Ovary follicle positioning and puncture path planning system based on multi-modal image fusion and ai

Abstract

The application discloses a kind of follicle positioning and puncture path planning system based on multimodal image fusion and AI, it is related to intelligent medical technology field, to solve the technical problem that existing technology does not integrate chromosome risk, blood vessel endothelium safety and blood perfusion evaluation, leading to high quality follicle screening difficult, complication high incidence, including: multimodal image-serological fusion module, obtain the multimodal image data and serological marker data of target patient, and generate comprehensive feature set by fusion;AI three-dimensional evaluation module, based on comprehensive feature set, follicle is jointly evaluated based on chromosome risk, blood perfusion state and blood vessel endothelium safety state.The application solves the core technical problem that high-quality ovum screening is difficult, improves the clinical available embryo rate, shortens the patient treatment cycle, reduces the physical damage and economic burden brought by repeated ovum extraction, and provides technical support for assisted reproductive treatment of elderly patients with polycystic ovary and vascular lesions.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of intelligent medical treatment, in particular to an ovulation follicle positioning and puncture path planning system based on multi-modal image fusion and AI. BACKGROUND

[0002] In the field of assisted reproduction, the treatment of egg retrieval for patients with advanced age and polycystic ovaries combined with vascular lesions has always been a clinical problem. Due to physiological decline, such patients not only face the problem of significantly increased risk of chromosomal aneuploidy of follicles, but also have characteristics of abnormal ovarian vascular endothelial function and uneven blood perfusion distribution, which superimpose on each other to cause the three dilemmas of "difficult selection of high-quality oocytes, difficult safety protection of blood vessels, and difficult guarantee of oocyte activity" in the process of egg retrieval.

[0003] Existing egg retrieval path planning systems mainly focus on follicle positioning or simple blood vessel avoidance based on single ultrasound image, and do not integrate chromosomal abnormality risk assessment, blood vessel endothelial safety state and follicle blood perfusion state. The quality of the follicle is determined by its morphological characteristics, which cannot accurately identify high-quality follicles at the genetic level, and lacks targeted protection strategies for blood vessel endothelial fragility. Moreover, the influence of blood perfusion state on oocyte activity is not considered, which leads to problems such as "low usable embryo rate with standard egg number" and "high incidence of blood vessel damage complications after puncture" in clinical practice, seriously affecting treatment effectiveness and patient safety, and becoming a key technical bottleneck restricting the success rate of assisted reproduction for such patients. Therefore, we propose an ovulation follicle positioning and puncture path planning system based on multi-modal image fusion and AI. SUMMARY

[0004] The application aims to provide an ovulation follicle positioning and puncture path planning system based on multi-modal image fusion and AI to solve the technical problems of the prior art that do not integrate chromosomal risk, blood vessel endothelial safety and blood perfusion assessment, leading to difficult selection of high-quality follicles and high incidence of complications.

[0005] To solve the above technical problems, the application provides the following technical solution: an ovulation follicle positioning and puncture path planning system based on multi-modal image fusion and AI, comprising:

[0006] A multi-modal image-serological fusion module acquires multi-modal image data and serological marker data of a target patient, and generates a comprehensive feature set by fusion;

[0007] An AI three-dimensional evaluation module performs joint evaluation of chromosomal risk, blood perfusion state and blood vessel endothelial safety state of the follicle based on the comprehensive feature set, and divides the puncture priority;

[0008] A multi-target dynamic path planning module, aiming at maximizing the high-quality low-risk follicle acquisition rate, minimizing the risk of vascular endothelial damage and adapting to the blood perfusion state, generates an initial puncture path;

[0009] An intraoperative dynamic correction module for real-time data updating and path adjustment to compensate for deviations.

[0010] A puncture parameter adaptive module dynamically adjusts the puncture execution parameters according to the vascular endothelium and follicle perfusion state.

[0011] The present application breaks through the limitations of the prior art relying only on morphological evaluation, and the system can accurately identify high-quality follicles with low chromosomal abnormality risk and high activity from the genetic level, while selecting follicles with sufficient oxygen supply based on blood perfusion state to avoid ineffective puncture of high-risk or hypoxic follicles. The core technical problem of difficult high-quality ovum screening is solved, which significantly improves the clinical usable embryo rate, effectively shortens the treatment cycle of patients, reduces the physical damage and economic burden caused by repeated ovum collection, and provides core technical support for assisted reproductive treatment of elderly patients with polycystic ovary and vascular disease.

[0012] Preferably, the multi-modal image-serological fusion module comprises:

[0013] An image data acquisition unit acquires transvaginal three-dimensional ultrasound contrast data, MRI diffusion weighted imaging data, MRI dynamic contrast enhancement data and ultrasound elastography data, and extracts blood flow distribution features, follicle cell activity features, blood perfusion features and vascular endothelium surrounding tissue mechanical features, respectively.

[0014] The serological marker data includes homocysteine, vascular endothelial growth factor and antithrombin III related data.

[0015] A feature fusion unit uses a feature correlation mapping algorithm to correlate image features and serological marker data, and constructs a comprehensive feature set containing follicle quality features, blood perfusion features, vascular endothelium safety features and serum correlation features.

[0016] The comprehensive feature set includes a fusion feature reflecting the correlation between follicle hypoxia state and vascular disease.

[0017] Preferably, the AI three-dimensional evaluation module comprises:

[0018] A hierarchical attention model dynamically adjusts the weight proportion of image features and serological features according to the target patient age stratification, and realizes age-adaptive feature weighting.

[0019] The weight adjustment logic of the hierarchical attention model is: for older patients in a lower age range, the weight proportion of follicular morphological features and granulosa cell activity features is increased; for older patients in a higher age range, the weight proportion of serological marker data and blood perfusion features is increased.

[0020] The two-dimensional classification matrix construction unit, through joint fitting of multi-modal image features and serological features, deduces the correlation between follicular hypoxic state and chromosomal abnormality risk, and divides the follicles into four categories: low-risk low-hypoxia, low-risk high-hypoxia, high-risk low-hypoxia, and high-risk high-hypoxia, corresponding to different puncture priorities and timing, and the correlation deduction formula is:

[0021] ;

[0022] In the formula, represents the follicular chromosomal abnormality probability; represents the follicular hypoxic index, wherein represents the apparent diffusion coefficient of MRI diffusion weighted imaging, represents the blood perfusion parameter of MRI dynamic contrast enhancement, 、 represents the feature weight coefficient; represents the serological correlation feature, wherein represents the homocysteine concentration, represents the vascular endothelial growth factor concentration, 、 represents the serum feature weight coefficient; 、 、 represents the fitting coefficient;

[0023] The safety threshold determination unit sets safety determination criteria according to the vascular endothelial peripheral tissue mechanical characteristics extracted by ultrasonic elastography, and down-regulates the puncture priority of the follicles that exceed the criteria.

