Endoscope-assisted modified miccoli precise treatment system for thyroid cancer with lateral neck lymph node metastasis

Abstract

The application discloses a thyroid cancer lateral neck lymph node metastasis endoscope-assisted modified Miccoli precise treatment system, and the system is operated through the following method, the method comprises the following steps: constructing a patient individualized three-dimensional anatomical model through preoperative multi-modal image fusion, and extracting the spatial topological relationship of the cervical sheath, accessory nerve, cervical plexus branch and lymphatic drainage path; planning a minimally invasive access path based on the three-dimensional anatomical model, and setting a double-channel puncture point at the posterior edge of the sternocleidomastoid muscle and 2 cm above the clavicle, which is respectively used for inserting a 30° oblique vision endoscope and a variable-angle ultrasonic aspirator; collecting tissue displacement data in real time through an electromagnetic navigation probe during the operation, and dynamically registering the preoperative model to generate a corrected navigation guide map; and the application aims to solve the problems of insufficient cleaning range, unclear anatomical level, limited operation space and disconnection between intraoperative real-time navigation and postoperative verification of the traditional Miccoli operation in the treatment of lateral neck (II-IV area) lymph node metastasis.

Description

Technical Field

[0001] This invention relates to the technical field of information technology, specifically to the laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis in thyroid cancer. Background Technology

[0002] In the surgical treatment of lateral cervical lymph node metastasis in thyroid cancer, the core technical challenge lies in how to minimize surgical trauma and improve the accuracy of lymph node dissection while ensuring tumor radicality. This challenge involves clear exposure and real-time identification of the complex anatomical structures in the lateral cervical region. However, traditional open surgery may affect the patient's postoperative appearance and recovery experience due to the large incision. In contrast, fully laparoscopic techniques in the lateral cervical region are limited by the narrow operating space and insufficient instrument freedom, which may affect the identification and protection of key neurovascular structures.

[0003] In addition, although the Miccoli procedure achieves central dissection through a small incision and endoscopic assistance, its original approach and operation path have limited coverage of lymph nodes in zones II, III, and IV when dealing with lateral cervical lymph node metastases. If it is directly used, it may result in insufficient dissection range or unclear anatomical layers, affecting the tumor control effect.

[0004] Furthermore, when introducing laparoscopic assistance for modified procedures, there are challenges in ensuring intraoperative visual stability and instrument coordination. If the light source angle, lens direction, and instrument movement are not optimized synchronously, surgical efficiency may be reduced and the risk of collateral damage may be increased.

[0005] Ultimately, when precise lymph node clearance is guided by the fusion of preoperative image assessment and intraoperative navigation information, image registration errors or individual anatomical variations may lead to positioning deviations. Furthermore, the existing system lacks a closed-loop verification mechanism for real-time feedback on lymph node clearance status, which may affect the consistency between postoperative pathological assessment and clinical decision-making.

[0006] This complex issue spans the entire process from surgical design and anatomical identification to operational execution and efficacy verification. It involves the multidimensional synergy of minimally invasive techniques, image navigation, and surgical precision, and directly affects the safety and effectiveness of treatment for lateral neck metastases of thyroid cancer. Summary of the Invention

[0007] This invention provides a modified Miccoli precision treatment system for lateral cervical lymph node metastasis in thyroid cancer, with the aim of solving the problems of insufficient dissection range, unclear anatomical layers, limited operating space, and disconnect between intraoperative real-time navigation and postoperative verification when the traditional Miccoli procedure is used to treat lateral cervical (II-IV) lymph node metastasis.

[0008] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0009] A modified Miccoli precision treatment system for lateral cervical lymph node metastasis in thyroid cancer, assisted by endoscopy, comprises: a system body, which operates through the following methods: constructing an individualized three-dimensional anatomical model of the patient through preoperative multimodal image fusion, extracting the spatial topological relationships of the cervical sheath, accessory nerve, cervical plexus branches, and lymphatic drainage pathways; planning a minimally invasive approach based on the three-dimensional anatomical model, and setting dual-channel puncture points at the posterior border of the sternocleidomastoid muscle and 2 cm above the clavicle, respectively for inserting a 30° oblique endoscope and a variable-angle ultrasonic aspirator; during the operation, acquiring tissue displacement data in real time using an electromagnetic navigation probe, dynamically registering it with the preoperative model to generate a corrected navigation guidance map; and using a near-infrared fluorescence module integrated into the endoscope light source to excite the preoperatively injected indocyanine green-marked area to identify potential metastases. Lymph node boundaries; the endoscopic image stream and navigation guidance map are aligned pixel-by-pixel, and high-risk areas are segmented using a convolutional neural network and superimposed onto the main field of view; if the overlap between the segmented area and the preoperative predicted lymph node distribution is higher than 85%, an automatic tracking mode is activated, driving the robotic arm to adjust the lens focal length and angle to keep the target structure centered; after completing the en bloc resection of the lymph nodes, the resected specimen is placed on the intraoperative rapid pathology stage, and the surface reflectance spectrum of the resection margin is scanned using a miniature spectral analyzer to determine whether there are residual tumor signals; if abnormal spectral features are detected, a secondary cleaning command is triggered, reactivating the adjacent high-risk subregions in the navigation guidance map; finally, the entire surgical process video, navigation trajectory, spectral analysis results, and pathological feedback data are packaged, encrypted, and written into a medical blockchain node to form a traceable closed-loop record of treatment.

