Image-guided rectal puncture visualization and irrigation method based on fluorescence contrast enhancement

Abstract

This invention relates to the field of medical auxiliary technology, and more particularly to an image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement. The method includes: firstly, using artificial intelligence to segment preoperative pelvic magnetic resonance images to generate a three-dimensional model containing the target lesion and key anatomical structures, and planning a safe puncture path accordingly; during the procedure, using elastic registration technology to spatially align real-time ultrasound images with the three-dimensional model to generate a fused image for real-time navigation of the puncture instrument to the target lesion; subsequently, injecting a fluorescent contrast agent into the lesion, and confirming the true extent of the lesion through fluorescence imaging; finally, while irrigating the lesion, monitoring the attenuation of the fluorescence signal intensity in real time, and objectively determining the irrigation endpoint when the signal intensity drops to the background noise level and the change curve enters a plateau phase. This invention achieves precision and objectivity throughout the entire process, from preoperative planning to intraoperative navigation and efficacy verification.

Description

Technical Field

[0001] This invention relates to the field of medical auxiliary technology, and in particular to an image-guided rectal puncture, imaging, and irrigation method based on fluorescence contrast enhancement. Background Technology

[0002] In the diagnosis and treatment of perianal abscesses and anal fistulas, especially complex anal fistulas, accurate localization of the lesion, determination of its extent, and its relationship with the anal sphincter are crucial for surgical success. Currently, clinical practice often relies on magnetic resonance imaging (MRI) for preoperative assessment to obtain anatomical information, followed by ultrasound or CT-guided puncture and drainage.

[0003] However, existing image-guided techniques have a significant limitation: intraoperative images, such as ultrasound, cannot intuitively and in real time determine whether the abscess cavity has been thoroughly cleaned. Doctors mainly rely on experience to observe whether the irrigation fluid has cleared up to make a judgment. However, for viscous pus or complex abscess cavities, this subjective judgment method is very likely to lead to incomplete irrigation, which becomes an important cause of postoperative infection recurrence. Summary of the Invention

[0004] To overcome the above shortcomings, this invention provides an image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement, which aims to improve the problem in the prior art that it is difficult to intuitively and in real time determine whether the abscess cavity has been thoroughly rinsed clean.

[0005] In a first aspect, the present invention provides the following technical solution: an image-guided rectal puncture and imaging irrigation method based on fluorescence contrast enhancement, the method comprising the following steps:

[0006] Preoperative pelvic magnetic resonance images of patients are acquired, input into a pre-trained artificial intelligence segmentation model, processed, and a three-dimensional model containing the target lesion area and key anatomical structures is generated.

[0007] Based on the three-dimensional model, one or more recommended puncture paths are calculated;

[0008] During the procedure, real-time ultrasound images are acquired and elastically registered with the three-dimensional model to spatially align the three-dimensional model with the real-time anatomical structure, generating a fused image.

[0009] The puncture procedure is performed according to the recommended puncture path shown in the fused image to reach the target lesion area;

[0010] A fluorescent contrast agent is injected into the target lesion area, and the distribution of the fluorescent contrast agent is observed based on the fluorescent image to confirm the extent of the target lesion.

[0011] The target lesion area, which has been filled with the fluorescent contrast agent, is flushed, and the endpoint of the flushing is determined by real-time observation of the changes in the signal intensity of the fluorescent contrast agent in the fluorescent image.

[0012] Preferably, the process for generating the 3D model includes:

[0013] The raw data of the preoperative pelvic magnetic resonance imaging (MRI) images in DICOM format were acquired, and the raw data in DICOM format was preprocessed.

[0014] The preprocessed image data is input into the pre-trained artificial intelligence segmentation model;

[0015] The artificial intelligence segmentation model outputs a pixel-level segmentation mask for the target lesion area and key anatomical structures.

[0016] Based on the pixel-level segmentation mask, three-dimensional reconstruction is performed to generate a three-dimensional model containing the target lesion area and key anatomical structures.

[0017] Preferably, the calculation process for the puncture path includes:

[0018] In the three-dimensional model, the starting region and the target region for puncture are defined;

[0019] In the three-dimensional model, one or more critical anatomical structures that need to be avoided are marked;

[0020] Based on a preset path planning algorithm, a path search is performed between the starting region and the target region;

[0021] During the path search process, the critical anatomical structure to be avoided is introduced as a constraint condition, and one or more paths that maintain a safe distance from the critical anatomical structure are calculated as candidate paths.

[0022] The candidate paths are scored based on path length and safety margin, and one or more recommended puncture paths are output based on the scoring results.

[0023] Preferably, the process for generating the fused image includes:

[0024] Real-time acquisition of the spatial pose of the ultrasound probe in the patient coordinate system;

[0025] Based on the spatial pose of the ultrasound probe, the real-time ultrasound image is mapped onto the patient coordinate system to form real-time three-dimensional ultrasound data.

