舒张期末心室容积均值计算出统计射血分数。
图3为本发明实施例二提供的一种基于心脏超声视频筛选舒张期和收缩期图像的装置的模块结构图，该装置可以为实现本发明实施例方法的终端设备或者服务器，也可以为与上述终端设备或者服务器连接的实现本发明实施例方法的装置，例如该装置可以是上述终端设备或者服务器的装置或芯片系统。如图3所示，该装置包括：获取模块201、图像预处理模块202、显著性处理模块203和图像筛选模块204。
获取模块201用于获取心脏超声视频。
图像预处理模块202用于对心脏超声视频进行视频图像分帧处理生成多个分帧图像；并基于图像语义分割模型对各个分帧图像进行左心室语义分割处理，从而在各个分帧图像中得到一个对应的左心室分割图像。
显著性处理模块203用于对各个左心室分割图像进行图像面积统计生成对应的左心室面积参数；并根据所有左心室面积参数进行面积-时间曲线转换处理生成对应的第一面积-时间曲线；并根据所有左心室面积参数进行显著性阈值确认处理生成对应的第一显著性阈值。
图像筛选模块204用于根据第一显著性阈值对第一面积-时间曲线现有的峰值点与谷值点进行显著性峰值点与显著性谷值点识别处理；并对第一面积-时间曲线上的显著性峰值点与显著性谷值点进行有效峰值点和有效谷值点筛选处理；并将第一面积-时间曲线上有效谷值点对应的分帧图像作为收缩期末筛选图像，有效峰值点对应的分帧图像作为舒张期末筛选图像，并按预设的心动周期排序规则对所有筛选图像进行排序生成对应的筛选图像序列。
本发明实施例提供的一种基于心脏超声视频筛选舒张期和收缩期图像的装置，可以执行上述方法实施例中的方法步骤，其实现原理和技术效果类似，在此不再赘述。
需要说明的是，应理解以上装置的各个模块的划分仅仅是一种逻辑功能的划分，实际实现时可以全部或部分集成到一个物理实体上，也可以物理上分开。且这些模块可以全部以软件通过处理元件调用的形式实现；也可以全部以硬件的形式实现；还可以部分模块通过处理元件调用软件的形式实现，部分模块通过硬件的形式实现。例如，获取模块可以为单独设立的处理元件，也可以集成在上述装置的某一个芯片中实现，此外，也可以以程序代码的形式存储于上述装置的存储器中，由上述装置的某一个处理元件调用并执行以上确定模块的功能。其它模块的实现与之类似。此外这些模块全部或部分可以集成在一起，也可以独立实现。这里所描述的处理元件可以是一种集成电路，具有信号的处理能力。在实现过程中，上述方法的各步骤或以上各个模块可以通过处理 器元件中的硬件的集成逻辑电路或者软件形式的指令完成。
例如，以上这些模块可以是被配置成实施以上方法的一个或多个集成电路，例如：一个或多个特定集成电路(Application Specific Integrated Circuit，ASIC)，或，一个或多个数字信号处理器(Digital Signal Processor，DSP)，或，一个或者多个现场可编程门阵列(Field Programmable Gate Array，FPGA)等。再如，当以上某个模块通过处理元件调度程序代码的形式实现时，该处理元件可以是通用处理器，例如中央处理器(Central Processing Unit，CPU)或其它可以调用程序代码的处理器。再如，这些模块可以集成在一起，以片上系统(System-on-a-chip，SOC)的形式实现。
在上述实施例中，可以全部或部分地通过软件、硬件、固件或者其任意组合来实现。当使用软件实现时，可以全部或部分地以计算机程序产品的形式实现。该计算机程序产品包括一个或多个计算机指令。在计算机上加载和执行该计算机程序指令时，全部或部分地产生按照本发明实施例所描述的流程或功能。上述计算机可以是通用计算机、专用计算机、计算机网络、或者其他可编程装置。上述计算机指令可以存储在计算机可读存储介质中，或者从一个计算机可读存储介质向另一个计算机可读存储介质传输，例如，上述计算机指令可以从一个网站站点、计算机、服务器或数据中心通过有线(例如同轴电缆、光纤、数字用户线路(Digital Subscriber Line，DSL))或无线(例如红外、无线、蓝牙、微波等)方式向另一个网站站点、计算机、服务器或数据中心进行传输。上述计算机可读存储介质可以是计算机能够存取的任何可用介质或者是包含一个或多个可用介质集成的服务器、数据中心等数据存储设备。上述可用介质可以是磁性介质，(例如，软盘、硬盘、磁带)、光介质(例如，DVD)、或者半导体介质(例如固态硬盘(solid state disk，SSD))等。