[0024] Preferably, the hierarchical attention model uses an adaptive weighting algorithm to calculate the weights:

[0025] When the target patient is an older patient in a lower age range, the weight formula is: ;

[0026] When the target patient is an older patient in a higher age range, : ;

[0027] In the formula, represents the feature comprehensive weight of the older patient in the lower age range; represents the feature comprehensive weight of the older patient in the higher age range; represents the image feature-based weight; represents the serological feature-based weight; represents the follicle chromosomal abnormality probability; represents the weight adjustment coefficient, wherein

[0028] Preferably, the multi-objective dynamic path planning module adopts a multi-objective optimization algorithm improved based on a non-dominated sorting genetic algorithm to construct a multi-objective function:

[0029]

[0030] wherein is a spatial coordinate vector of the puncture path is a high-quality low-risk follicle acquisition rate function, wherein represents the total number of follicles in the ovary, represents the weight value of the th follicle, represents the spatial coverage area of the th follicle, represents the chromosomal abnormality probability of the th follicle; is a vascular endothelial abnormal area length function, wherein represents the vascular endothelial abnormality state indication function corresponding to a certain point on the path; is a perfusion adaptation function, wherein represents the local follicle blood flow perfusion value corresponding to a certain point on the path, represents the minimum threshold value of the follicle high perfusion area;

[0031] The multi-objective dynamic path planning module reconstructs the three-dimensional blood flow perfusion atlas and the blood vessel distribution atlas of the ovary based on the multi-modal image, and searches for the optimal initial puncture path based on the multi-objective function.

[0032] Preferably, the path generation constraint conditions include that the shortest distance between the puncture path and the vascular endothelium meets the safety requirements, the path preferentially passes through the follicle high perfusion area, and avoids the small blood vessel branch dense area;

[0033] ​​​​​​​The multi-target dynamic path planning module constructs an integrated three-dimensional model of the ovarian anatomical structure and blood perfusion through three-dimensional reconstruction technology, and performs path search and optimization in the model combined with constraint conditions, to ensure that the initial puncture path meets the actual needs of clinical puncture operation and adapts to the physiological state of the patient's ovary.

[0034] Preferably, the intraoperative dynamic correction module comprises:

[0035] A real-time data updating unit is configured to continuously acquire intraoperative real-time ultrasound contrast data and ultrasound elastography data, and dynamically update follicle blood perfusion state data and ovarian spatial position data.

[0036] A path adjustment unit is configured to calculate a puncture angle adjustment amount based on the updated real-time data, and dynamically adjust the puncture angle and path trajectory by using an intelligent optimization algorithm to compensate for path deviation caused by physiological movement or blood flow fluctuation of the patient.

[0037] The dynamic adjustment amount is calculated as follows:

[0038] ;

[0039] In the formula: represents the puncture angle adjustment amount; represents an angle adjustment coefficient of the blood perfusion difference value; represents an angle adjustment coefficient of the displacement amount; represents a difference between the real-time blood perfusion value and the initial perfusion value, wherein represents a real-time updated blood perfusion parameter, represents a blood perfusion parameter at the initial path planning; represents a relative change amount of blood perfusion; represents a displacement amount of the real-time spatial position of the ovary and the initial position, wherein , , represents displacement components of the ovary in the three-dimensional space, corresponding to the displacement in the x, y, and z axis directions, respectively; represents an ovary standard displacement reference value; represents a relative change amount of displacement; represents an angle conversion coefficient.

[0040] Preferably, the puncture parameter adaptive module comprises:

[0041] A parameter adjustment rule library is configured to pre-establish a corresponding relationship among the vascular endothelial state, the follicle perfusion state, and the puncture parameters, including the puncture needle advancing speed, the needle tip pressure, and the puncture suction speed.

[0042] The real-time control unit dynamically adjusts the puncture execution parameters according to the blood vessel endothelium safety state result output by the AI three-dimensional evaluation module and the updated follicle perfusion state data in the operation, so as to avoid blood vessel endothelium injury and reduce the exposure time of oocytes in an adverse perfusion environment.

[0043] Preferably, the adaptive adjustment formula of the puncture parameters is as follows:

[0044] Puncture needle advancing speed: ;

[0045] Needle tip pressure: ;

[0046] Puncture suction speed: ;

[0047] In the formula, represents the puncture needle advancing speed; represents the basic puncture needle advancing speed; represents the elastic modulus of the tissue around the blood vessel endothelium; represents the blood vessel endothelium safety elastic modulus threshold value; represents the elastic modulus corresponding to the limit fragile state of the blood vessel endothelium; represents the speed adjustment coefficient, ; represents the needle tip pressure; represents the basic needle tip pressure; represents the upper limit of the safety pressure; represents the puncture suction speed; represents the basic suction speed; represents the suction speed adjustment coefficient, ; represents the normal perfusion threshold value; represents the real-time follicle perfusion value;

[0048] When the blood vessel endothelium is in a fragile state, the puncture needle advancing speed is reduced and the needle tip pressure is controlled in a safety range; when the follicle is in a mild hypoxic state, the puncture suction speed is increased to guarantee the activity of oocytes.

[0049] Preferably, the system further comprises a data storage and traceability module, which is used for storing multi-modal image data, serological marker data, AI three-dimensional evaluation results, puncture path planning data, intraoperative real-time update data, puncture parameter adjustment records and puncture execution data, and forms a complete diagnosis and treatment data chain.

[0050] Compared with the prior art, the present application has the following beneficial effects:

[0051] 1、The application breaks through the limitation of the prior art relying only on morphological evaluation by fusing multi-modal images and an AI three-dimensional evaluation module, deeply correlating multi-modal image features and serological marker data, constructing a quantitative correlation model of follicle hypoxic state and chromosome abnormality risk, and accurately identifying low chromosome abnormality risk and high activity high-quality follicles from the genetic level, while screening follicles with sufficient oxygen supply in combination with blood perfusion state to avoid ineffective puncture of high-risk or hypoxic follicles. The core technical problem of high-quality oocyte screening difficulty is solved, the clinical usable embryo rate is significantly improved, the patient treatment cycle is effectively shortened, the physical damage and economic burden caused by repeated oocyte retrieval are reduced, and core technical support is provided for assisted reproductive treatment of elderly patients with polycystic ovary combined with vascular lesions.