[0026] Extract the corresponding set of anatomical feature points from the real-time three-dimensional ultrasound data and the three-dimensional model;

[0027] The spatial transformation relationship between the three-dimensional model and the real-time three-dimensional ultrasound data is calculated using a non-rigid iterative nearest point algorithm.

[0028] Based on the calculated spatial transformation relationship, the three-dimensional model is elastically deformed to achieve spatial alignment with the anatomical structures in the real-time three-dimensional ultrasound data.

[0029] The aligned 3D model is overlaid with the real-time 3D ultrasound data to generate the fused image.

[0030] Preferably, the process for generating the real-time three-dimensional ultrasound data includes:

[0031] Acquire two-dimensional real-time ultrasound image frames and their corresponding spatial pose matrices;

[0032] Based on the spatial pose matrix, the spatial position and imaging plane orientation of each frame of the two-dimensional real-time ultrasound image in the patient coordinate system are determined;

[0033] Constrained by the spatial position and imaging plane direction, each frame of the two-dimensional real-time ultrasound image is used as a slice of data and filled into a common three-dimensional voxel grid.

[0034] Null interpolation is performed on the regions in the three-dimensional voxel mesh that are not filled by the slice data to generate complete, spatially continuous real-time three-dimensional ultrasound data.

[0035] Preferably, the process of elastically deforming the three-dimensional model includes:

[0036] Obtain the spatial transformation relationship calculated by the non-rigid iterative nearest point algorithm, wherein the spatial transformation relationship is displacement field data used to drive the deformation of the three-dimensional model;

[0037] Based on the deformation parameters, a displacement vector is calculated for each vertex of the three-dimensional model using a spatial interpolation algorithm;

[0038] Based on the calculated displacement vector, the spatial position of each vertex of the three-dimensional model is moved, thereby completing the overall elastic deformation of the three-dimensional model;

[0039] The anatomical contour of the deformed 3D model is compared with the anatomical structure in the real-time 3D ultrasound data to verify the accuracy of spatial alignment.

[0040] Preferably, the procedure for performing a puncture operation according to the recommended puncture path shown in the fused image to reach the target lesion area includes:

[0041] On the real-time navigation view of the fused image, the recommended puncture path and the real-time position and orientation of the puncture instrument are rendered synchronously.

[0042] Adjust the spatial orientation of the puncture instrument so that its projection in the real-time navigation view coincides with the extension line of the recommended puncture path;

[0043] Insert the needle along the recommended puncture path, and monitor the deviation between the tip of the puncture instrument and the recommended puncture path in real time during the needle insertion process;

[0044] If the detected deviation exceeds the preset tolerance, the needle insertion direction is adjusted to correct the deviation until the tip of the puncture instrument reaches the target lesion area.

[0045] Preferably, the process for confirming the extent of the target lesion includes:

[0046] A preset dose of fluorescent contrast agent is slowly injected at constant pressure through the channel of the puncture instrument that has reached the target lesion area;

[0047] Switch the imaging mode to fluorescence mode to acquire and display dynamic images of the diffusion process of the fluorescent contrast agent in the tissue in real time;

[0048] In the dynamic image, the region filled with fluorescence signal is identified, and the boundary of the region is the real-time morphological boundary of the target lesion;

[0049] Based on the real-time morphological boundaries, the final extent of the target lesion is confirmed, wherein the dynamic images are used to identify fistula branches or satellite abscesses that were not anticipated in the preoperative images.

[0050] Preferably, the process for determining the rinsing endpoint includes:

[0051] The irrigation fluid is pulsed into the target lesion area using the puncture instrument.

[0052] In fluorescence imaging mode, the overall intensity value of the fluorescent contrast agent signal in the target lesion area is monitored in real time;

[0053] Record the decrease curve of the overall strength value during the rinsing process;

[0054] When the overall intensity value drops to the level of background noise intensity and the decline curve enters a plateau period, it is determined that the rinsing has reached its end.

[0055] The present invention has the following beneficial effects:

[0056] 1. In this invention, by seamlessly connecting multiple links such as artificial intelligence segmentation, three-dimensional path planning, multimodal image fusion, fluorescence navigation and quantitative verification, a complete closed loop is constructed from precise preoperative planning to precise intraoperative execution and immediate postoperative effect verification, which significantly improves the systematicness and intelligence level of diagnosis and treatment of perianal diseases.

[0057] 2. In this invention, a safe path is automatically planned based on an AI 3D model before surgery, and the pre-planned path is dynamically integrated with the real-time anatomical structure by combining the intraoperative flexible registration technology. This provides doctors with an augmented reality navigation view that is both forward-looking and real-time. This not only guides the puncture instrument to the target accurately, but also effectively avoids critical anatomical structures and protects anal function to the greatest extent.