图4为本发明实施例三提供的一种电子设备的结构示意图。该电子设备可以为前述的终端设备或者服务器，也可以为与前述终端设备或者服务器连接的实现本发明实施例方法的终端设备或服务器。如图4所示，该电子设备 可以包括：处理器301(例如CPU)、存储器302、收发器303；收发器303耦合至处理器301，处理器301控制收发器303的收发动作。存储器302中可以存储各种指令，以用于完成各种处理功能以及实现本发明上述实施例中提供的方法和处理过程。优选的，本发明实施例涉及的电子设备还包括：电源304、系统总线305以及通信端口306。系统总线305用于实现元件之间的通信连接。上述通信端口306用于电子设备与其他外设之间进行连接通信。
在图4中提到的系统总线可以是外设部件互连标准(Peripheral Component Interconnect，PCI)总线或扩展工业标准结构(Extended Industry Standard Architecture，EISA)总线等。该系统总线可以分为地址总线、数据总线、控制总线等。为便于表示，图中仅用一条粗线表示，但并不表示仅有一根总线或一种类型的总线。通信接口用于实现数据库访问装置与其他设备(例如客户端、读写库和只读库)之间的通信。存储器可能包含随机存取存储器(Random Access Memory，RAM)，也可能还包括非易失性存储器(Non-Volatile Memory)，例如至少一个磁盘存储器。
上述的处理器可以是通用处理器，包括中央处理器CPU、网络处理器(Network Processor，NP)等；还可以是数字信号处理器DSP、专用集成电路ASIC、现场可编程门阵列FPGA或者其他可编程逻辑器件、分立门或者晶体管逻辑器件、分立硬件组件。
需要说明的是，本发明实施例还提供一种计算机可读存储介质，该存储介质中存储有指令，当其在计算机上运行时，使得计算机执行上述实施例中提供的方法和处理过程。
本发明实施例还提供一种运行指令的芯片，该芯片用于执行上述实施例中提供的方法和处理过程。
本发明实施例提供了一种基于心脏超声视频筛选舒张期和收缩期图像的方法、装置、电子设备及计算机可读存储介质，使用图像语义分割模型来提高了左心室识别精度，通过识别左心室面积-时间曲线的显著性峰值点、谷值点 来确保了对左心室收缩期末、舒张期末极值点的识别精度，通过进一步识别显著性峰值点、谷值点的有效时间间隔来提高了对左心室收缩期末、舒张期末极值点的识别准确度。通过本发明，摆脱了传统做法中的人工干预环节，提高了对收缩期末、舒张期末图像的筛选精度与准确度，对图像筛选质量给出了稳定性保障。
专业人员应该还可以进一步意识到，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、计算机软件或者二者的结合来实现，为了清楚地说明硬件和软件的可互换性，在上述说明中已经按照功能一般性地描述了各示例的组成及步骤。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发明的范围。
结合本文中所公开的实施例描述的方法或算法的步骤可以用硬件、处理器执行的软件模块，或者二者的结合来实施。软件模块可以置于随机存储器(RAM)、内存、只读存储器(ROM)、电可编程ROM、电可擦除可编程ROM、寄存器、硬盘、可移动磁盘、CD-ROM、或技术领域内所公知的任意其它形式的存储介质中。
以上所述的具体实施方式，对本发明的目的、技术方案和有益效果进行了进一步详细说明，所应理解的是，以上所述仅为本发明的具体实施方式而已，并不用于限定本发明的保护范围，凡在本发明的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本发明的保护范围之内。