[0010] In one aspect of this disclosure, the step of constructing a patient-specific three-dimensional anatomical model through preoperative multimodal image fusion and extracting the spatial topological relationships of the cervical sheath, accessory nerve, cervical plexus branches, and lymphatic drainage pathways includes:

[0011] The patient's preoperative enhanced CT, MRI and PET-CT raw DICOM sequences were acquired, and the three sets of images were voxel aligned using a non-rigid registration algorithm to generate a fused image volume.

[0012] U-Net semantic segmentation was performed on the fused image volume, and the internal jugular vein, common carotid artery, vagus nerve, accessory nerve, C2–C4 branches of the cervical plexus and Level II–IV lymph node capsules were labeled respectively.

[0013] Based on the annotation results, a surface mesh model with vector direction is constructed, and elastic modulus parameters are assigned to each structure.

[0014] The tissue deformation of the head and neck under different body positions was simulated by finite element simulation, and the maximum displacement field was output as the intraoperative registration compensation factor.

[0015] The displacement field is superimposed on the static mesh model to generate a dynamic anatomical model;

[0016] The intersection of the accessory nerve course and the omohyoid muscle was extracted as a key anchoring landmark to establish a local coordinate system.

[0017] Based on this coordinate system, the spatial bounding boxes of each lymph node subregion are defined, and a structured anatomical topology map is generated.

[0018] In one aspect of this disclosure, the step of planning a minimally invasive approach based on the three-dimensional anatomical model and setting a dual-channel puncture point at the posterior border of the sternocleidomastoid muscle and 2 cm above the clavicle, for inserting a 30° oblique laparoscope and a variable-angle ultrasonic aspirator respectively, includes:

[0019] The instrument entry path is simulated in a dynamic anatomical model to avoid the perforating branches of the external jugular vein and the cutaneous branches of the supraclavicular nerve.

[0020] The first puncture point was determined to be located at the junction of the middle and lower 1 / 3 of the posterior border of the sternocleidomastoid muscle, with the depth controlled in the superficial fascia layer, for the placement of a 30° rigid strabismus endoscope with a diameter of 5 mm.

[0021] The second puncture point is set 2cm above the clavicle and 4cm lateral to the midline, penetrating the superficial layer of the deep cervical fascia but not breaking through the prevertebral fascia, and is used to insert an ultrasonic aspirator with a terminal deflection of ±45°.

[0022] The angle between the two puncture channels is maintained at 60°–75° to ensure the stability of the operating triangle;

[0023] A cross mark is projected onto the skin surface at the puncture point location using a laser locator;

[0024] The puncture cannula uses a spiral fixing structure with a sealing valve and has anti-slip threads on the outer wall. After being screwed in, it forms a mechanical interlock with the subcutaneous tissue.

[0025] The endoscope and suction device are connected to the suspension bracket via a universal joint, and the bracket base is attached to the side rail of the operating table.

[0026] In one aspect of this disclosure, the step of acquiring tissue displacement data in real time during surgery using an electromagnetic navigation probe, dynamically registering it with a preoperative model, and generating a corrected navigation guidance map includes:

[0027] A miniature electromagnetic sensor is embedded in the tip of the endoscope and the working end of the suction device, with the sampling frequency set to 100Hz.

[0028] The electromagnetic field generator above the surgical area is activated simultaneously to establish a three-dimensional positioning space;

[0029] The probe spatial coordinates are collected every 5 seconds, and the nearest point is iteratively matched with the corresponding anatomical points in the preoperative model.

[0030] Calculate the current tissue deformation matrix and update the positions of each structure in the dynamic anatomical model;

[0031] The corrected cervical sheath and accessory nerve pathway are overlaid onto the main laparoscopic view with a semi-transparent green outline;

[0032] When the probe approaches a preset safe distance threshold (e.g., 1.5mm from the accessory nerve), an audible and visual alarm is triggered and the suction power output is frozen.

[0033] The navigation guide map is streamed to the main control processor in H.265 encoding with a latency of less than 80ms.

[0034] In one aspect of this disclosure, the step of identifying the boundaries of potential metastatic lymph nodes by exciting a preoperatively injected indocyanine green-labeled region using a near-infrared fluorescence module integrated with a laparoscopic light source includes:

[0035] 0.5 mg / kg indocyanine green was injected peritumorally 24 hours before surgery to enrich it in metastatic lesions via lymphatic drainage;

[0036] The cavity mirror light source module incorporates a 785nm laser diode and an 820nm bandpass filter.

[0037] When near-infrared mode is activated, the CCD sensor switches to the 900–1700nm response band;

[0038] Fluorescence intensity distribution maps were collected, and the Otsu thresholding method was used to segment the bright areas.

[0039] The segmentation mask is aligned with the white light image through an affine transformation to generate a fused view.

[0040] Perform morphological closing operations on the merged view to eliminate noise speckles;

[0041] Output the continuous boundary curve as a reference line for lymph node resection and overlay it onto the picture-in-picture window in the lower right corner of the main field of view.

[0042] In one aspect of this disclosure, the step of pixel-level aligning the endoscopic image stream with the navigation guidance map, segmenting high-risk regions using a convolutional neural network, and overlaying them onto the main field of view includes:

[0043] Deploy a lightweight ResNet-18 backbone semantic segmentation network with an input resolution of 512×512;

[0044] The training dataset contains intraoperative images of 2000 cases with labeled lateral cervical lymph node metastases;

[0045] The network outputs a four-channel probability map, corresponding to normal fat, fibrous tissue, suspicious lymph nodes, and blood vessels, respectively.

[0046] Bilinear interpolation is used to map the probability map back to the original image size;

[0047] Set the threshold for suspicious lymph node pathways to 0.72 and generate a binary mask;

[0048] The mask is overlaid on the real-time cavity mirror image using a red semi-transparent color gradation;

[0049] Each frame takes no more than 35ms to process, ensuring smooth video playback.