[0058] 3. In this invention, the dynamic change of the signal intensity of the fluorescent contrast agent is used as an objective quantitative indicator. By monitoring the signal intensity decline curve in real time and setting clear endpoint criteria, the judgment of the flushing endpoint is changed from relying on subjective experience to relying on objective data, which ensures the thoroughness of flushing and reduces the risk of infection recurrence from the root. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the process of the image-guided rectal puncture, imaging, and irrigation method based on fluorescence contrast enhancement proposed in this invention. Detailed Implementation

[0060] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0061] In a first embodiment of the present invention, the present invention provides an image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement, such as... Figure 1 As shown, it includes the following steps:

[0062] Preoperative pelvic magnetic resonance images of the patient are acquired and input into a pre-trained artificial intelligence segmentation model to process and generate a three-dimensional model containing the target lesion area and key anatomical structures.

[0063] Furthermore, the process of generating a 3D model includes:

[0064] Acquire raw DICOM format data of preoperative pelvic magnetic resonance imaging and preprocess the raw DICOM format data;

[0065] The preprocessed image data is input into a pre-trained artificial intelligence segmentation model;

[0066] The artificial intelligence segmentation model outputs a pixel-level segmentation mask for the target lesion area and key anatomical structures.

[0067] Based on pixel-level segmentation masking, 3D reconstruction is performed to generate a 3D model containing the target lesion area and key anatomical structures.

[0068] Specifically, preoperative pelvic MRI images of the patient are acquired. These images are typically stored in DICOM format and contain a series of two-dimensional tomographic images acquired from the patient's pelvic region. Next, the acquired raw DICOM data is preprocessed. Preprocessing steps include image normalization and resolution normalization. Image normalization specifically adjusts the pixel value range of the image to a standard interval, such as between 0 and 1, to eliminate data differences caused by different scanning devices or protocols. The calculation formula is as follows:

[0069] ;

[0070] in, Represents the original pixel value. and These represent the minimum and maximum pixel values ​​in the image, respectively. Represents the normalized pixel value;

[0071] Resolution standardization specifically involves resampling all two-dimensional tomographic images to a uniform isotropic resolution, such as 1 mm x 1 mm x 1 mm, to ensure the geometric accuracy of subsequent three-dimensional reconstruction.

[0072] The preprocessed image data is input into a pre-trained artificial intelligence segmentation model, preferably a deep learning convolutional neural network based on the U-Net architecture. This network extracts high-level features of the image step by step through the encoder path, and then recovers spatial information step by step through the decoder path combined with skip connections. Finally, a class label is assigned to each pixel in the output layer. The model has been trained using a large dataset of pelvic magnetic resonance images annotated by experts. Its optimization objective is to minimize the cross-entropy loss function between the predicted segmentation result and the real annotation.

[0073] After the AI ​​segmentation model runs, it outputs a pixel-level segmentation mask for the target lesion area and key anatomical structures. This segmentation mask is a binary or polynomial image with the same size as the input image, where the value of each pixel identifies whether it belongs to the target lesion area, the internal anal sphincter, the external anal sphincter, the puborectalis muscle, or the background. Finally, based on the obtained pixel-level segmentation mask, three-dimensional reconstruction is performed. This process stacks a series of two-dimensional segmentation masks in three-dimensional space according to the layer thickness and interlayer spacing in their original DICOM file header information, and uses the moving cube algorithm or surface rendering algorithm to generate a continuous, smooth three-dimensional mesh model containing the target lesion area and key anatomical structures. This three-dimensional model uses a set of vertices and faces to represent the surface geometry of different anatomical structures, providing an accurate anatomical space reference for subsequent path planning.

[0074] Through the above steps, a pre-trained artificial intelligence segmentation model can be used to automatically process pelvic magnetic resonance images, which can efficiently and accurately generate a three-dimensional model containing the target lesion and key anatomical structures. This transforms the traditional two-dimensional, subjective image assessment into an objective, quantitative three-dimensional anatomical map, providing an indispensable and accurate spatial benchmark for subsequent planning of safe puncture paths, and improving the scientific nature and safety of surgical planning.

[0075] Based on the 3D model, one or more recommended puncture paths are calculated.

[0076] Furthermore, the calculation process for the puncture path includes:

[0077] In the 3D model, define the starting and target areas for puncture;

[0078] In the 3D model, mark one or more critical anatomical structures that need to be avoided;

[0079] Based on a preset path planning algorithm, a path search is performed between the starting region and the target region;

[0080] During the path search process, critical anatomical structures that need to be avoided are introduced as constraints, and one or more paths that maintain a safe distance from the critical anatomical structures are calculated as candidate paths.

[0081] Candidate paths are scored based on path length and safety margin, and one or more recommended puncture paths are output based on the scoring results.