[0052] 2、The application is directed to the characteristics of fragile vascular endothelium and easy damage of patients with polycystic ovary combined with vascular lesions, and the application realizes active protection of the vascular endothelium by multi-target dynamic path planning and adaptive adjustment of puncture parameters. The path planning module not only avoids blood vessel dense areas when generating a puncture path, but also identifies fragile areas of the vascular endothelium through ultrasonic elastography data to ensure a safe distance between the path and the vascular endothelium. The puncture parameter adaptive module dynamically adjusts the puncture needle advancement speed and needle tip pressure according to the state of the vascular endothelium to avoid damage to the vascular endothelium caused by improper operation. On the basis of solving the core problem, the pain point of blood vessel safety protection difficulty is further overcome, the incidence of complications such as post-puncture bleeding, hematoma and vascular endothelial damage is greatly reduced, and the treatment safety of patients is ensured, leaving good ovarian reserve function for subsequent treatment.

[0053] 3、The application also realizes personalized evaluation of patients of different ages by a hierarchical attention model, dynamically adjusts the feature weight proportion according to the physiological characteristics of younger and older patients, and combines an intraoperative dynamic correction module to capture real-time changes in ovarian displacement and blood perfusion, continuously optimize the puncture path and operation parameters, further break through the limitations of the prior art, realize the precise adaptation of the treatment plan to the physiological state of the individual patient, and effectively avoid evaluation bias and operation errors caused by individual differences in follicle quality evaluation, path planning and puncture operation, improve the precision and reliability of treatment, and provide a more personalized oocyte retrieval solution for patients with polycystic ovary combined with vascular lesions. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 The figure is a schematic diagram of the system architecture of the application;

[0055] Figure 2 The figure is a schematic diagram of the implementation process of Example Two in the application. DETAILED DESCRIPTION

[0056] Embodiment one: as shown, the application relates to a kind of follicle positioning and puncture path planning system based on multimodal image fusion and AI, comprising: multimodal image-serological fusion module, AI three-dimensional evaluation module, multi-target dynamic path planning module, intraoperative dynamic correction module and puncture parameter adaptive module; Figure 1

[0057] The multimodal image-serological fusion module obtains the multimodal image data and serological marker data of the target patient, and generates a comprehensive feature set by fusion;

[0058] In the embodiment of the application, the multimodal image-serological fusion module includes an image data acquisition unit and a feature fusion unit.

[0059] The image data acquisition unit obtains transvaginal three-dimensional ultrasound contrast data, MRI diffusion weighted imaging data, MRI dynamic contrast enhancement data and ultrasound elastography data, and extracts blood flow distribution features, follicle cell activity features, blood perfusion features and vascular endothelial tissue mechanical features, respectively.

[0060] Among them, the transvaginal three-dimensional ultrasound contrast data obtained by the image data acquisition unit is used to evaluate the distribution of ovarian artery blood flow and the situation of microvessel branches; the MRI diffusion weighted imaging data is used to evaluate the activity of follicular granulosa cells; the MRI dynamic contrast enhancement data is used to quantify the local blood perfusion level of follicle; the ultrasound elastography data is used to detect the mechanical properties of the tissue around the vascular endothelium, including the uniformity of tissue elasticity and the distribution of elastic state.

[0061] The serological marker data includes homocysteine, vascular endothelial growth factor and antithrombin III related data.

[0062] The feature fusion unit associates and maps the image features and serological marker data, constructs a comprehensive feature set containing follicle quality features, blood perfusion features, vascular endothelial safety features and serum correlation features, and the comprehensive feature set contains fusion features reflecting the correlation between follicle hypoxic state and vascular lesions.

[0063] Among them, the feature fusion unit uses a feature association mapping algorithm to associate the vascular lesion related state reflected by the serological marker data with the follicle and ovarian tissue features extracted from the multimodal image, to generate a fusion feature that can comprehensively reflect the follicle quality, blood flow environment and vascular safety state, and provide comprehensive feature input for the AI three-dimensional evaluation module.

[0064] ​The AI three-dimensional evaluation module performs joint evaluation on the chromosome risk, blood perfusion state and vascular endothelial safety state of the follicle based on the comprehensive feature set, and divides the follicle puncture priority;

[0065] In the embodiment of the application, the AI three-dimensional evaluation module comprises a hierarchical attention model, a two-dimensional classification matrix construction unit and a safety threshold determination unit;

[0066] The hierarchical attention model dynamically adjusts the weight proportion of image features and serological features according to the age stratification of the target patient, and realizes age-adaptive feature weighting.

[0067] The weight adjustment logic of the hierarchical attention model is: for the elderly patient in the lower age group, the weight proportion of follicle morphological features and granulosa cell activity features is increased; for the elderly patient in the higher age group, the weight proportion of serological marker data and blood perfusion features is increased, so that the evaluation model adapts to the physiological characteristics of the elderly patients in different age groups.

[0068] In the embodiment of the application, the weight calculation of the hierarchical attention model adopts an adaptive weighting algorithm, and the specific formula is as follows:

[0069] When the target patient is an elderly patient in the lower age group, the weight calculation formula is:

[0070] ;

[0071] When the target patient is an elderly patient in the higher age group, the weight calculation formula is:

[0072] ;

[0073] In the formula:

[0074] represents the feature comprehensive weight of the elderly patient in the lower age group;

[0075] represents the feature comprehensive weight of the elderly patient in the higher age group;

[0076] represents the image feature basic weight, reflecting the basic contribution of image data to follicle evaluation;

[0077] represents the serological feature basic weight, reflecting the basic contribution of serological marker data to follicle evaluation;

[0078] represents the follicle chromosome abnormality probability, which is derived from the formula and reflects the quality risk of the follicle at the genetic level;

[0079] , 、 、 represents a weight adjustment coefficient, used to adapt to the physiological feature differences of patients in different age groups, wherein 、 , the dynamic adjustment of the feature weights of different age groups is realized through the difference of the coefficients.