[0050] In one aspect of this disclosure, the step of placing the resected specimen on an intraoperative rapid pathology stage after en bloc lymph node resection and scanning the surface reflectance spectrum of the resection margin with a miniature spectrometer to determine whether there is residual tumor signal includes:

[0051] The stage has a built-in temperature control module to maintain the specimen temperature at 4℃±0.5℃;

[0052] The spectrometer emits a broadband white light in the 400–1000 nm range, and the receiver uses an optical fiber array to collect diffuse reflection signals.

[0053] Ten spectra were collected for each square millimeter region, and the average was then input into a pre-trained SVM classifier.

[0054] The classifier is trained on 200 samples with known cutting edge states and outputs a label of "clean" or "positive".

[0055] If any region is identified as positive, then that coordinate is marked in the three-dimensional specimen model;

[0056] Reverse map the coordinates to the surgical area navigation map and highlight the sub-areas that need additional cleaning.

[0057] The system automatically generates suggested paths for secondary operations and pushes them to the main screen.

[0058] In one aspect of this disclosure, the system body includes:

[0059] The individualized anatomical modeling unit is configured to construct an individualized three-dimensional anatomical model of the patient through preoperative multimodal image fusion and extract the spatial topological relationships of the cervical sheath, accessory nerve, cervical plexus branches and lymphatic drainage pathways.

[0060] The dual-channel puncture positioning unit is configured to plan the minimally invasive approach path based on the three-dimensional anatomical model and set dual-channel puncture points at the posterior border of the sternocleidomastoid muscle and 2cm above the clavicle, respectively for inserting a 30° oblique laparoscope and a variable-angle ultrasonic aspirator.

[0061] The electromagnetic navigation registration unit is configured to perform the following steps: during the operation, the electromagnetic navigation probe collects tissue displacement data in real time, combines it with the preoperative model for dynamic registration, and generates a corrected navigation guidance map.

[0062] The near-infrared fluorescence recognition unit is configured to identify the boundaries of potential metastatic lymph nodes by exciting the preoperatively injected indocyanine green-labeled area using the near-infrared fluorescence module integrated with the endoscopic light source.

[0063] The intelligent image fusion unit is configured to perform pixel-level alignment between the endoscope image stream and the navigation guidance map, segment out high-risk areas through a convolutional neural network, and overlay them onto the main field of view;

[0064] The intraoperative margin verification unit is configured to, after completing the en bloc resection of the lymph nodes, place the resected specimen on the intraoperative rapid pathology stage and scan the surface reflectance spectrum of the margin using a miniature spectral analyzer to determine whether there are residual tumor signals.

[0065] The treatment closed-loop recording unit is configured to package and encrypt the entire surgical process video, navigation trajectory, spectral analysis results, and pathological feedback data, and write them into a medical blockchain node to form a traceable treatment closed-loop record.

[0066] Compared with the prior art, the present invention has the following beneficial effects:

[0067] This invention mainly includes Attached Figure Description

[0068] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0069] Figure 1 This is one of the flowcharts for the laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to the present invention.

[0070] Figure 2 This is the second flowchart of the laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to the present invention.

[0071] Figure 3 This is the third flowchart of the laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to the present invention. Detailed Implementation

[0072] The present invention will be further described below with reference to embodiments. These embodiments are merely some, not all, of the embodiments of the present invention. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the protection scope of the present invention.

[0073] Please see Figures 1-3 As shown, this embodiment discloses a laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer, which is operated in the following way:

[0074] In the preoperative preparation phase, the patient's preoperative enhanced CT, MRI, and PET-CT raw DICOM sequence data were first acquired. These three sets of image data originated from different modal imaging devices, possessing varying spatial resolutions and tissue contrasts. A voxel-level alignment of the three images was performed using a non-rigid registration algorithm (such as a B-spline-based free deformation model) to generate a fused image volume. This fused image volume preserved the bony anatomy of CT, the soft tissue resolution of MRI, and the metabolic activity information of PET-CT. Subsequently, the fused image volume was processed using the U-Net semantic segmentation network to annotate the internal jugular vein, common carotid artery, vagus nerve, accessory nerve, C2–C4 branches of the cervical plexus, and the lymph node capsules of Level II–IV regions. These annotations were output as 3D point clouds or curved meshes, and each anatomical structure was assigned corresponding elastic modulus parameters for subsequent mechanical simulations. Next, a head and neck biomechanical model was constructed based on the finite element method to simulate tissue deformation in commonly used surgical positions such as supine, head tilted back at 15°, and shoulder height increased by 10cm, calculating the maximum displacement field of each key structure. This displacement field, acting as a compensating factor for intraoperative dynamic registration, is superimposed on the static mesh model to generate a dynamic anatomical model. In this model, the intersection of the accessory nerve and the omohyoid muscle is defined as a key anchoring landmark, and a local coordinate system is established based on this landmark. According to this coordinate system, spatial bounding boxes are defined for each subregion of Level IIa, IIb, III, and IV, forming a structured anatomical topology map. This map includes the relative positions, distances, angles, and spatial occlusion relationships between the structures.