[0082] Specifically, firstly, the starting and target areas for puncture are defined in the 3D model. The starting area is usually defined as the 3D spatial range corresponding to the preset skin puncture point, and the target area is defined as the target lesion area, that is, the spatial range occupied by the core location of the abscess or fistula in the 3D model. Next, one or more critical anatomical structures that need to be avoided are marked in the 3D model. These critical anatomical structures include at least the internal anal sphincter, external anal sphincter, puborectalis muscle, and important vascular nerve bundles. These structures are marked as insurmountable obstacle areas in the 3D model. Based on a preset path planning algorithm, a path search is performed between the starting and target areas. The preferred path planning algorithm is the A algorithm. The A algorithm searches for the optimal path by evaluating the cost function of the path. Its cost function f(n) consists of two parts: f(n) = g(n) + h(n), where g(n) represents the actual cost of moving from the starting area to the current node n, and h(n) represents the estimated cost from the current node n to the target area, which is usually calculated using Euclidean distance.

[0083] During the path search process, critical anatomical structures to be avoided are introduced as constraints. The algorithm calculates the shortest spatial distance between each point on the path and the surfaces of all critical anatomical structures. The goal of this path search is to find one or more connected paths that maintain a minimum preset safety distance from all critical anatomical structures. These paths are recorded as candidate paths. The preset safety distance can be set based on clinical experience, for example, 3 to 5 millimeters. Then, all candidate paths are scored based on path length and safety margin. The path scoring function S can be expressed as:

[0084] ;

[0085] Where L represents the length of the path, and D represents the minimum distance between all points on the path and the critical anatomical structure, i.e., the safety margin. and The weighting coefficient is used to balance the relative importance of path length and safety. The system sorts all candidate paths according to the scoring results and outputs one or more paths with the best scores as the final recommended puncture path for doctors to refer to during surgery.

[0086] Through the above steps, automated path planning and multi-objective optimization scoring can be performed within the anatomical space of the three-dimensional maintenance model, with the avoidance of critical structures as a hard constraint. This transforms the puncture path selection process, which relies on personal experience and spatial imagination, into a decision-making process driven by objective algorithms that quantitatively assesses safety and efficiency. As a result, the optimal solution that combines the shortest path with the maximum safety margin can be recommended to the surgeon, reducing the risk of accidental damage to key tissues and anal dysfunction during surgery.

[0087] During the procedure, real-time ultrasound images are acquired and elastically registered with a 3D model to spatially align the 3D model with the real-time anatomical structures, generating a fused image.

[0088] Furthermore, the process of generating fused images includes:

[0089] Real-time acquisition of the spatial pose of the ultrasound probe in the patient coordinate system;

[0090] Based on the spatial pose of the ultrasound probe, real-time ultrasound images are mapped onto the patient's coordinate system to form real-time three-dimensional ultrasound data.

[0091] Extract the corresponding set of anatomical feature points from real-time 3D ultrasound data and 3D models;

[0092] The spatial transformation relationship between the 3D model and real-time 3D ultrasound data is calculated using a non-rigid iterative nearest point algorithm.

[0093] Based on the calculated spatial transformation relationship, the three-dimensional model is elastically deformed to achieve spatial alignment with the anatomical structures in the real-time three-dimensional ultrasound data.

[0094] The aligned 3D model is overlaid with real-time 3D ultrasound data to generate a fused image.

[0095] Furthermore, the process for generating real-time three-dimensional ultrasound data includes:

[0096] Acquire two-dimensional real-time ultrasound image frames and their corresponding spatial pose matrices;

[0097] Based on the spatial pose matrix, the spatial position and imaging plane orientation of each frame of two-dimensional real-time ultrasound image in the patient coordinate system are determined;

[0098] Constrained by spatial location and imaging plane orientation, each frame of two-dimensional real-time ultrasound image is treated as a slice of data and filled into a common three-dimensional voxel grid.

[0099] Null interpolation is performed on regions in the three-dimensional voxel mesh that are not filled by slice data to generate complete, spatially continuous real-time three-dimensional ultrasound data.

[0100] Furthermore, the process of elastically deforming a 3D model includes:

[0101] Obtain the spatial transformation relationship calculated by the non-rigid iterative nearest point algorithm, where the spatial transformation relationship is the displacement field data used to drive the deformation of the three-dimensional model;

[0102] Based on the deformation parameters, a displacement vector is calculated for each vertex of the 3D model using a spatial interpolation algorithm;

[0103] Based on the calculated displacement vector, the spatial position of each vertex of the three-dimensional model is moved, thereby completing the overall elastic deformation of the three-dimensional model;

[0104] The anatomical contours of the deformed 3D model were compared with the anatomical structures in real-time 3D ultrasound data to verify the accuracy of spatial alignment.