[0080] Operation logic: for elderly patients in different age groups, an adaptive weighting algorithm is used to dynamically allocate the weight proportion of image features and serological features, and the probability of follicle chromosomal abnormality is integrated to construct an age-adapted comprehensive weight, realize the feature weighting matching the physiological characteristics of patients, and provide accurate weight basis for subsequent follicle priority division. Through the adaptive weight calculation of age stratification, the evaluation model can accurately match the physiological characteristics of elderly patients in different age groups, while integrating the chromosomal abnormality probability association, improving the relevance of feature weight and follicle actual quality and safety state, providing more targeted basis for subsequent follicle priority division, and ensuring the accuracy and individualization of the evaluation results.

[0081] The two-dimensional classification matrix construction unit divides the follicles into different puncture priority categories based on the chromosomal risk assessment results and the follicle hypoxia degree assessment results;

[0082] In the embodiments of the present application, the two-dimensional classification matrix construction unit derives the correlation between the follicle hypoxia state and the chromosomal abnormality risk through the joint fitting of multi-modal image features and serological features, and divides the follicle categories based on the correlation, including low-risk low-hypoxia follicles, low-risk high-hypoxia follicles, high-risk low-hypoxia follicles and high-risk high-hypoxia follicles. Different categories of follicles correspond to different puncture priorities and puncture timing suggestions.

[0083] The correlation derivation formula of the follicle hypoxia state and the chromosomal abnormality risk is as follows:

[0084] ;

[0085] In the formula:

[0086] represents the probability of follicle chromosomal abnormality, the value range is 0-1, the larger the value, the higher the risk of follicle chromosomal abnormality;

[0087] represents the Sigmoid activation function, used to map the linear combination result to the 0-1 interval to realize probabilistic output;

[0088] represents the follicle hypoxia index, which comprehensively reflects the oxygen supply state of the follicle and is calculated from the multi-modal image features, , wherein apparent diffusion coefficient (ADC) representing diffusion weighted imaging of MRI, reflecting the activity state of follicular granulosa cells, blood perfusion parameter representing dynamic contrast enhanced of MRI, reflecting the local blood perfusion level of follicle, 、 representing feature weight coefficient, used to adjust and contribution to the hypoxia index;

[0089] representing serological correlation feature, comprehensively reflecting the vascular lesion related state of the patient, calculated from serological marker data, wherein representing homocysteine concentration, one of the serological markers, related to the vascular lesion state, representing vascular endothelial growth factor concentration, one of the serological markers, related to the vascular endothelial function and blood flow state, 、 representing serum feature weight coefficient, used to adjust and contribution to the serological correlation feature;

[0090] 、 、 representing fitting coefficient, obtained by training clinical data, used to adjust the influence degree of hypoxia index and serological correlation feature on the probability of chromosomal abnormality.

[0091] Operation logic: construct follicle hypoxia index through multi-modal image features, construct serological correlation feature combined with serological marker data, use Sigmoid activation function to map the linear combination of the two to the probability of chromosomal abnormality between 0-1, realize the quantitative correlation of follicle hypoxia state and chromosomal abnormality risk, provide quality basis at gene level for follicle priority division, through deep fusion of multi-modal image features and serological features, accurately quantify the correlation between follicle hypoxia state and chromosomal abnormality risk, realize accurate evaluation of follicle quality at gene level, provide scientific basis for screening of high-quality low-risk follicles, and improve the accuracy and pertinence of follicle screening.

[0092] The safety threshold determination unit sets the vascular endothelium safety determination standard according to the mechanical characteristics of the tissue around the vascular endothelium extracted by ultrasonic elastography, and performs puncture priority down processing on the follicle corresponding to the area exceeding the safety standard.

[0093] The multi-target dynamic path planning module generates an initial puncture path with the goals of maximizing high-quality low-risk follicle acquisition rate, minimizing vascular endothelium damage risk, and adapting to blood perfusion state.

[0094] In the embodiment of the present application, the multi-target dynamic path planning module adopts an improved multi-target optimization algorithm to construct a multi-target function;

[0095] The improved multi-target optimization algorithm is improved based on a non-dominated sorting genetic algorithm, introduces ovary dissection structure constraints and physiological state constraints, optimizes a path search strategy, improves the generation efficiency and adaptability of an optimal puncture path, and ensures that the generated path meets the actual requirements of clinical puncture operations while meeting multi-target requirements.

[0096] The selection operator optimization formula of the improved multi-target optimization algorithm is:

[0097] ;

[0098] In the formula, f represents the selection probability of the path, and the greater the value, the more the path meets the multi-target optimization requirements and the higher the probability of being selected;

[0099] f represents the fitness function of the path, which comprehensively reflects the degree to which the path meets the multi-target optimization requirements, wherein f1, f2, and f3 represent the normalized results of the three sub-target functions, respectively, eliminate the dimensional differences of different target functions, and realize collaborative optimization,

[0100] f1, f2, and f3 represent the normalized results of the three sub-target functions, respectively, eliminate the dimensional differences of different target functions, and realize collaborative optimization, f1, f2, and f3 represent the normalized results of the three sub-target functions, respectively, eliminate the dimensional differences of different target functions, and realize collaborative optimization, f1, f2, and f3 represent the normalized results of the three sub-target functions, respectively, eliminate the dimensional differences of different target functions, and realize collaborative optimization,

[0101] f1, f2, and f3 represent the normalized results of the three sub-target functions, respectively, eliminate the dimensional differences of different target functions, and realize collaborative optimization,

[0102] f1, f2, and f3 represent the normalized results of the three sub-target functions, respectively, eliminate the dimensional differences of different target functions, and realize collaborative optimization, f1, f2, and f3 represent the normalized results of the three sub-target functions, respectively, eliminate the dimensional differences of different target functions, and realize collaborative optimization,

[0103] ​​​​​​​​​Operation logic: based on the normalized results of the three sub-objective functions, the path fitness function is constructed, the path selection probability is obtained by calculating the fitness proportion of each candidate path, the efficient screening of the optimal puncture path is realized, and the selected path not only meets the multi-objective optimization requirements, but also meets the actual needs of clinical puncture operation, through the construction of the fitness function and the calculation of the selection probability, the efficient screening of the multi-objective optimization path is realized, and the selected path can balance the requirements of high-quality low-risk follicle acquisition, vascular endothelial protection and perfusion adaptation, and adapt to the actual scene of clinical puncture operation, improve the practicability and execution efficiency of path planning.

[0104] In the embodiment of the application, the multi-objective function includes a first target of maximizing high-quality low-risk follicle acquisition rate, a second target of minimizing the length of abnormal vascular endothelial region through the puncture path, and a third target of making the puncture path conform to the target patient's follicle high perfusion region throughout the path.