[0075] During the surgical approach planning phase, based on the aforementioned dynamic anatomical model, the instrument entry path is simulated in computer-aided surgical planning software. The path planning must avoid vulnerable structures such as the external jugular vein perforators and the cutaneous branches of the supraclavicular nerve. The first puncture point is located at the junction of the middle and lower thirds of the posterior border of the sternocleidomastoid muscle, above the superficial layer of the deep cervical fascia, with a depth controlled at 8–10 mm, for inserting a 5 mm diameter, 30° rigid oblique endoscope. The second puncture point is located 2 cm above the clavicle and 4 cm lateral to the midline. This point penetrates the superficial layer of the deep cervical fascia but does not break through the prevertebral fascia, with a depth of approximately 12–15 mm, for inserting an ultrasonic aspirator with a terminal deflection of ±45°. The angle between the two puncture channels in three-dimensional space is maintained between 60° and 75° to ensure the stability of the operating triangle and avoid instrument interference. On the skin surface, the two puncture points are projected as crosshairs using a laser locator, with an accuracy controlled at ±0.5 mm. The cannula used for puncture has a spiral fixation structure with a sealing valve and continuous anti-slip threads on the outer wall. When screwed in, it rotates and advances to form a mechanical engagement with the subcutaneous tissue, preventing intraoperative displacement. The endoscope and ultrasonic aspirator are connected to the suspension bracket via universal joints. The bracket base is fixed to the side rail of the operating table by magnetic attraction or snap-fit, and can be adjusted in three degrees of freedom in terms of height, front-back, and left-right position to accommodate patients of different body types.

[0076] Once the intraoperative procedure begins, the electromagnetic navigation system is activated. Miniature electromagnetic sensors are pre-embedded in the endoscope tip and the working end of the ultrasonic aspirator, with a sampling frequency set to 100Hz. An electromagnetic field generator installed above the surgical area establishes a 30cm cube-shaped positioning space, covering the entire lateral neck surgical area. The system acquires the probe's spatial coordinates every 5 seconds and performs iterative nearest point (ICP) matching with corresponding anatomical points in the preoperative dynamic anatomical model (such as the bifurcation of the internal jugular vein and the exit point of the accessory nerve), calculating the current tissue deformation matrix. This matrix is ​​used to update the position of each structure in the dynamic anatomical model in real time, generating a corrected navigation guidance map. The corrected cervical sheath and accessory nerve pathways are superimposed onto the main laparoscopic view as semi-transparent green outlines, achieved using pixel-level affine transformation to ensure strict alignment of the anatomical structures with the actual field of view. When the probe approaches a preset safe distance threshold (e.g., 1.5mm for the accessory nerve and 2.0mm for the internal jugular vein), the system triggers an audible and visual alarm and freezes the power output of the ultrasonic aspirator via a hardware signal line to prevent accidental injury. The navigation guidance map is transmitted to the main control processor via Gigabit Ethernet in H.265 encoding format, with end-to-end latency controlled within 80ms, meeting real-time requirements.

[0077] Concurrently, 0.5 mg / kg indocyanine green (ICG) was injected around the primary tumor 24 hours preoperatively to enrich it in metastatic lymph nodes via lymphatic drainage. During the procedure, the near-infrared fluorescence function of the endoscopic light source module was activated. This module uses a built-in 785 nm laser diode as the excitation source, coupled with an 820 nm bandpass filter to remove stray light. The CCD sensor was switched to the 900–1700 nm response band to acquire fluorescence images. The system used the Otsu adaptive thresholding method to segment the fluorescence intensity distribution map and extract the bright areas. Subsequently, an affine transformation was used to align the fluorescence segmentation mask with the white light image to generate a fused view. To eliminate noise, a morphological closing operation (structuring element: 5×5 circular kernel) was performed on the fused view, outputting a continuous boundary curve. This curve served as a reference line for lymph node resection, superimposed in a picture-in-picture format in the lower right corner of the main field of view. The window size was 1 / 4 of the original image, and its position was fixed and did not change with the movement of the main field of view.

[0078] At the image processing level, a lightweight ResNet-18 was deployed as the backbone of the semantic segmentation network, with an input image resolution of 512×512 pixels. During training, the network used a dataset of 2000 intraoperative images annotated by experienced head and neck surgeons, with annotation categories including normal fat, fibrous tissue, suspicious lymph nodes, and blood vessels. The network output is a four-channel probability map, with each channel representing the probability of presence for the corresponding category. The system sets the activation threshold for the suspicious lymph node channel to 0.72; pixels exceeding this value are identified as high-risk areas, generating a binary mask. This mask is mapped back to the original endoscopic image size (typically 1920×1080) using bilinear interpolation and overlaid on the real-time image with a red semi-transparent color gradation (40% transparency). The image processing module runs on a dedicated GPU accelerator card, with a processing time of no more than 35ms per frame, ensuring a video frame rate above 25fps. When the system detects a high-risk area with a higher overlap of 85% with the preoperative predicted lymph node distribution (calculated using the Dice coefficient), it automatically initiates tracking mode. At this point, the main control processor sends commands to the robotic arm via the CAN bus to adjust the focal length and angle of the endoscope lens, ensuring that the target structure remains centered in the image. The robotic arm's range of motion is limited to ±10° pitch and ±15° yaw to prevent excessive movement that could lead to loss of field of view.

[0079] After en bloc lymph node resection, the excised specimen was immediately placed on an intraoperative rapid pathology stage. This stage incorporated a Peltier temperature control module, using PID feedback to maintain the specimen temperature at 4℃±0.5℃ to prevent tissue degradation. A miniature spectrometer was mounted 5cm above the stage, emitting a broad-spectrum white light of 400–1000nm, illuminating an area of ​​10mm×10mm. A 64-channel fiber optic array, evenly spaced at 1mm intervals, collected diffuse reflectance signals from the specimen's cut edge surface. Ten spectra were continuously acquired for each square millimeter area, and the average value was input into a pre-trained support vector machine (SVM) classifier. This classifier was trained based on spectral data from 200 specimens with known cut edge states (negative or positive), using a radial basis function (RBF) kernel, a penalty parameter C=10, and γ=0.01. The classifier output a label of "clean" or "positive". If any region is identified as positive, the system marks the coordinates (x, y, z) in the 3D specimen model and transforms these coordinates to the corresponding anatomical location in the surgical area navigation map using an inverse mapping algorithm. The lymph node subregion where this location is located is highlighted as a flashing yellow area, and a secondary operation suggestion window pops up on the right side of the main screen, displaying the recommended supplementary dissection path and the required adjustment of the puncture angle. After the operator confirms, the system reactivates the navigation guidance map for that subregion and restores the power output of the ultrasound aspirator.