[0105] Specifically, the first step is to acquire real-time ultrasound images and perform elastic registration with the three-dimensional model. To achieve this, the spatial pose of the ultrasound probe in the patient coordinate system needs to be acquired in real time. This process is achieved through an optical positioning and tracking system. The system continuously tracks optical markers fixed on the ultrasound probe and outputs the spatial position and pose of the probe in the patient coordinate system at a frequency of tens of times per second. This pose is usually represented by a 4x4 homogeneous transformation matrix, i.e., the spatial pose matrix.

[0106] Based on the spatial pose of the ultrasound probe, real-time ultrasound images are mapped onto the patient coordinate system to form real-time three-dimensional ultrasound data. The specific process is as follows: the system synchronously acquires each frame of two-dimensional real-time ultrasound image output by the ultrasound probe and records the spatial pose matrix corresponding to the frame image. According to the spatial pose matrix, the spatial position of each frame of two-dimensional image in the three-dimensional patient coordinate system and the orientation of its imaging plane can be accurately determined. Then, using a common three-dimensional voxel grid with a preset resolution as a container, each frame of two-dimensional image is treated as a layer of slice data and filled into the corresponding position of the three-dimensional grid according to its spatial position and orientation. For voxels in the three-dimensional voxel grid that are not filled by any slice data, a trilinear interpolation algorithm is used to perform null interpolation processing based on the pixel values ​​of its neighboring voxels, and finally a complete and spatially continuous real-time three-dimensional ultrasound volume data is generated.

[0107] Next, the corresponding set of anatomical feature points is extracted from the reconstructed real-time 3D ultrasound data and the preoperative 3D model. These feature points include stable anatomical landmarks such as the anterior margin of the pubic symphysis, the tip of the coccyx, and the junction of the anorectal canal. The spatial transformation relationship between the 3D model and the real-time 3D ultrasound data is calculated using a non-rigid iterative nearest point algorithm. This algorithm iteratively performs two steps: first, it finds the nearest point in the real-time 3D ultrasound data for each feature point on the 3D model and establishes a point-to-point correspondence; second, based on these point pairs, it solves a spatial transformation relationship that minimizes the average distance between all corresponding point pairs. The transformation relationship here is not limited to rigid body transformation but allows elastic deformation. Its output is a displacement field data used to drive the deformation of the 3D model.

[0108] Based on the calculated spatial transformation relationship, the 3D model undergoes elastic deformation. Specifically, the system acquires the displacement field data and uses spatial interpolation algorithms such as thin-plate spline interpolation to calculate a specific displacement vector for each vertex on the 3D model mesh. Subsequently, based on the calculated displacement vector, the spatial position of each vertex of the 3D model is moved, thereby completing the overall elastic deformation of the 3D model. After deformation, the anatomical contours of the deformed 3D model, such as lesion boundaries and sphincter contours, are visually compared with the corresponding structures in the real-time 3D ultrasound data to verify whether the spatial alignment accuracy meets the requirements of surgical navigation. Finally, the aligned, elastically deformed 3D model is superimposed with the real-time 3D ultrasound data. During superposition, the 3D model is rendered as a semi-transparent colored curved surface, and the real-time 3D ultrasound data is rendered as a grayscale volume, both displayed on the navigation screen to generate the final fused image that can be used for real-time guidance.

[0109] Through the above steps, the preoperative three-dimensional model and the intraoperative real-time three-dimensional ultrasound data can be elastically registered and fused with high precision. This effectively compensates for the displacement of anatomical structures caused by changes in patient position and tissue deformation, generating a fused navigation image that can truly reflect the real-time anatomical situation. This allows the static preoperative planning to be accurately mapped to the dynamic surgical environment, providing a reliable spatial reference and guidance for subsequent puncture and irrigation operations.

[0110] The puncture procedure is performed according to the recommended puncture path shown in the fused image to reach the target lesion area.

[0111] Furthermore, the procedure for performing a puncture based on the recommended puncture path shown in the fused images to reach the target lesion area includes:

[0112] On the real-time navigation view of the fused images, the recommended puncture path and the real-time position and orientation of the puncture instrument are rendered simultaneously.

[0113] Adjust the spatial orientation of the puncture instrument so that its projection in the real-time navigation view coincides with the extension of the recommended puncture path;

[0114] Insert the needle along the recommended puncture path, and monitor the deviation between the tip of the puncture instrument and the recommended puncture path in real time during the needle insertion process;

[0115] If the detected deviation exceeds the preset tolerance, the needle insertion direction is adjusted to correct the deviation until the tip of the puncture instrument reaches the target lesion area.