[0105] The multi-objective dynamic path planning module reconstructs the three-dimensional blood flow perfusion atlas and blood vessel distribution atlas of the ovary through multi-modal imaging, establishes path generation constraints, and generates an initial puncture path that meets the requirements of blood vessel avoidance and perfusion adaptation;

[0106] The mathematical expression of the multi-objective optimization function is as follows:

[0107] ;

[0108] In the formula, is the spatial coordinate vector of the puncture path The three sub-objective functions are as follows:

[0109] High-quality low-risk follicle acquisition rate function: ;

[0110] Vascular endothelial abnormal region length function: ;

[0111] Perfusion adaptation function: ;

[0112] In the formula,

[0113] The multi-objective optimization function vector is represented by minimizing the vector to realize the coordinated optimization of the three core targets.

[0114] is the three-dimensional spatial coordinate vector of the puncture path represents the trajectory of the puncture path in the three-dimensional space of the ovary.

[0115] represents a high-quality low-risk follicle acquisition rate function, the larger the value, the more high-quality low-risk follicles that can be acquired by the puncture path;

[0116] represents a vascular endothelium abnormal area length function, the smaller the value, the smaller the potential damage to the vascular endothelium by the puncture path;

[0117] represents a perfusion adaptation function, the smaller the value, the better the puncture path fits the high-perfusion area of the follicle;

[0118] represents the total number of follicles in the ovary;

[0119] represents the weight value of the th follicle, calculated by the hierarchical attention model, reflecting the priority of the follicle;

[0120] represents an indicator function, taking a value of 1 when (the spatial coverage area of the puncture path through the th follicle) and a value of 0 otherwise, used to determine whether the path covers the target follicle;

[0121] represents the spatial coverage area of the th follicle, i.e. the range occupied by the follicle in three-dimensional space;

[0122] represents the chromosomal abnormality probability of the th follicle, calculated by the formula, reflecting the genetic quality risk of the follicle;

[0123] represents the vascular endothelium abnormality state indicator function corresponding to a point on the path, taking a value of 1 when the vascular endothelium state at that point exceeds the safety threshold, and a value of 0 otherwise, used to identify high-risk areas on the path;

[0124] represents the follicle local blood perfusion value corresponding to a point on the path, reflecting the oxygen supply and nutritional status at that location;

[0125] represents the minimum threshold value of the follicle high-perfusion area, used to define the judgment criteria for the high-perfusion area.

[0126] Operation logic: build a multi-objective optimization function, maximize the high-quality low-risk follicle acquisition rate, minimize the risk of vascular endothelial damage, and convert the three core objectives of path adaptation to blood perfusion state into a mathematical optimization problem. Through the solution of the puncture path space coordinates, the initial puncture path that meets the efficiency, safety and oocyte quality protection requirements is generated. Through multi-objective collaborative optimization, the initial puncture path can maximize the avoidance of vascular endothelial damage risk while ensuring the efficiency of high-quality low-risk follicle acquisition, and adhere to the high perfusion area to protect the activity of oocytes, achieving the triple balance of efficiency, safety and quality, and providing scientific path guidance for clinical puncture.

[0127] In the embodiment of the application, the path generation constraint conditions include: the shortest distance between the puncture path and the vascular endothelium meets the safety requirement, the path preferentially passes through the high perfusion area of the follicle, and avoids the dense area of the microvessel branch; the multi-objective dynamic path planning module constructs an integrated three-dimensional model of the ovarian anatomy and blood perfusion through three-dimensional reconstruction technology, and performs path search and optimization in the three-dimensional model to ensure that the generated initial puncture path takes into account puncture efficiency, safety and oocyte quality protection.

[0128] The intraoperative dynamic correction module updates the image and physiological state data in real time, and dynamically adjusts the puncture path;

[0129] In the embodiment of the application, the intraoperative dynamic correction module includes a real-time data updating unit and a path adjustment unit;

[0130] The real-time data updating unit continuously acquires intraoperative real-time ultrasound contrast data and ultrasound elastography data, and dynamically updates the follicle blood perfusion state data and ovarian spatial position data;

[0131] The path adjustment unit dynamically adjusts the puncture angle and path trajectory based on the updated real-time data through an intelligent optimization algorithm, compensates for the path deviation caused by physiological movement of the patient or blood flow fluctuation, and ensures that the puncture path always adapts to the follicle state and ovarian anatomical position.

[0132] Wherein, the dynamic adjustment amount of the path adjustment unit is calculated as follows:

[0133] ;

[0134] In the formula:

[0135] represents the puncture angle adjustment amount, that is, the puncture angle size that needs to be corrected, which is used to compensate for the path deviation;

[0136] represents the angle adjustment coefficient of the blood perfusion difference, which is used to adjust the influence degree of blood perfusion change on the puncture angle;

[0137] Angle adjustment coefficient representing displacement amount, used to adjust the degree of influence of ovarian displacement on puncture angle;

[0138] Difference between real-time blood perfusion value and initial perfusion value, Reflecting the dynamic changes of blood perfusion during the operation, wherein Real-time updated blood perfusion parameter, obtained from intraoperative real-time ultrasound contrast data, Blood perfusion parameter at the initial path planning, providing a reference for path adjustment;

[0139] Relative change amount of blood perfusion, eliminating the influence of initial perfusion value difference on adjustment amount, making the adjustment more universal;

[0140] Displacement amount of ovarian real-time spatial position and initial position, Reflecting the displacement degree of the ovary in three-dimensional space, 、 、 Displacement component of the ovary in three-dimensional space, derived from the distribution change of tissue elasticity state of ultrasound elastography, corresponding to x, y, z axis direction displacement respectively;

[0141] Ovarian standard displacement reference value, providing a reference for the calculation of relative change amount of displacement;

[0142] Relative change amount of displacement, eliminating the influence of individual ovarian size difference on adjustment amount;

[0143] Angle conversion coefficient, used to convert the relative change amount into the actual angle adjustment value.

[0144] Operation logic: based on the intraoperative real-time updated blood perfusion data and ovarian spatial position data, through the calculation of blood perfusion relative change amount and displacement relative change amount, combined with the corresponding angle adjustment coefficient and angle conversion coefficient, the puncture angle adjustment amount is obtained, realizing the real-time compensation of path deviation caused by physiological movement and blood flow fluctuation. Through real-time capture of blood perfusion dynamic change and ovarian spatial displacement, the puncture angle adjustment amount is accurately calculated, realizing the real-time compensation of path deviation caused by various interference factors during the operation, ensuring that the puncture path always fits the target follicle and the safety area, improving the real-time accuracy and stability of puncture.