[0080] The entire surgical procedure's data flow is managed centrally by a closed-loop treatment recording unit. This unit includes a video capture card, a navigation trajectory recorder, a spectral analysis log module, and a pathology feedback interface. All data is encrypted using AES-256, packaged according to the HL7FHIR standard format, and transmitted to a medical blockchain node via the hospital's intranet. This node is deployed on a private blockchain that meets the Level 3 requirements of the National Medical Information Security Protection System. Each block contains a timestamp, operator ID, device serial number, and hash checksum. Once written, the data is immutable, forming a complete closed-loop treatment record that can be used for postoperative review, quality control, and multi-center research.

[0081] The aforementioned units are interconnected via a high-speed data bus. The individualized anatomical modeling unit runs on the preoperative workstation and interfaces with the PACS system via the DICOM protocol; the dual-channel puncture positioning unit is integrated into the surgical navigation host and communicates with the laser positioning device via USB 3.0; the electromagnetic navigation registration unit includes an electromagnetic field generator, sensor interface board, and registration engine, and connects to the main control processor via PCIe x4; the near-infrared fluorescence recognition unit shares the same image processing board with the endoscope host and receives sensor data via the MIPICSI-2 interface; the intelligent image fusion unit is deployed on a GPU server and outputs superimposed images to the surgical monitor via DP1.4; the intraoperative margin verification unit communicates with the stage temperature control module via RS-485 and synchronizes results with the main control system via TCP / IP; the treatment closed-loop recording unit is connected to the hospital data center via a fiber optic channel. All hardware devices have passed IEC60601-1 medical electrical equipment safety certification, and the software system complies with the YY / T0664 medical device software lifecycle process standard.

[0082] In a clinical application, a 52-year-old female patient with papillary thyroid carcinoma and right-sided Level III lymph node metastasis received the treatment described in this invention. Preoperative image fusion showed that the metastatic lesion was adjacent to the accessory nerve, making safe exposure difficult with the traditional Miccoli procedure. During the operation, a dual-channel puncture point was set up according to the above procedure. After the endoscope was successfully inserted, the navigation system displayed the location of the accessory nerve in real time. The ultrasonic aspirator avoided the nerve under the guidance of the green outline and completely removed the lymph node mass. Near-infrared fluorescence showed clear boundaries, and the AI ​​segmentation area overlapped with the preoperative prediction by 89%, triggering automatic tracking. Postoperative spectral analysis of the resection margin revealed a small positive lesion. The system reversed the localization to the Level IIb subregion, and a second dissection was performed, which was negative again. All data was encrypted and uploaded to the blockchain. No recurrence was observed during the 12-month follow-up. This embodiment fully verifies the technical feasibility and clinical value of this invention in terms of anatomical identification, operational safety, thorough resection, and data traceability.

[0083] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principles of this invention are further supplemented below with a specific application scenario.

[0084] Step 1: In the preoperative individualized anatomical modeling process, the patient's preoperative enhanced CT, MRI, and PET-CT raw DICOM data are first imported into the individualized anatomical modeling unit. This unit uses a B-spline-based non-rigid registration algorithm to perform voxel-level spatial alignment of the three sets of images, generating a fused image volume. In the fused image volume, CT provides high-contrast contours of bony structures such as the transverse processes of the cervical vertebrae and the hyoid bone; MRI presents the soft tissue boundaries of the intracervical vascular bundles; and PET-CT marks metabolically active areas, corresponding to potential metastatic lymph nodes. Subsequently, the U-Net semantic segmentation network performs pixel-level classification on the fused image volume, outputting a three-dimensional curved mesh of the internal jugular vein, common carotid artery, vagus nerve, accessory nerve, C2–C4 branches of the cervical plexus, and the lymph node capsules of Level II–IV regions. Each mesh vertex is assigned an elastic modulus parameter, for example, 0.8 MPa for the accessory nerve and 1.2 MPa for the internal jugular vein, to reflect their biomechanical properties. Next, the finite element solver, under boundary conditions of a supine position, head tilted back 15°, and shoulder pad height of 10cm, calculated the displacement field of each structure under gravity and postural traction. This displacement field, as a dynamic compensation factor, was superimposed on the static mesh to generate a dynamic anatomical model. In this model, the intersection of the accessory nerve and the omohyoid muscle was extracted as a fixed anchor point, and a local coordinate system was established with this as the origin. Then, the spatial bounding boxes of Level IIa, IIb, III, and IV subregions were defined, forming a structured topological map containing relative positions, occlusion relationships, and safety distance thresholds. This process automatically acquired images from the PACS system via the DICOM protocol and was completed on the preoperative workstation, ensuring that the model accuracy error was less than 1.0mm.