[0116] Specifically, in the real-time navigation view of the fused images, the recommended puncture path and the real-time position and orientation of the puncture instrument are rendered simultaneously. The recommended puncture path is presented as a highlighted virtual line with depth markings. The real-time position and orientation of the puncture instrument are obtained by installing optical positioning markers on the proximal end of the puncture needle and tracking its spatial pose matrix in real time by the optical positioning tracking system. Based on this pose matrix, the system draws a 3D model of the puncture instrument in the navigation view in real time. Then, the doctor adjusts the spatial pose of the puncture instrument. The goal of this adjustment is to make the 3D projection of the puncture instrument in the real-time navigation view visually coincide with the virtual extension line of the recommended puncture path. This is achieved by observing the screen and moving the puncture needle to align the axis of its virtual projection with the axis of the planned path.

[0117] The doctor inserts the needle along the recommended puncture path. During the insertion, the system monitors the deviation between the tip of the puncture instrument and the recommended puncture path in real time. This deviation is the Euclidean distance from the three-dimensional spatial coordinates of the needle tip to the straight line of the recommended puncture path. The system has a preset tolerance value, which is usually set according to clinical safety requirements, such as 2 mm. If the calculated real-time deviation exceeds this preset tolerance during monitoring, the system will issue a visual warning to the doctor, such as changing the path color from green to red or flashing a prompt. After receiving the warning, the doctor pauses the insertion and fine-tunes the insertion direction to correct the deviation, so that the needle tip trajectory returns to the tolerance range. The doctor repeats the above steps of insertion, real-time monitoring and deviation correction until the tip of the puncture instrument, based on its spatial positioning coordinates, has entered the three-dimensional spatial range of the target lesion area, thus completing the precise puncture.

[0118] Through the above steps, the virtual recommended puncture path and the real-time spatial pose of the puncture instrument can be superimposed and displayed on the fused image navigation view, providing doctors with intuitive visual guidance and real-time spatial feedback. This enables them to accurately align the instrument with the planned path and ensure that the puncture trajectory is always within the safety tolerance through continuous monitoring and active correction during dynamic needle insertion, thereby ultimately achieving precise and safe arrival of the instrument tip at the target lesion area.

[0119] Fluorescent contrast agent is injected into the target lesion area, and the distribution of the fluorescent contrast agent is observed based on the fluorescence imaging to confirm the extent of the target lesion.

[0120] Furthermore, the process for confirming the extent of the target lesion includes:

[0121] A preset dose of fluorescent contrast agent is slowly injected at constant pressure through the channel of the puncture instrument that has reached the target lesion area;

[0122] Switch the imaging mode to fluorescence mode to acquire and display dynamic images of the diffusion process of the fluorescent contrast agent in the tissue in real time;

[0123] In dynamic images, the region filled with fluorescence signal is identified, and the boundary of this region is the real-time morphological boundary of the target lesion;

[0124] Based on real-time morphological boundaries, the final extent of the target lesion is confirmed, with dynamic imaging used to identify fistula branches or satellite abscesses that were not anticipated in preoperative imaging.

[0125] Specifically, through the internal channel of the puncture instrument that has reached the target lesion area, a preset dose of fluorescent contrast agent is slowly injected with a syringe at constant pressure. The fluorescent contrast agent is preferably indocyanine green solution, with a preset dose range of 0.1 to 0.5 ml and a concentration of 0.125 mg / ml. The constant pressure is achieved by using an infusion pump with pressure feedback or by the doctor manually and evenly pushing the contrast agent, which aims to allow the contrast agent to diffuse naturally along the interstitial space and avoid tissue tearing due to excessive pressure. Then, the acquisition mode of the imaging equipment is switched to fluorescence mode. This mode uses near-infrared light to excite the fluorescent contrast agent and acquires the fluorescence signal of a specific wavelength emitted by it. The system acquires and displays the dynamic image sequence of the diffusion process of the fluorescent contrast agent in tissues such as abscesses and fistulas in real time.

[0126] In the displayed dynamic fluorescence images, areas filled with fluorescence signals are identified. Since the fluorescent contrast agent fills the cavities such as abscesses and fistulas, these areas will show bright fluorescence signals in the images. The boundary of the fluorescence signal area, that is, the position where the fluorescence intensity significantly decreases from strong to the background level, is defined as the real-time morphological boundary of the target lesion. Finally, based on the identified real-time morphological boundary, the final extent of the target lesion is confirmed. In this process, by observing the flow and filling process of the fluorescent contrast agent in the dynamic images, small fistula branches or satellite abscesses that were not visualized on preoperative MRI and other images due to resolution limitations or early infection can be clearly displayed. The system can compare this finally confirmed lesion extent with the preoperative planned extent and update the three-dimensional model, providing a precise target area for subsequent irrigation treatment.

[0127] Through the above steps, a fluorescent contrast agent can be injected into the lesion and its dynamic diffusion characteristics can be used to transform the originally invisible or blurred infected cavities and duct systems into a high-signal area that is clearly visible under fluorescent imaging. This allows for precise delineation of the real-time morphological boundaries of the lesion and, in particular, can effectively reveal occult fistula branches and satellite abscesses that are difficult to detect in conventional preoperative imaging, thus achieving a panoramic and accurate assessment of the true extent of the lesion.