[0145] In the embodiments of the present application, the physiological movement of the patient includes the displacement of the ovary caused by respiratory movement; the path adjustment unit locates the real-time spatial position of the ovary by analyzing the distribution change of the tissue elasticity state in the ultrasonic elastography data, and synchronously adjusts the angle and position of the puncture path in combination with the dynamic change of blood perfusion in the real-time ultrasonic contrast data, so as to ensure that the puncture target is always a high-quality low-risk follicle, and the path always avoids the abnormal area of the blood vessel endothelium.

[0146] The puncture parameter adaptive module adjusts the puncture execution parameter according to the blood vessel endothelium state and the follicle perfusion state, realizes accurate and safe follicle puncture guidance, and is suitable for the assisted reproductive ovum collection scene of the patient with advanced polycystic ovary combined with blood vessel lesions.

[0147] In the embodiments of the present application, the puncture parameter adaptive module includes a parameter adjustment rule library and a real-time control unit.

[0148] The parameter adjustment rule library pre-establishes the corresponding relationship among the blood vessel endothelium state, the follicle perfusion state and the puncture parameter, and the puncture parameter includes a puncture needle advancing speed and a needle tip pressure.

[0149] The real-time control unit calls the corresponding rule in the parameter adjustment rule library according to the blood vessel endothelium safety state result output by the AI three-dimensional evaluation module and the follicle perfusion state data updated during the operation, dynamically adjusts the puncture execution parameter, avoids damage to the blood vessel endothelium during the puncture process, and reduces the exposure time of the oocyte in the poor perfusion environment.

[0150] The adaptive adjustment formula of the puncture parameter is as follows:

[0151] The puncture needle advancing speed is: ;

[0152] The needle tip pressure is: ;

[0153] The puncture suction speed is: ;

[0154] In the formula, E represents the puncture needle advancing speed, which is dynamically adjusted according to the blood vessel endothelium state to ensure the safety of puncture;

[0155] E0 represents the basic advancing speed of the puncture needle, which provides a reference for parameter adjustment;

[0156] E1 represents the elastic modulus of the tissue around the blood vessel endothelium, which is extracted from the ultrasonic elastography data and reflects the fragility of the blood vessel endothelium;

[0157] E2 represents the elastic modulus of the tissue around the blood vessel endothelium, which is extracted from the ultrasonic elastography data and reflects the fragility of the blood vessel endothelium;

[0158] Elasticity modulus threshold of blood vessel endothelium, defining the elasticity range of blood vessel endothelium in a safe state;

[0159] Elasticity modulus corresponding to the limit fragile state of blood vessel endothelium, defining the risk threshold of blood vessel endothelium;

[0160] Speed adjustment coefficient, for adjusting the influence range of elasticity modulus change on the speed of advancement;

[0161] Needle tip pressure, adapting to the state of blood vessel endothelium to avoid excessive damage;

[0162] Basic needle tip pressure, providing a reference for pressure adjustment;

[0163] Upper limit of safe pressure, ensuring that the needle tip pressure does not exceed the tolerance range of blood vessel endothelium;

[0164] Puncture and aspiration speed, adjusted according to the perfusion state of the follicle to protect the activity of oocytes;

[0165] Basic aspiration speed, providing a reference for aspiration speed adjustment;

[0166] Aspiration speed adjustment coefficient, for adjusting the influence range of perfusion state change on the aspiration speed;

[0167] Normal perfusion threshold, defining the perfusion standard for the follicle in a normal oxygen supply state;

[0168] Real-time follicle perfusion value, obtained from real-time ultrasound contrast data during the operation, reflecting the real-time oxygen supply state of the follicle.

[0169] Operation logic: based on the elasticity modulus of the tissue around the blood vessel endothelium and the real-time perfusion state of the follicle, self-adaptive adjustment models of the puncture needle advancement speed, needle tip pressure, and puncture and aspiration speed are constructed respectively. Through the quantitative correlation between physiological state and puncture parameters, dynamic adaptation of puncture parameters is realized, balancing the safety of puncture and protection of oocyte activity. Through the dynamic adaptation of puncture parameters and the state of blood vessel endothelium and the perfusion state of the follicle, the fragile blood vessel endothelium is effectively protected during the puncture process, avoiding excessive damage, while reducing the exposure time of oocytes in adverse perfusion environment, protecting the activity of oocytes, and improving the safety and effectiveness of puncture operation.

[0170] When the mechanical characteristics of the tissue around the vascular endothelium show that the vascular endothelium is in a fragile state, the puncture parameter adaptive module reduces the puncture needle advancing speed, and controls the needle tip pressure to be within a safe range; when the follicle perfusion state shows that the follicle is in a mild hypoxic state, the module increases the puncture and suction speed, shortens the residence time of the oocyte in the follicle, and protects the activity of the oocyte.

[0171] In the embodiment of the application, the system further comprises a data storage and traceability module for storing multi-modal image data, serological data, evaluation results, puncture path planning data and intraoperative execution data, forming a complete diagnosis and treatment data chain, supporting subsequent clinical data analysis, algorithm optimization and diagnosis and treatment process traceability, and providing data support for precision and personalized treatment of oocyte retrieval in assisted reproduction.

[0172] Embodiment two: as shown in the oocyte retrieval test data embodiment of a follicle positioning and puncture path planning system based on multi-modal image fusion and AI: Figure 2

[0173] I. Test objects and grouping design

[0174] Test object basic information

[0175] 200 patients with advanced polycystic ovary and vascular lesions were selected from the hospital reproductive medicine center, aged 35-42 years old, all met the diagnostic criteria for polycystic ovary syndrome, and were confirmed by vascular ultrasound to have abnormal ovarian vascular endothelial function (resistance index ≥0.7) and uneven blood perfusion distribution. Patients with severe organ disease, coagulation dysfunction and other complications affecting oocyte retrieval were excluded.

[0176] Grouping method

[0177] A randomized controlled trial design was used to divide the patients into a test group (100 cases) and a control group (100 cases), wherein the test group was further divided into a lower age sub-group (35-38 years old, 52 cases) and a higher age sub-group (39-42 years old, 48 cases), and the control group was not divided by age.