Claims (10)

  1. 一种基于心脏超声视频筛选舒张期和收缩期图像的方法，其特征在于，所述方法包括：
    获取心脏超声视频；
    对所述心脏超声视频进行视频图像分帧处理生成多个分帧图像；
    基于图像语义分割模型对各个所述分帧图像进行左心室语义分割处理，从而在各个所述分帧图像中得到一个对应的左心室分割图像；
    对各个所述左心室分割图像进行图像面积统计生成对应的左心室面积参数；
    根据所有所述左心室面积参数进行面积-时间曲线转换处理生成对应的第一面积-时间曲线；
    根据所有所述左心室面积参数进行显著性阈值确认处理生成对应的第一显著性阈值；
    根据所述第一显著性阈值对所述第一面积-时间曲线现有的峰值点与谷值点进行显著性峰值点与显著性谷值点识别处理；
    对所述第一面积-时间曲线上的显著性峰值点与显著性谷值点进行有效峰值点和有效谷值点筛选处理；
    将所述第一面积-时间曲线上有效谷值点对应的所述分帧图像作为收缩期末筛选图像，有效峰值点对应的所述分帧图像作为舒张期末筛选图像，并按预设的心动周期排序规则对所有筛选图像进行排序生成对应的筛选图像序列。
  2. 根据权利要求1所述的基于心脏超声视频筛选舒张期和收缩期图像的方法，其特征在于，所述对各个所述左心室分割图像进行图像面积统计生成对应的左心室面积参数，具体包括：
    统计各个所述左心室分割图像的像素点总数，作为对应的所述左心室面积参数。
  3. 根据权利要求1所述的基于心脏超声视频筛选舒张期和收缩期图像的 方法，其特征在于，所述根据所有所述左心室面积参数进行面积-时间曲线转换处理生成对应的第一面积-时间曲线，具体包括：
    按对应的所述分帧图像的时间先后顺序，对所有所述左心室面积参数进行排序得到第一面积参数序列；并以面积参数为纵轴、时间为横轴，对所述第一面积参数序列进行面积-时间曲线转换处理生成所述第一面积-时间曲线。
  4. 根据权利要求1所述的基于心脏超声视频筛选舒张期和收缩期图像的方法，其特征在于，所述根据所有所述左心室面积参数进行显著性阈值确认处理生成对应的第一显著性阈值，具体包括：
    按所述左心室面积参数从小到大的顺序，对所有所述左心室面积参数进行排序得到第二面积参数序列；并将所述第二面积参数序列中，排序靠前的第一预设采样位置处的所述左心室面积参数作为较小边界值，排序靠后的第二预设采样位置处的所述左心室面积参数作为较大边界值；并将所述较大边界值与所述较小边界值的差值的一半作为对应的所述第一显著性阈值。
  5. 根据权利要求1所述的基于心脏超声视频筛选舒张期和收缩期图像的方法，其特征在于，所述根据所述第一显著性阈值对所述第一面积-时间曲线现有的峰值点与谷值点进行显著性峰值点与显著性谷值点识别处理，具体包括：
    将所述第一面积-时间曲线上各个现有峰值点作为待评估峰值点；并对各个所述待评估峰值点进行显著性特征评估处理生成对应的显著性评估参数；并将所述显著性评估参数超过所述第一显著性阈值的所述待评估峰值点作为第一峰值点；
    对所述第一面积-时间曲线进行曲线倒置处理生成对应的第二面积-时间曲线；并将所述第二面积-时间曲线上各个现有峰值点作为待评估峰值点；并对各个所述待评估峰值点进行显著性特征评估处理生成对应的显著性评估参数；并将所述显著性评估参数超过所述第一显著性阈值的所述待评估峰值点作为第二峰值点；并根据所述第二面积-时间曲线与所述第一面积-时间曲线 的曲线倒置对应关系，将所述第一面积-时间曲线中与各个所述第二峰值点倒置对应的现有谷值点作为第一谷值点；
    将所述第一面积-时间曲线上各个所述第一峰值点作为所述显著性峰值点，各个所述第一谷值点作为所述显著性谷值点。
  6. 根据权利要求5所述的基于心脏超声视频筛选舒张期和收缩期图像的方法，其特征在于，所述对各个所述待评估峰值点进行显著性特征评估处理生成对应的显著性评估参数，具体包括：
    将所述待评估峰值点所在的面积-时间曲线记为当前面积-时间曲线；
    在所述当前面积-时间曲线上，过所述待评估峰值点做时间横轴的平行线记为第一平行线；
    将所述第一平行线与所述待评估峰值点左侧曲线下降沿的第一个交点记为第一左侧交点，将所述第一左侧交点到所述待评估峰值点的曲线片段记为第一左侧曲线片段；并将所述第一平行线与所述待评估峰值点右侧曲线上升沿的第一个交点记为第一右侧交点，将所述待评估峰值点到所述第一右侧交点的曲线片段，记为第一右侧曲线片段；