[0128] The target lesion area, which is filled with fluorescent contrast agent, is flushed, and the endpoint of flushing is determined by real-time observation of changes in the intensity of the fluorescent contrast agent signal in the fluorescent images.

[0129] Furthermore, the procedure for determining the end point of the rinse includes:

[0130] The irrigation fluid is injected pulsatially into the target lesion area using a puncture instrument.

[0131] In fluorescence imaging mode, the overall intensity value of the fluorescent contrast agent signal in the target lesion area is monitored in real time;

[0132] Record the overall intensity value decrease curve during the rinsing process;

[0133] When the overall intensity value drops to the level of background noise intensity and the decline curve enters a plateau period, it is determined that the rinsing has reached its end.

[0134] Specifically, through the working channel of the puncture instrument, a pulsed irrigation solution is injected into the target lesion area. The irrigation solution is sterile saline. The pulsed injection is performed by controlling the syringe in an intermittent and rapid pushing manner. The instantaneous kinetic energy generated by the fluid flow agitates and flushes the viscous pus and fluorescent contrast agent mixture in the abscess cavity, thereby improving the irrigation efficiency. During irrigation, the system remains in fluorescence imaging mode and monitors the overall intensity value of the fluorescent contrast agent signal in the target lesion area in real time. The overall intensity value is calculated by defining a zone of interest that includes the entire target lesion area in the real-time fluorescence video stream, and then calculating the average fluorescence intensity of all pixels in the zone of interest. The system continuously records the sequence of changes in the overall intensity value over time during the irrigation process and plots its decline curve. This curve visually reflects the dynamic process of the fluorescent contrast agent being diluted and carried away by the irrigation solution.

[0135] The determination of the flushing endpoint is based on two quantitative conditions that must be met simultaneously. First, the overall intensity value monitored in real time must decrease to the background noise level. This background noise level is determined by collecting the fluorescence intensity of normal tissue far from the lesion area and calculating its average value. Second, the decline curve must enter the plateau phase. The criteria for determining the plateau phase is that, within a continuous period of time, such as 10 seconds, the decrease in the overall intensity value is less than 5% of its initial value. When the system determines that the above two conditions are met simultaneously, it automatically determines that the flushing has reached the endpoint and can issue an audible or visual prompt to inform the operator. This judgment method based on objective signal indicators ensures the thoroughness of the flushing and avoids the possibility of insufficient or excessive flushing that may be caused by traditional experience-based judgment.

[0136] Through the above steps, the dynamic attenuation of fluorescence signal intensity in the lesion area during the irrigation process can be monitored and quantified in real time. This transforms the traditional judgment of the irrigation endpoint, which relies on subjective experience, into a precise judgment based on objective data indicators. When the signal intensity drops to the background level and the attenuation curve enters the plateau phase, it indicates that the contrast agent in the abscess cavity has been completely removed. This ensures the thoroughness and standardization of the irrigation operation and reduces the risk of postoperative infection recurrence due to insufficient irrigation.

[0137] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for image-guided rectal puncture, imaging, and irrigation based on fluorescence contrast enhancement, characterized in that, The method includes the following steps: Preoperative pelvic magnetic resonance images of patients are acquired, input into a pre-trained artificial intelligence segmentation model, processed, and a three-dimensional model containing the target lesion area and key anatomical structures is generated. Based on the three-dimensional model, one or more recommended puncture paths are calculated; During the procedure, real-time ultrasound images are acquired and elastically registered with the three-dimensional model to spatially align the three-dimensional model with the real-time anatomical structure, generating a fused image. The puncture procedure is performed according to the recommended puncture path shown in the fused image to reach the target lesion area; A fluorescent contrast agent is injected into the target lesion area, and the distribution of the fluorescent contrast agent is observed based on the fluorescent image to confirm the extent of the target lesion. The target lesion area, which has been filled with the fluorescent contrast agent, is flushed, and the endpoint of the flushing is determined by real-time observation of the changes in the signal intensity of the fluorescent contrast agent in the fluorescent image.

2. The image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement according to claim 1, characterized in that, The process of generating the 3D model includes: The raw data of the preoperative pelvic magnetic resonance imaging (MRI) images in DICOM format were acquired, and the raw data in DICOM format was preprocessed. The preprocessed image data is input into the pre-trained artificial intelligence segmentation model; The artificial intelligence segmentation model outputs a pixel-level segmentation mask for the target lesion area and key anatomical structures. Based on the pixel-level segmentation mask, three-dimensional reconstruction is performed to generate a three-dimensional model containing the target lesion area and key anatomical structures.

