CN-121998979-A - Quality improvement method and system for hysteroscope operation image data

CN121998979ACN 121998979 ACN121998979 ACN 121998979ACN-121998979-A

Abstract

The application provides a quality improvement method and a quality improvement system for hysteroscopic operation image data, which relate to the technical field of image processing, and the method comprises the steps of acquiring hysteroscopic images of a target patient at a specific time point and acquiring multi-mode physiological time sequence data of the target patient before and after the specific time point; the method comprises the steps of carrying out time sequence feature fusion analysis on multi-mode physiological time sequence data to obtain individual physiological cycle phase information corresponding to a specific time point, inputting the individual physiological cycle phase information into a pre-trained healthy uterine cavity image to generate an intelligent body, outputting to obtain a standard healthy uterine cavity reference image corresponding to the specific time point, comparing a hysteroscope image with the standard healthy uterine cavity reference image to generate a pathological residual feature map, and carrying out enhanced display on the pathological residual feature map. Solves the technical problem that the hysteroscope image has unobvious abnormal characteristics in the prior art.

Inventors

LIU JIN
HE XIAO
XIAO XIFENG

Assignees

中国人民解放军空军军医大学

Dates

Publication Date: 20260508
Application Date: 20260409

Claims (10)

1. A method for improving quality of hysteroscopic image data, the method comprising: Acquiring hysteroscope images of a target patient at a specific time point, and acquiring multi-mode physiological time sequence data of the target patient before and after the specific time point; performing time sequence feature fusion analysis on the multi-mode physiological time sequence data to obtain individual physiological cycle phase information corresponding to the specific time point; inputting the individual physiological cycle phase information into a pre-trained healthy uterine cavity image to generate an intelligent body, and outputting to obtain a standard healthy uterine cavity reference image corresponding to the specific time point; And comparing the hysteroscope image with the standard healthy uterine cavity reference image, generating a pathology residual feature map, and performing enhanced display on the pathology residual feature map.
2. The method of claim 1, wherein the multi-modal physiological time series data includes a continuously monitored base body temperature sequence, serum hormone level detection data, endometrial morphology feature time series data extracted from pelvic ultrasound images, and menstrual history information of electronic medical records.
3. The quality improvement method of hysteroscopic surgical image data according to claim 1, wherein performing a temporal feature fusion analysis on the multi-modal physiological temporal data to obtain individual physiological cycle phase information corresponding to the specific time point includes: calculating menstrual cycle regularity scores of the target patient based on menstrual history information of electronic medical records in the multi-modal physiological time series data; obtaining a pre-constructed phase estimation submodule array, wherein the phase estimation submodule array comprises M phase estimation submodules; multiplying and rounding the menstrual cycle regularity score with M to obtain the calling number N of the phase estimation submodule, wherein N is a positive integer between 1 and M; Randomly selecting N phase estimation submodules from the phase estimation submodule array, respectively processing the multi-mode physiological time sequence data by the N selected phase estimation submodules, and correspondingly outputting N pieces of intermediate phase information; And carrying out fusion calculation on the N pieces of intermediate phase information to generate individual physiological cycle phase information.
4. The method of claim 3, wherein calculating a menstrual cycle regularity score for the target patient based on menstrual history information recorded by electronic medical records in the multi-modal physiological time series data comprises: Extracting a continuous historical menstrual cycle length sequence of the target patient from menstrual history information recorded by the electronic medical record; Calculating standard deviation and average value of the historical menstrual cycle length sequence, and calculating the ratio of the standard deviation to the average value to obtain a cycle length variation coefficient; Calculating the ratio of the sum of the absolute values of the change rates of each adjacent cycle in the historical menstrual cycle length sequence to the total number of adjacent cycles to obtain the average absolute cycle change rate; dividing the period length variation coefficient by a first preset clinical threshold to obtain a basic fluctuation sub-score, wherein the value range of the basic fluctuation sub-score is 0-1; dividing the average absolute period change rate by a second preset clinical threshold to obtain a severe change sub-score, wherein the value range of the severe change sub-score is 0-1; and calculating the arithmetic average value of the basic fluctuation sub-score and the severe variation sub-score to obtain a menstrual cycle regularity score, wherein the numerical value of the menstrual cycle regularity score is positively correlated with the irregular degree of menstrual cycle.
5. The method for improving the quality of hysteroscopic image data as claimed in claim 4, wherein the determining process of the first preset clinical threshold and the second preset clinical threshold comprises: Collecting a menstrual cycle data set comprising a plurality of samples, wherein each sample comprises at least menstrual history information, age and body mass index; dividing samples in the menstrual cycle data set into a plurality of group subsets by taking age segmentation and body mass index classification as joint classification basis, wherein the age segmentation at least comprises puberty, childbirth age and perimenopause, and the body mass index classification at least comprises a lean range, a normal range, an overweight range and an obese range; Calculating an arithmetic average value of the cycle length variation coefficients of all samples in each group subset as a first clinical reference threshold corresponding to each group subset; calculating an arithmetic mean of the average absolute period change rates of all samples in each group subset as a second clinical reference threshold corresponding to each group subset; Constructing a threshold mapping relation library for storing mapping relation between each group subset and the corresponding first clinical reference threshold and second clinical reference threshold; Determining a group subset which belongs to the target patient according to the age and body quality index of the target patient, and inquiring from the threshold mapping relation library to obtain a corresponding first clinical reference threshold and a corresponding second clinical reference threshold which are respectively used as a first preset clinical threshold and a second preset clinical threshold which are matched with the target patient.