    将所述第一左侧曲线片段中，曲线幅值最低的谷值点作为第一参考谷值点；并将所述第一右侧曲线片段中，曲线幅值最低的谷值点作为第二参考谷值点；从所述第一、第二参考谷值点中，选择曲线幅值相对较高的作为第三参考谷值点；
    将所述待评估峰值点与所述第三参考谷值点的曲线幅值差，作为与所述待评估峰值点对应的所述显著性评估参数。
  7. 根据权利要求1所述的基于心脏超声视频筛选舒张期和收缩期图像的方法，其特征在于，所述对所述第一面积-时间曲线上的显著性峰值点与显著性谷值点进行有效峰值点和有效谷值点筛选处理，具体包括：
    将所述第一面积-时间曲线上的所述显著性峰值点记为待筛峰值点；对所述第一面积-时间曲线上的所述待筛峰值点进行峰值点过滤处理，具体为：当 相邻的两个待筛峰值点的时间间隔低于预设的有效极值点间隔阈值时，将其中曲线幅值较低的待筛峰值点从当前所有待筛峰值点中滤除；对所述第一面积-时间曲线上剩余的所述待筛峰值点持续进行所述峰值点过滤处理，直到所述第一面积-时间曲线上任意一对相邻的待筛峰值点的时间间隔都不低于所述有效极值点间隔阈值为止；
    将所述第一面积-时间曲线上的所述显著性谷值点记为待筛谷值点；对所述第一面积-时间曲线上的所述待筛谷值点进行谷值点过滤处理，具体为：当相邻的两个待筛谷值点的时间间隔低于所述有效极值点间隔阈值时，将其中曲线幅值较高的待筛谷值点从当前所有待筛谷值点中滤除；对所述第一面积-时间曲线上剩余的所述待筛谷值点持续进行所述谷值点过滤处理，直到所述第一面积-时间曲线上任意一对相邻的待筛谷值点的时间间隔都不低于所述有效极值点间隔阈值为止；
    将所述第一面积-时间曲线上最后剩余的所述待筛峰值点作为所述有效峰值点，最后剩余的所述待筛谷值点作为所述有效谷值点。
  8. 一种用于实现权利要求1-7任一项所述的基于心脏超声视频筛选舒张期和收缩期图像的方法步骤的装置，其特征在于，所述装置包括：获取模块、图像预处理模块、显著性处理模块和图像筛选模块；
    所述获取模块用于获取心脏超声视频；
    所述图像预处理模块用于对所述心脏超声视频进行视频图像分帧处理生成多个分帧图像；并基于图像语义分割模型对各个所述分帧图像进行左心室语义分割处理，从而在各个所述分帧图像中得到一个对应的左心室分割图像；
    所述显著性处理模块用于对各个所述左心室分割图像进行图像面积统计生成对应的左心室面积参数；并根据所有所述左心室面积参数进行面积-时间曲线转换处理生成对应的第一面积-时间曲线；并根据所有所述左心室面积参数进行显著性阈值确认处理生成对应的第一显著性阈值；
    所述图像筛选模块用于根据所述第一显著性阈值对所述第一面积-时间 曲线现有的峰值点与谷值点进行显著性峰值点与显著性谷值点识别处理；并对所述第一面积-时间曲线上的显著性峰值点与显著性谷值点进行有效峰值点和有效谷值点筛选处理；并将所述第一面积-时间曲线上有效谷值点对应的所述分帧图像作为收缩期末筛选图像，有效峰值点对应的所述分帧图像作为舒张期末筛选图像，并按预设的心动周期排序规则对所有筛选图像进行排序生成对应的筛选图像序列。
  9. 一种电子设备，其特征在于，包括：存储器、处理器和收发器；
    所述处理器用于与所述存储器耦合，读取并执行所述存储器中的指令，以实现权利要求1-7任一项所述的方法步骤；
    所述收发器与所述处理器耦合，由所述处理器控制所述收发器进行消息收发。
  10. 一种计算机可读存储介质，其特征在于，所述计算机可读存储介质存储有计算机指令，当所述计算机指令被计算机执行时，使得所述计算机执行权利要求1-7任一项所述的方法的指令。
PCT/CN2022/097244 2022-01-07 2022-06-07 基于心脏超声视频筛选舒张期和收缩期图像的方法和装置 WO2023130662A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202210018424.5A CN114419500A (zh) 2022-01-07 2022-01-07 基于心脏超声视频筛选舒张期和收缩期图像的方法和装置
CN202210018424.5 2022-01-07