[0178] Test group: oocyte retrieval was performed using the follicle positioning and puncture path planning system based on multi-modal image fusion and AI described in the application;

[0179] Control group: oocyte retrieval was performed using the existing conventional oocyte retrieval path planning system (single ultrasound image positioning + simple blood vessel avoidance).

[0180] II. Test method and evaluation index

[0181] Test method

[0182] ​Test group process: Collect three-dimensional transvaginal ultrasound contrast, MRI diffusion weighted imaging, MRI dynamic contrast enhancement, ultrasound elastography data and serological markers (homocysteine, vascular endothelial growth factor, antithrombin III) data, generate comprehensive feature set through multi-modal image-serological fusion module; Through AI three-dimensional evaluation module to divide the follicle puncture priority; The initial path is generated by the multi-objective dynamic path planning module, the deviation is compensated by the intraoperative dynamic correction module, and the execution parameters are adjusted by the puncture parameter self-adaptive module to complete the oocyte retrieval.

[0183] Control group process: Only follicle positioning is performed by transvaginal two-dimensional ultrasound, the doctor plans the puncture path according to experience, and completes the oocyte retrieval with fixed puncture parameters.

[0184] Evaluation index;

[0185] Core index: high-quality follicle recognition accuracy, available embryo rate, post-puncture vascular injury complication rate, oocyte viability;

[0186] Auxiliary index: path planning time, puncture times, postoperative ovarian blood flow recovery time.

[0187] III. Test results;

[0188]

[0189]

[0190] IV. Analysis of test results;

[0191] High-quality egg screening problem solved: The control group only uses ultrasonic morphological evaluation, and the high-quality follicle recognition accuracy is only 63.2%, and the available embryo rate is 31.5%; The test group uses multi-modal image-serological fusion and AI three-dimensional evaluation, accurately associates follicle hypoxic state with chromosomal abnormality risk, and the recognition accuracy is improved to 82.5%, and the available embryo rate is improved to 56.8%, among which the younger subgroup has better follicle shape and activity characteristics, and the effect is more significant (60.3%), and the older subgroup is still improved by 21.4 percentage points compared with the control group, solving the core problem of "oocyte number meeting the standard but available embryo rate being low".

[0192] Vascular safety protection effect is significant: The control group has a post-puncture vascular injury complication rate of 14.8% due to the lack of targeted protection of vascular endothelial fragility; The test group avoids fragile vascular areas through multi-objective path planning, combined with puncture parameter self-adaptive adjustment, and the complication rate is reduced to 3.2%, both subgroups remain at a low level, effectively solving the pain point of "vascular safety protection difficulty" and ensuring patient treatment safety.

[0193] Oocyte activity guarantee in place: the control group does not consider the influence of blood perfusion on oocyte activity, and the survival rate is only 74.6%; the test group path is preferentially attached to the high perfusion area, and the suction speed is adaptively increased during the puncture of the anoxic follicle, the oocyte survival rate reaches 91.3%, and the difference between different age subgroups is small, which can effectively reduce the influence of the adverse perfusion environment on the oocyte, and solve the problem of "oocyte activity guarantee difficulty".

[0194] Treatment efficiency and recovery speed are improved: the path planning time of the test group is shortened by 50.6% compared with the control group, the puncture times are reduced by 48.5%, and the postoperative ovarian blood flow recovery time is shortened by 43.1%, which not only reduces the surgical trauma of the patient, but also improves the clinical treatment efficiency, and meets the actual needs of assisted reproduction clinics.

[0195] The embodiments of the present application are disclosed, but are not limited to this, and those skilled in the art can easily understand the spirit of the present application according to the above embodiments, and make different inferences and changes, as long as they do not deviate from the spirit of the present application, they are within the protection scope of the present application.

Claims

1. A follicle localization and puncture path planning system based on multimodal image fusion and AI, characterized in that, include: The multimodal imaging-serology fusion module acquires multimodal imaging data and serological biomarker data of the target patient and fuses them to generate a comprehensive feature set. The AI ​​three-dimensional assessment module, based on a comprehensive feature set, performs a joint assessment of follicle chromosomal risk, blood perfusion status, and vascular endothelial safety status, and classifies puncture priorities. The multi-objective dynamic path planning module aims to maximize the acquisition rate of high-quality, low-risk follicles, minimize the risk of vascular endothelial injury, and adapt the path to the blood perfusion state, thereby generating an initial puncture path. The multi-objective dynamic path planning module reconstructs the three-dimensional blood flow perfusion map and vascular distribution map of the ovary through multimodal imaging, establishes path generation constraints, and searches for the optimal initial puncture path based on a multi-objective function. The multi-objective dynamic path planning module constructs an integrated three-dimensional model of ovarian anatomy and blood perfusion using three-dimensional reconstruction technology. In the model, it performs path search and optimization based on constraints to adapt to the ovarian physiological state of the target patient. The intraoperative dynamic correction module is used to update data and adjust the path in real time to compensate for deviations; The puncture parameter adaptive module dynamically adjusts the puncture execution parameters based on the vascular endothelium and follicle perfusion status.

2. The follicle localization and puncture path planning system based on multimodal image fusion and AI according to claim 1, characterized in that, The multimodal image-serology fusion module includes: The image data acquisition unit acquires transvaginal three-dimensional ultrasound contrast imaging data, MRI diffusion-weighted imaging data, MRI dynamic contrast-enhanced data and ultrasound elastography data, and extracts blood flow distribution characteristics, follicular cell activity characteristics, blood flow perfusion characteristics and mechanical characteristics of tissues surrounding vascular endothelium, respectively. The serological marker data include data related to homocysteine, vascular endothelial growth factor, and antithrombin III. The feature fusion unit uses a feature association mapping algorithm to associate image features with serological marker data, and constructs a comprehensive feature set that includes follicle quality features, blood perfusion features, vascular endothelial safety features and serological association features. The comprehensive feature set includes fusion features that reflect the association between follicular hypoxia and vascular lesions.