6. The quality improvement method for hysteroscopic image data according to claim 3, wherein performing fusion calculation on the N pieces of intermediate phase information to generate individual physiological cycle phase information includes: calculating an arithmetic average value of the N pieces of intermediate phase information as an average phase vector; And inputting the average phase vector to a pre-trained phase decoder, and mapping to obtain individual physiological cycle phase information, wherein the individual physiological cycle phase information is a scalar value used for quantitatively representing the position of a specific time point in the individual menstrual cycle of the target patient.
7. The quality improvement method of hysteroscopic image data as claimed in claim 1, wherein the construction process of the healthy uterine cavity image generating agent comprises: Collecting hysteroscopic images acquired by a plurality of healthy volunteers at a plurality of preset time points in a complete physiological cycle to form a training image set; aiming at each hysteroscope image in the training image set, calculating corresponding individual physiological cycle phase information according to multi-mode physiological time sequence data synchronously recorded during acquisition, and taking the individual physiological cycle phase information as a label of the hysteroscope image; Constructing a condition generation countermeasure network comprising a generator and a discriminator; Inputting the individual physiological cycle phase information to the generator under the condition that the individual physiological cycle phase information is used as a condition, inputting the corresponding hysteroscope image as a real sample to the discriminator, and enabling the generator to learn the mapping relation from the individual physiological cycle phase information to the healthy hysteroscope image under the individual physiological cycle phase information through countermeasure training; After training is completed, the generator in the condition generating countermeasure network is defined as a healthy uterine cavity image generating agent.
8. The method of claim 1, wherein comparing the hysteroscopic image with the standard healthy uterine reference image to generate a pathology residual feature map that enhances display of areas of the hysteroscopic image that deviate from normal physiological structures comprises: Adopting an encoder-decoder network structure, sharing part of network layer parameters with a generator in the healthy uterine cavity image generation intelligent agent, and constructing a pathological feature decoupler; Training the pathological feature decoupler by taking a reconstruction loss function and an contrast difference loss function as joint constraint, so as to obtain a trained pathological feature decoupler, wherein the reconstruction loss function is used for constraining that the hysteroscope image can be reconstructed after pixel-level addition of the standard healthy uterine cavity reference image and a pathological residual feature image to be generated, and the contrast difference loss function is used for constraining that the distance between the pathological residual feature image and residual feature distribution of a group of real pathological samples in a feature space is minimized; And the trained pathological feature decoupler takes the standard healthy uterine cavity reference image as a health standard, performs feature decoupling calculation on the hysteroscope image, separates out cycle-related features and cycle-independent residual features, and maps the cycle-independent residual features to generate a pathological residual feature map.
9. The quality improvement method for hysteroscopic image data as claimed in claim 1, wherein said pathology residual feature map is enhanced displayed, comprising: performing pseudo-color mapping on the pathological residual feature map to generate a pseudo-color enhancement map; performing spatial registration and pixel level fusion on the pseudo-color enhancement map and the hysteroscope image to generate a difference enhancement display image; And carrying out morphological analysis on the high-response area in the difference enhanced display image, extracting boundary outline and area parameters of the high-response area, and superposing and displaying the boundary outline and the area parameters on the difference enhanced display image in a graphical labeling mode.
10. A quality enhancement system for hysteroscopic image data, characterized by being adapted to perform a quality enhancement method for hysteroscopic image data as claimed in any one of claims 1 to 9, comprising: the data acquisition module is used for acquiring hysteroscope images of a target patient at a specific time point and acquiring multi-mode physiological time sequence data of the target patient before and after the specific time point; The phase information calculation module is used for carrying out time sequence feature fusion analysis on the multi-mode physiological time sequence data to obtain individual physiological cycle phase information corresponding to the specific time point; the standard image generation module is used for inputting the physiological cycle phase information of the individual to a pre-trained healthy uterine cavity image generation agent and outputting a standard healthy uterine cavity reference image corresponding to the specific time point; the contrast enhancement module is used for comparing the hysteroscope image with the standard healthy uterine cavity reference image, generating a pathology residual characteristic diagram and enhancing and displaying the pathology residual characteristic diagram.