3. The image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement according to claim 1, characterized in that, The calculation process for the puncture path includes: In the three-dimensional model, the starting region and the target region for puncture are defined; In the three-dimensional model, one or more critical anatomical structures that need to be avoided are marked; Based on a preset path planning algorithm, a path search is performed between the starting region and the target region; During the path search process, the critical anatomical structure to be avoided is introduced as a constraint condition, and one or more paths that maintain a safe distance from the critical anatomical structure are calculated as candidate paths. The candidate paths are scored based on path length and safety margin, and one or more recommended puncture paths are output based on the scoring results.

4. The image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement according to claim 1, characterized in that, The process for generating the fused image includes: Real-time acquisition of the spatial pose of the ultrasound probe in the patient coordinate system; Based on the spatial pose of the ultrasound probe, the real-time ultrasound image is mapped onto the patient coordinate system to form real-time three-dimensional ultrasound data. Extract the corresponding set of anatomical feature points from the real-time three-dimensional ultrasound data and the three-dimensional model; The spatial transformation relationship between the three-dimensional model and the real-time three-dimensional ultrasound data is calculated using a non-rigid iterative nearest point algorithm. Based on the calculated spatial transformation relationship, the three-dimensional model is elastically deformed to achieve spatial alignment with the anatomical structures in the real-time three-dimensional ultrasound data. The aligned 3D model is overlaid with the real-time 3D ultrasound data to generate the fused image.

5. The image-guided rectal puncture, imaging, and irrigation method based on fluorescence contrast enhancement according to claim 4, characterized in that, The process for generating the real-time three-dimensional ultrasound data includes: Acquire two-dimensional real-time ultrasound image frames and their corresponding spatial pose matrices; Based on the spatial pose matrix, the spatial position and imaging plane orientation of each frame of the two-dimensional real-time ultrasound image in the patient coordinate system are determined; Constrained by the spatial position and imaging plane direction, each frame of the two-dimensional real-time ultrasound image is used as a slice of data and filled into a common three-dimensional voxel grid. Null interpolation is performed on the regions in the three-dimensional voxel mesh that are not filled by the slice data to generate complete, spatially continuous real-time three-dimensional ultrasound data.

6. The image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement according to claim 4, characterized in that, The process of elastically deforming the three-dimensional model includes: Obtain the spatial transformation relationship calculated by the non-rigid iterative nearest point algorithm, wherein the spatial transformation relationship is displacement field data used to drive the deformation of the three-dimensional model; Based on the deformation parameters, a displacement vector is calculated for each vertex of the three-dimensional model using a spatial interpolation algorithm; Based on the calculated displacement vector, the spatial position of each vertex of the three-dimensional model is moved, thereby completing the overall elastic deformation of the three-dimensional model; The anatomical contour of the deformed 3D model is compared with the anatomical structure in the real-time 3D ultrasound data to verify the accuracy of spatial alignment.

7. The image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement according to claim 1, characterized in that, The procedure for performing a puncture operation to reach the target lesion area based on the recommended puncture path shown in the fused image includes: On the real-time navigation view of the fused image, the recommended puncture path and the real-time position and orientation of the puncture instrument are rendered synchronously. Adjust the spatial orientation of the puncture instrument so that its projection in the real-time navigation view coincides with the extension line of the recommended puncture path; Insert the needle along the recommended puncture path, and monitor the deviation between the tip of the puncture instrument and the recommended puncture path in real time during the needle insertion process; If the detected deviation exceeds the preset tolerance, the needle insertion direction is adjusted to correct the deviation until the tip of the puncture instrument reaches the target lesion area.

8. The image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement according to claim 1, characterized in that, The process for confirming the extent of the target lesion includes: A preset dose of fluorescent contrast agent is slowly injected at constant pressure through the channel of the puncture instrument that has reached the target lesion area; Switch the imaging mode to fluorescence mode to acquire and display dynamic images of the diffusion process of the fluorescent contrast agent in the tissue in real time; In the dynamic image, the region filled with fluorescence signal is identified, and the boundary of the region is the real-time morphological boundary of the target lesion; Based on the real-time morphological boundaries, the final extent of the target lesion is confirmed, wherein the dynamic images are used to identify fistula branches or satellite abscesses that were not anticipated in the preoperative images.

9. The image-guided rectal puncture and irrigation method based on fluorescence contrast enhancement according to claim 1, characterized in that, The process for determining the end point of rinsing includes: The irrigation fluid is pulsed into the target lesion area using the puncture instrument. In fluorescence imaging mode, the overall intensity value of the fluorescent contrast agent signal in the target lesion area is monitored in real time; Record the decrease curve of the overall strength value during the rinsing process; When the overall intensity value drops to the level of background noise intensity and the decline curve enters a plateau period, it is determined that the rinsing has reached its end.

Images