[0085] Step Two: In the dual-channel puncture localization stage, the surgical navigation host loads the aforementioned dynamic anatomical model and runs a path avoidance algorithm in the computer-aided planning interface. This algorithm uses the external jugular vein perforators and the supraclavicular cutaneous branches as prohibited areas, searching for a collision-free path from the skin surface to the target lymph node area. The first puncture point is ultimately determined to be located at the junction of the middle and lower 1 / 3 of the posterior border of the sternocleidomastoid muscle. This point corresponds to a depth of 8–10 mm above the superficial layer of the deep cervical fascia in the dynamic model, avoiding damage to the cervical plexus cutaneous branches. The second puncture point is located 2 cm above the clavicle and 4 cm lateral to the midline, penetrating the superficial layer of the deep cervical fascia but terminating in the prevertebral fascial space, at a depth of 12–15 mm, ensuring that the working end of the ultrasonic aspirator can reach the Level III area without touching the brachial plexus. The three-dimensional angle between the two points is maintained at 68° through vector calculation, meeting the operational triangulation stability requirements. The laser locator receives the puncture coordinates via USB 3.0 and projects a crosshair onto the skin, with the localization error controlled within ±0.5 mm. During puncture, the spiral cannula is screwed in at a preset angle. Its continuous anti-slip threads on the outer wall form a mechanical interlock with the subcutaneous fascia fibers to prevent slippage during the operation. The endoscope and ultrasonic aspirator are connected to the suspension bracket via universal joints. The bracket base is magnetically fixed to the side rail of the operating table, allowing for fine adjustments in the X, Y, and Z directions to accommodate different neck lengths and body shapes.

[0086] Step 3: After the intraoperative electromagnetic navigation registration is initiated, the miniature electromagnetic sensor embedded in the endoscope tip and the working end of the ultrasonic aspirator sends a position signal to the electromagnetic field generator at a frequency of 100Hz. The electromagnetic field generator establishes a uniform magnetic field within a 30cm cube, and the sensor receives the induced voltage and calculates the spatial coordinates. Every 5 seconds, the system performs ICP matching between the current probe coordinates and feature points in the dynamic anatomical model, such as the bifurcation point of the internal jugular vein and the exit point of the accessory nerve, to calculate the current tissue deformation matrix. This matrix is ​​used to update the spatial position of each structure in the model in real time, generating a corrected navigation guidance map. The navigation guidance map is processed by the GPU to perform pixel-level affine transformation, accurately superimposing the accessory nerve path as a semi-transparent green outline onto the main endoscope screen, with an alignment error of less than 0.7 pixels. When the ultrasonic aspirator probe is less than 1.5mm away from the accessory nerve, the hardware interrupt circuit immediately cuts off its power output and triggers an audible and visual alarm. The entire navigation data stream is H.265 encoded and compressed, and transmitted to the main control processor via gigabit Ethernet, with a measured end-to-end latency of 72ms, meeting clinical real-time requirements.

[0087] Step 4: The near-infrared fluorescence recognition unit is activated synchronously during the operation. Because indocyanine green (ICG) was injected 24 hours preoperatively, the ICG concentration in the metastatic lymph nodes was significantly higher than in the surrounding tissue. The endoscopic light source module is switched to near-infrared mode, where a 785nm laser excites the ICG to produce fluorescence at 900–1700nm, and an 820nm bandpass filter removes background light interference. After the CCD sensor acquires the fluorescence image, the Otsu adaptive thresholding method automatically segments the bright areas to generate an initial mask. This mask is aligned with the white light image through an affine transformation matrix to eliminate displacement caused by viewing angle differences. Subsequently, morphological closing operations fill the internal voids with 5×5 circular structuring elements, smoothing the edges and outputting a continuous boundary curve. This curve serves as a resection reference line, fixedly displayed in a picture-in-picture format in the lower right quarter of the main field of view, remaining constant regardless of camera movement, ensuring it is always visible to the surgeon.

[0088] Step 5: The intelligent image fusion unit processes the endoscopic video stream in real time. A lightweight ResNet-18 network receives a 512×512 cropped image, extracts features through convolution and pooling layers, and outputs a four-channel probability map. If a pixel in the suspected lymph node channel has a probability value ≥0.72, it is identified as a high-risk area, and a binary mask is generated. This mask is restored to 1920×1080 resolution using bilinear interpolation and overlaid on the original image with 40% transparency in red. When the system calculates that the Dice coefficient between this mask and the preoperative predicted lymph node distribution reaches 89%, it is determined that the target area highly overlaps, and the tracking mode is automatically triggered. The main control processor sends angle commands to the robotic arm via the CAN bus, adjusting the lens pitch ±8° and yaw ±12° to keep the target structure within ±50 pixels of the image center. The robotic arm movement is constrained by limit switches to prevent exceeding the safe travel range.

[0089] Step Six: Intraoperative Margin Verification. The excised specimen is immediately placed on a rapid pathology stage. The Peltier temperature control module adjusts the current using a PID algorithm to maintain the stage surface temperature at 4℃±0.5℃, inhibiting tissue autolysis. A miniature spectrometer emits a 400–1000nm broadband light spectrum to illuminate a 10mm×10mm area, and a 64-channel fiber optic array acquires diffuse reflectance spectra at 1mm intervals. Each square millimeter area is sampled 10 times, and the average spectrum is input into an SVM classifier. This classifier uses the RBF kernel function and learns the spectral differences between 200 negative and positive margin samples during the training phase, such as the absorption peak intensity at 850nm and the scattering slope at 600nm. If a region is judged as "positive," the system records its (x,y,z) position in the specimen's three-dimensional coordinate system and transforms it to its anatomical location on the surgical navigation map using an inverse mapping algorithm. The Level IIb subzone corresponding to this location immediately flashes brightly on the main screen, and a secondary cleaning suggestion window pops up, prompting the user to adjust the angle of the second puncture point by 5° and shift it 3mm towards the head. After the operator confirms, the system unlocks the ultrasonic aspirator power and reactivates the navigation guidance map for that subzone.