Publications (1)

Publication Number Publication Date
WO2023130662A1 true WO2023130662A1 (zh) 2023-07-13

Family

ID=81271104

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/097244 WO2023130662A1 (zh) 2022-01-07 2022-06-07 基于心脏超声视频筛选舒张期和收缩期图像的方法和装置

Country Status (2)

Country Link
CN (1) CN114419500A (zh)
WO (1) WO2023130662A1 (zh)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118013257A (zh) * 2024-04-07 2024-05-10 一网互通(北京)科技有限公司 基于数据序列的峰值查找方法、装置及电子设备
CN118013257B (zh) * 2024-04-07 2024-06-07 一网互通(北京)科技有限公司 基于数据序列的峰值查找方法、装置及电子设备

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114419500A (zh) * 2022-01-07 2022-04-29 乐普（北京）医疗器械股份有限公司 基于心脏超声视频筛选舒张期和收缩期图像的方法和装置
CN114898882B (zh) * 2022-06-21 2023-04-18 四川大学华西医院 基于超声对右心功能进行评估的方法和系统
CN116019435A (zh) * 2022-12-27 2023-04-28 北京镁伽机器人科技有限公司 一种类心脏的心率确定方法、装置、电子设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090028412A1 (en) * 2007-07-27 2009-01-29 Siemens Medical Solutions Usa, Inc. System and Method for Automatic Detection of End of Diastole and End of Systole Image Frames in X-Ray Ventricular Angiography
CN105637360A (zh) * 2013-10-17 2016-06-01 株式会社岛津制作所 波形中的峰值端点检测方法及检测装置
CN106175742A (zh) * 2016-07-19 2016-12-07 北京心量科技有限公司 一种心脏体征获取方法以及装置
CN112381895A (zh) * 2020-10-19 2021-02-19 深圳蓝韵医学影像有限公司 一种心脏射血分数计算方法与装置
CN114419500A (zh) * 2022-01-07 2022-04-29 乐普（北京）医疗器械股份有限公司 基于心脏超声视频筛选舒张期和收缩期图像的方法和装置