Description

Quality improvement method and system for hysteroscope operation image data Technical Field The invention relates to the technical field of image processing, in particular to a quality improvement method and system for hysteroscopic operation image data. Background In the clinical diagnosis and treatment process of hysteroscopic surgery, hysteroscopic images are core image bases for doctors to carry out hysteroscopic lesion diagnosis and surgery operation, but the image quality of the hysteroscopic images is obviously limited, and the practical requirements of clinical accurate diagnosis and treatment are difficult to meet. The method is characterized in that the normal physiological form of the female uterine cavity can generate dynamic physiological change along with the menstrual cycle of an individual, the physiological change is easily confused with pathological abnormal characteristics, so that a static hysteroscopic image cannot accurately distinguish a normal physiological structure from a pathological abnormal region, and meanwhile, the visual recognition degree of a tiny abnormal region in the hysteroscopic image is low, so that the problem that the abnormal characteristics of the hysteroscopic image are not outstanding is solved. Disclosure of Invention The invention provides a quality improvement method and a system for hysteroscope operation image data, aiming at the technical problems of obvious physiological interference, fuzzy abnormal characteristics, insufficient effective detail identification and poor overall visual quality of hysteroscope images caused by the fact that the normal physiological form of a female hysteroscope cavity can generate dynamic physiological change along with the menstrual cycle of an individual and the visual identification of an abnormal area is low in an original hysteroscope operation image in the prior art. The technical scheme for solving the technical problems is as follows: in a first aspect, the present invention provides a quality improvement method for hysteroscopic image data, including: Acquiring hysteroscope images of a target patient at a specific time point, and acquiring multi-mode physiological time sequence data of the target patient before and after the specific time point; performing time sequence feature fusion analysis on the multi-mode physiological time sequence data to obtain individual physiological cycle phase information corresponding to the specific time point; inputting the individual physiological cycle phase information into a pre-trained healthy uterine cavity image to generate an intelligent body, and outputting to obtain a standard healthy uterine cavity reference image corresponding to the specific time point; And comparing the hysteroscope image with the standard healthy uterine cavity reference image, generating a pathology residual feature map, and performing enhanced display on the pathology residual feature map. In a second aspect, the present invention provides a quality enhancement system for hysteroscopic surgical image data, comprising: the data acquisition module is used for acquiring hysteroscope images of a target patient at a specific time point and acquiring multi-mode physiological time sequence data of the target patient before and after the specific time point; The phase information calculation module is used for carrying out time sequence feature fusion analysis on the multi-mode physiological time sequence data to obtain individual physiological cycle phase information corresponding to the specific time point; the standard image generation module is used for inputting the physiological cycle phase information of the individual to a pre-trained healthy uterine cavity image generation agent and outputting a standard healthy uterine cavity reference image corresponding to the specific time point; the contrast enhancement module is used for comparing the hysteroscope image with the standard healthy uterine cavity reference image, generating a pathology residual characteristic diagram and enhancing and displaying the pathology residual characteristic diagram. The beneficial effects of the invention are as follows: Compared with the prior art, the method comprises the steps of firstly, providing complete and relevant basic data support for subsequent image quality improvement processing by acquiring hysteroscope images of a target patient at a specific time point and multi-mode physiological time sequence data before and after the specific time point, ensuring pertinence and accuracy of the subsequent processing, secondly, carrying out time sequence feature fusion analysis on the multi-mode physiological time sequence data to obtain individual physiological cycle phase information, accurately capturing the current physiological state of the uterine cavity of the target patient, providing a basis for avoiding image interference caused by a physiological cycle, thirdly, inputting the individual physiological cycle ph