[0090] Step Seven: The treatment closed-loop recording unit synchronously collects multi-source data throughout the entire surgery. The video capture card captures the main screen at 1920×1080@25fps, the navigation trajectory recorder stores the probe coordinate time series, the spectral analysis log module records the results of each resection margin detection, and the pathology feedback interface receives the operator's confirmation signal. All data is processed by the AES-256 encryption engine, encapsulated into JSON objects according to the HL7FHIR standard, and pushed to the private blockchain node through the hospital's intranet. Each block contains a UTC timestamp, the surgeon's ID, the device serial number, and a SHA-256 hash value, and cannot be tampered with after being written. This record can be used for postoperative quality retrospective analysis, multi-center efficacy comparison, and AI model iterative training.

[0091] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer, characterized in that, include: The system itself operates through the following methods: By constructing an individualized three-dimensional anatomical model of the patient through preoperative multimodal image fusion, the spatial topological relationships of the cervical sheath, accessory nerve, cervical plexus branches, and lymphatic drainage pathways were extracted. Based on the three-dimensional anatomical model, a minimally invasive approach was planned, and a dual-channel puncture point was set at the posterior border of the sternocleidomastoid muscle and 2 cm above the clavicle, which were used to insert a 30° oblique laparoscope and a variable-angle ultrasonic aspirator, respectively. During the operation, tissue displacement data is collected in real time using an electromagnetic navigation probe, and dynamic registration is performed in combination with the preoperative model to generate a corrected navigation guidance map. The boundaries of potential metastatic lymph nodes were identified by exciting the preoperatively injected indocyanine green-labeled area using a near-infrared fluorescence module integrated with a laparoscopic light source. The endoscopic image stream is aligned pixel-level with the navigation guidance map, and high-risk areas are segmented using a convolutional neural network and superimposed onto the main field of view. If the overlap between the segmented area and the preoperatively predicted lymph node distribution is greater than 85%, the automatic tracking mode is activated, driving the robotic arm to adjust the lens focal length and angle to keep the target structure centered. After completing the en bloc resection of the lymph nodes, the resected specimen was placed on the intraoperative rapid pathology stage, and the surface reflectance spectrum of the resection margin was scanned by a miniature spectral analyzer to determine whether there were any residual tumor signals. If abnormal spectral features are detected, a secondary cleaning command is triggered to reactivate the adjacent high-risk sub-region in the navigation guidance map; Finally, the entire surgical process video, navigation trajectory, spectral analysis results, and pathological feedback data are packaged, encrypted, and written into the medical blockchain node to form a traceable closed-loop record of treatment.

2. The laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to claim 1, characterized in that, The process involves constructing an individualized three-dimensional anatomical model of the patient through preoperative multimodal image fusion, extracting the spatial topological relationships of the cervical sheath, accessory nerve, cervical plexus branches, and lymphatic drainage pathways, including: The patient's preoperative enhanced CT, MRI and PET-CT raw DICOM sequences were acquired, and the three sets of images were voxel aligned using a non-rigid registration algorithm to generate a fused image volume. U-Net semantic segmentation was performed on the fused image volume, and the internal jugular vein, common carotid artery, vagus nerve, accessory nerve, C2–C4 branches of the cervical plexus and Level II–IV lymph node capsules were labeled respectively. Based on the annotation results, a surface mesh model with vector direction is constructed, and elastic modulus parameters are assigned to each structure. The tissue deformation of the head and neck under different body positions was simulated by finite element simulation, and the maximum displacement field was output as the intraoperative registration compensation factor. The displacement field is superimposed on the static mesh model to generate a dynamic anatomical model; The intersection of the accessory nerve course and the omohyoid muscle was extracted as a key anchoring landmark to establish a local coordinate system. Based on this coordinate system, the spatial bounding boxes of each lymph node subregion are defined, and a structured anatomical topology map is generated.

3. The laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to claim 1, characterized in that: The minimally invasive approach path is planned based on the three-dimensional anatomical model, and a dual-channel puncture point is set at the posterior border of the sternocleidomastoid muscle and 2 cm above the clavicle, for inserting a 30° oblique laparoscope and a variable-angle ultrasonic aspirator, respectively, including: The instrument entry path is simulated in a dynamic anatomical model to avoid the perforating branches of the external jugular vein and the cutaneous branches of the supraclavicular nerve. The first puncture point was determined to be located at the junction of the middle and lower 1 / 3 of the posterior border of the sternocleidomastoid muscle, with the depth controlled in the superficial fascia layer, for the placement of a 30° rigid strabismus endoscope with a diameter of 5mm. The second puncture point is set 2cm above the clavicle and 4cm lateral to the midline, penetrating the superficial layer of the deep cervical fascia but not breaking through the prevertebral fascia, and is used to insert an ultrasonic aspirator with a terminal deflection of ±45°. The angle between the two puncture channels is maintained at 60°–75° to ensure the stability of the operating triangle; A cross mark is projected onto the skin surface at the puncture point location using a laser locator; The puncture cannula uses a spiral fixing structure with a sealing valve and has anti-slip threads on the outer wall. After being screwed in, it forms a mechanical interlock with the subcutaneous tissue. The endoscope and suction device are connected to the suspension bracket via a universal joint, and the bracket base is attached to the side rail of the operating table.