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20090028412A1 (en) * 2007-07-27 2009-01-29 Siemens Medical Solutions Usa, Inc. System and Method for Automatic Detection of End of Diastole and End of Systole Image Frames in X-Ray Ventricular Angiography
CN105637360A (zh) * 2013-10-17 2016-06-01 株式会社岛津制作所 波形中的峰值端点检测方法及检测装置
CN106175742A (zh) * 2016-07-19 2016-12-07 北京心量科技有限公司 一种心脏体征获取方法以及装置
CN112381895A (zh) * 2020-10-19 2021-02-19 深圳蓝韵医学影像有限公司 一种心脏射血分数计算方法与装置
CN114419500A (zh) * 2022-01-07 2022-04-29 乐普（北京）医疗器械股份有限公司 基于心脏超声视频筛选舒张期和收缩期图像的方法和装置

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118013257A (zh) * 2024-04-07 2024-05-10 一网互通(北京)科技有限公司 基于数据序列的峰值查找方法、装置及电子设备
CN118013257B (zh) * 2024-04-07 2024-06-07 一网互通(北京)科技有限公司 基于数据序列的峰值查找方法、装置及电子设备

Also Published As

Publication number Publication date
CN114419500A (zh) 2022-04-29

Similar Documents

Publication Publication Date Title
WO2023130662A1 (zh) 基于心脏超声视频筛选舒张期和收缩期图像的方法和装置
WO2021082272A1 (zh) 心动周期的确定方法及超声设备
CN110232383B (zh) 一种基于深度学习模型的病灶图像识别方法及病灶图像识别系统
CN108133476A (zh) 一种肺结节自动检测方法及系统
WO2021109987A1 (zh) 电子装置及非暂态计算机可读存储介质
WO2023130661A1 (zh) 一种二维频谱多普勒超声心动图像的处理方法和装置
CN111462049B (zh) 一种乳腺超声造影视频中病灶区形态自动标注方法
WO2022077888A1 (zh) 一种光体积描计信号的散点图分类方法和装置
CN105590324A (zh) 医学图像的分割方法及其装置
CN112381895A (zh) 一种心脏射血分数计算方法与装置
CN114121232A (zh) 分类模型训练方法、装置、电子设备及存储介质
CN111738992A (zh) 肺部病灶区域提取方法、装置、电子设备和存储介质
CN111639555B (zh) 一种指静脉图像噪声精确提取和自适应滤波去噪方法及装置
WO2024051018A1 (zh) 一种pet参数图像的增强方法、装置、设备及存储介质
CN114419499A (zh) 一种基于心脏超声视频的射血分数数据处理方法和装置
CN115661187A (zh) 用于中药制剂分析的图像增强方法
CN109003264B (zh) 一种视网膜病变图像类型识别方法、装置和存储介质
CN115775233A (zh) 一种基于心脏超声视频测量特征尺寸的处理方法和装置
CN116130090A (zh) 射血分数测量方法和装置、电子设备和存储介质
WO2022089266A1 (zh) 血管管腔提取方法、装置、电子设备和存储介质
CN111466894A (zh) 一种基于深度学习的射血分数计算方法及系统
CN112365474A (zh) 血管提取方法、装置、电子设备和存储介质
CN112258533A (zh) 一种超声图像中小脑蚓部的分割方法
Zheng et al. 3D consistent biventricular myocardial segmentation using deep learning for mesh generation
Yan et al. Automatic detection and localization of pulmonary nodules in ct images based on yolov5

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: 22918118

Country of ref document: EP

Kind code of ref document: A1