3. The follicle localization and puncture path planning system based on multimodal image fusion and AI according to claim 1, characterized in that, The AI ​​three-dimensional evaluation module includes: The hierarchical attention model dynamically adjusts the weight ratio of imaging features and serological features based on the age stratification of target patients, thereby achieving age-adaptive feature weighting. The weight adjustment logic of the hierarchical attention model is as follows: for older patients in the younger age group, the weight ratio of follicular morphology characteristics and granulosa cell activity characteristics is increased; for older patients in the older age group, the weight ratio of serological marker data and blood perfusion characteristics is increased. A two-dimensional classification matrix construction unit, through joint fitting of multimodal imaging features and serological features, derives the association between follicular hypoxia status and chromosomal abnormality risk, classifying follicles into four categories: low-risk hypoxia, low-risk hyperxia, high-risk hypoxia, and high-risk hyperxia, corresponding to different puncture priorities and timings. The association derivation formula is as follows: ； In the formula, Indicates the probability of chromosomal abnormalities in follicles; This indicates the follicular hypoxia index. ,in This represents the apparent diffusion coefficient in MRI diffusion-weighted imaging. This represents the blood perfusion parameters in MRI dynamic contrast enhancement. , Indicates the feature weight coefficient; Indicates serological association characteristics, ,in This indicates the concentration of homocysteine. Indicates the concentration of vascular endothelial growth factor. , Indicates the serum characteristic weighting coefficient; , , Represents the fitting coefficient; The safety threshold determination unit sets safety criteria based on the mechanical characteristics of the tissue surrounding the vascular endothelium extracted by ultrasound elastography, and lowers the puncture priority for follicles that exceed the standard area.

4. The follicle localization and puncture path planning system based on multimodal image fusion and AI according to claim 3, characterized in that, The hierarchical attention model uses an adaptive weighted algorithm to calculate the weights: When the target patients are younger or older patients, the weighting formula is: ; When the target patients are older patients... : ; In the formula, This indicates the comprehensive weighting of characteristics of older patients in the younger age group; This indicates the comprehensive weighting of characteristics of older patients in the higher age group; Indicates the basic weights of image features; Indicates the basic weights of serological characteristics; Indicates the probability of chromosomal abnormalities in follicles; , , , Represents the weighting adjustment coefficient, where , .

5. The follicle localization and puncture path planning system based on multimodal image fusion and AI according to claim 1, characterized in that, The multi-objective dynamic path planning module employs a multi-objective optimization algorithm based on an improved non-dominated sorting genetic algorithm to construct a multi-objective function: ； In the formula Spatial coordinate vector of the puncture path ; The function for the high-quality, low-risk follicle retrieval rate. ,in This indicates the total number of follicles in the ovary. Indicates the first The weight value of each follicle. Indicates the first The spatial coverage area of ​​each follicle Indicates the first The probability of chromosomal abnormalities in each follicle; This is a function of the length of the abnormal region in the vascular endothelium. ,in Represents a point on the path The corresponding vascular endothelial abnormality indicator function; For the infusion adaptation function, ,in Represents a point on the path The corresponding local blood perfusion value of the follicle. This represents the minimum threshold for areas of high follicular perfusion.

6. The follicle localization and puncture path planning system based on multimodal image fusion and AI according to claim 5, characterized in that, The path generation constraints include: the shortest distance between the puncture path and the vascular endothelium meets safety requirements, the path preferentially passes through follicular hyperperfusion areas, and avoids areas with dense microvascular branches.

7. The follicle localization and puncture path planning system based on multimodal image fusion and AI according to claim 1, characterized in that, The intraoperative dynamic correction module includes: The real-time data update unit is used to continuously acquire intraoperative real-time ultrasound contrast imaging data and ultrasound elastography data, and dynamically update follicular blood flow perfusion status data and ovarian spatial location data. The path adjustment unit, based on updated real-time data, calculates the puncture angle adjustment amount through an intelligent optimization algorithm, dynamically adjusts the puncture angle and path trajectory, and compensates for path deviations caused by patient physiological movements or blood flow fluctuations. The dynamic adjustment amount is calculated as follows: ； In the formula: Indicates the amount of adjustment for the puncture angle; Angle adjustment factor representing the difference in blood perfusion; An angular adjustment coefficient representing the displacement; This represents the difference between the real-time blood perfusion value and the initial perfusion value. ,in This indicates real-time updated blood perfusion parameters. This represents the blood perfusion parameters during initial path planning; This indicates the relative change in blood perfusion; This represents the displacement of the ovary from its initial spatial position in real time. ,in , , These represent the displacement components of the ovary in three-dimensional space, corresponding to displacements along the x, y, and z axes, respectively. This represents the standard displacement reference value for the ovary; This represents the relative change in displacement; This represents the angle conversion factor.

8. The follicle localization and puncture path planning system based on multimodal image fusion and AI according to claim 1, characterized in that, The puncture parameter adaptive module includes: A parameter adjustment rule base is used to pre-establish the correspondence between vascular endothelial status, follicle perfusion status and puncture parameters, including puncture needle advance speed, needle tip pressure and puncture aspiration speed. The real-time control unit, based on the vascular endothelial safety status results output by the AI ​​three-dimensional assessment module and the follicle perfusion status data updated during the operation, calls the corresponding rules in the parameter adjustment rule base to dynamically adjust the puncture execution parameters, thereby avoiding vascular endothelial damage and reducing the exposure time of oocytes in adverse perfusion environments.

9. The follicle localization and puncture path planning system based on multimodal image fusion and AI according to claim 8, characterized in that, The adaptive adjustment formula for the puncture parameters is as follows: Needle advance speed: ; Needle tip pressure: ; Puncture and aspiration rate: ; In the formula, Indicates the speed at which the puncture needle is advanced; Indicates the basic advance speed of the puncture needle; Indicates the elastic modulus of the tissue surrounding the vascular endothelium; This represents the threshold value for the safe elastic modulus of the vascular endothelium. This represents the elastic modulus corresponding to the extreme fragility state of the vascular endothelium. Indicates the speed adjustment coefficient. ; Indicates needle tip pressure; Indicates the basic needle tip pressure; Indicates the upper limit of safety pressure; Indicates the puncture and aspiration rate; Indicates the basic suction rate; This represents the suction speed adjustment coefficient. ; Indicates the normal perfusion threshold; This represents the real-time follicle perfusion value; When the vascular endothelium is in a fragile state, reduce the speed of needle advancement and control the needle tip pressure within a safe range; when the follicle is in a state of mild hypoxia, increase the puncture and aspiration speed to ensure oocyte viability.

10. The follicle localization and puncture path planning system based on multimodal image fusion and AI according to claim 1, characterized in that, It also includes a data storage and traceability module, which is used to store multimodal image data, serological marker data, AI three-dimensional assessment results, puncture path planning data, intraoperative real-time update data, puncture parameter adjustment records and puncture execution data, forming a complete diagnosis and treatment data chain.

Images