4. The laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to claim 1, characterized in that, The process of acquiring tissue displacement data in real time during the operation using an electromagnetic navigation probe, dynamically registering it with the preoperative model, and generating a corrected navigation guidance map includes: A miniature electromagnetic sensor is embedded in the tip of the endoscope and the working end of the suction device, with the sampling frequency set to 100Hz. The electromagnetic field generator above the surgical area is activated simultaneously to establish a three-dimensional positioning space; The probe spatial coordinates are collected every 5 seconds, and the nearest point is iteratively matched with the corresponding anatomical points in the preoperative model. Calculate the current tissue deformation matrix and update the positions of each structure in the dynamic anatomical model; The corrected cervical sheath and accessory nerve pathway are overlaid onto the main laparoscopic view with a semi-transparent green outline; When the probe approaches the preset safe distance threshold of 1.5mm, an audible and visual alarm is triggered and the suction power output is frozen; The navigation guide map is streamed to the main control processor in H.265 encoding with a latency of less than 80ms.

5. The laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to claim 1, characterized in that, The method of using a near-infrared fluorescence module integrated with a laparoscopic light source to excite the preoperatively injected indocyanine green-labeled area to identify the boundaries of potential metastatic lymph nodes includes: 0.5 mg / kg indocyanine green was injected peritumorally 24 hours before surgery to enrich it in metastatic lesions via lymphatic drainage; The cavity mirror light source module incorporates a 785nm laser diode and an 820nm bandpass filter. When near-infrared mode is activated, the CCD sensor switches to the 900–1700nm response band; Fluorescence intensity distribution maps were collected, and the Otsu thresholding method was used to segment the bright areas. The segmentation mask is aligned with the white light image through an affine transformation to generate a fused view. Perform morphological closing operations on the merged view to eliminate noise speckles; Output the continuous boundary curve as a reference line for lymph node resection and overlay it onto the picture-in-picture window in the lower right corner of the main field of view.

6. The laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to claim 3, characterized in that, The step of aligning the endoscopic image stream with the navigation guidance map at the pixel level, segmenting high-risk areas using a convolutional neural network, and overlaying them onto the main field of view includes: Deploy a lightweight ResNet-18 backbone semantic segmentation network with an input resolution of 512×512; The training dataset contains intraoperative images of 2000 cases with labeled lateral cervical lymph node metastases; The network outputs a four-channel probability map, corresponding to normal fat, fibrous tissue, suspicious lymph nodes, and blood vessels, respectively. Bilinear interpolation is used to map the probability map back to the original image size; Set the threshold for suspicious lymph node pathways to 0.72 and generate a binary mask; The mask is overlaid on the real-time cavity mirror image using a red semi-transparent color gradation; Each frame takes no more than 35ms to process.

7. The laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to claim 1, characterized in that, After en bloc lymph node resection, the resected specimen is placed on an intraoperative rapid pathology stage, and the surface reflectance spectrum of the resection margin is scanned using a miniature spectral analyzer to determine whether there are residual tumor signals, including: The stage has a built-in temperature control module to maintain the specimen temperature at 4℃±0.5℃; The spectrometer emits a broadband white light in the 400–1000 nm range, and the receiver uses an optical fiber array to collect diffuse reflection signals. Ten spectra were collected for each square millimeter region, and the average was then input into a pre-trained SVM classifier. The classifier is trained on 200 samples with known cutting edge states and outputs a label of "clean" or "positive". If any region is identified as positive, then that coordinate is marked in the three-dimensional specimen model; Reverse map the coordinates to the surgical area navigation map and highlight the sub-areas that need additional cleaning. The system automatically generates suggested paths for secondary operations and pushes them to the main screen.

8. The laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to claim 1, characterized in that, The system body includes: The individualized anatomical modeling unit is used to construct an individualized three-dimensional anatomical model of the patient through preoperative multimodal image fusion, and to extract the spatial topological relationships of the cervical sheath, accessory nerve, cervical plexus branches and lymphatic drainage pathways. The dual-channel puncture positioning unit is used to execute the minimally invasive approach path planned based on the three-dimensional anatomical model, and sets dual-channel puncture points at the posterior border of the sternocleidomastoid muscle and 2cm above the clavicle, respectively for inserting a 30° oblique laparoscope and a variable-angle ultrasonic aspirator. The electromagnetic navigation registration unit is used to collect tissue displacement data in real time during the operation via electromagnetic navigation probes, perform dynamic registration with the preoperative model, and generate a corrected navigation guidance map. The near-infrared fluorescence recognition unit is used to perform the excitation of the preoperatively injected indocyanine green-labeled area using the near-infrared fluorescence module integrated with the endoscopic light source to identify the boundaries of potential metastatic lymph nodes. The intelligent image fusion unit is used to perform pixel-level alignment of the endoscope image stream with the navigation guidance map, segment high-risk areas through a convolutional neural network and overlay them onto the main field of view; The intraoperative margin verification unit is used to perform the following procedure after en bloc lymph node resection: placing the resected specimen on the intraoperative rapid pathology stage and scanning the surface reflectance spectrum of the margin with a miniature spectral analyzer to determine whether there are residual tumor signals. The treatment closed-loop recording unit is used to package and encrypt the entire surgical process video, navigation trajectory, spectral analysis results, and pathological feedback data, and write them into the medical blockchain node to form a traceable treatment closed-loop record.

9. The laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to claim 8, characterized in that, The individualized anatomical modeling unit includes a non-rigid registration module, a U-Net semantic segmentation module, a finite element simulation module, and a local coordinate system construction module, which are used to generate dynamic anatomical models and structured anatomical topology maps.

10. The laparoscopic-assisted modified Miccoli precision treatment system for lateral cervical lymph node metastasis of thyroid cancer according to claim 8, characterized in that, The intraoperative margin verification unit includes a temperature-controlled stage, a miniature spectrometer, an SVM classifier, and a coordinate inverse mapping module, which are used to determine the margin status and locate the secondary cleaning area.

Images