Auxiliary treatment system and method for difficult-to-reverse suppurative periapical periodontitis

Abstract

The invention relates to the technical field of dental intelligent adjuvant therapy, and discloses an adjuvant therapy system and method for difficult-to-inverse suppurative periapical periodontitis, and the system comprises a multi-modal image fusion module which fuses CT / OCT images to generate a three-dimensional diagram, segments an infected region through a lightweight CNN, and outputs a probability diagram; the reinforcement learning debridement navigation module is used for optimizing the dentin injury risk by combining a biological membrane thickness dynamic planning path on the basis of a DDPG algorithm; the intelligent medicine control module is used for positioning a catheter according to an infection map, triggering medicine release by a pH / enzyme sensor and adjusting delivery pressure by PID; the biosensor monitoring module is used for detecting inflammatory factors, predicting the recurrence risk through LSTM, and triggering image rescanning and path updating when a threshold value is exceeded. Through multi-modal image fusion, reinforcement learning dynamic debridement navigation, microenvironment response type drug control and time sequence prediction closed-loop feedback, accurate elimination and relapse active inhibition of periapical periodontitis focuses are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of dental intelligent auxiliary treatment, and in particular to an auxiliary treatment system and method for irreversible suppurative periapical periodontitis. Background Art

[0002] Due to the complex anatomy of the root canal system and the presence of drug-resistant biofilms, refractory suppurative apical periodontitis can easily progress to chronic, intractable infection. Traditional treatments often face challenges such as incomplete lesion removal, high postoperative recurrence rates, and significant secondary surgical trauma. With the advancement of precision medicine and intelligent diagnosis and treatment technologies, a closed-loop treatment system integrating multimodal image navigation, dynamic path planning, targeted drug control, and proactive recurrence warning is urgently needed to achieve a paradigm shift in apical periodontitis treatment from an experience-driven to a data-driven approach.

[0003] In the existing technology, the clinical diagnosis and treatment of apical periodontitis mainly relies on debridement surgery guided by cone-beam CT images, combined with manual instrument operation to remove infected tissue, and local drug irrigation or systemic antibiotic treatment to inhibit recurrence of infection after surgery. Some improved solutions use optical coherence tomography (OCT) to assist in the detection of biofilm distribution, or use electrochemical sensors to monitor the concentration of inflammatory factors. Debridement path planning is mostly based on preoperative static imaging data, drug release depends on preset time or dosage strategy, and recurrence risk assessment is achieved through regular imaging review combined with retrospective analysis of clinical symptoms.

[0004] However, existing technologies still have some shortcomings. First, a single imaging modality cannot simultaneously capture the cross-scale characteristics of bone tissue and biofilms, resulting in deviations in debridement range planning. Second, static path planning lacks dynamic feedback of intraoperative changes in biofilm thickness, causing damage to healthy tissue or residual lesions. In addition, empirical drug release is not linked to the infection probability distribution and microenvironmental parameters, resulting in insufficient local drug concentration or excessive diffusion. Finally, discrete monitoring methods cannot predict the risk of recurrence in real time, delaying the opportunity for secondary intervention. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention provides an auxiliary treatment system and method for irreversible suppurative apical periodontitis, which solves the problems of incomplete lesion removal and high recurrence rate caused by data isolation, static operation planning and lack of feedback mechanism in the existing technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an auxiliary treatment system for irreversible suppurative periapical periodontitis, comprising:

[0007] A multimodal image fusion module is used to fuse low-dose cone-beam CT images with optical coherence tomography (OCT) images to generate a three-dimensional fused image of the periapical region. It then segments the infected area using a lightweight convolutional neural network and outputs an infection probability map.

[0008] A reinforcement learning debridement navigation module uses a deep deterministic policy gradient algorithm to dynamically plan the debridement path based on the three-dimensional fusion image and biofilm thickness data detected by real-time OCT, and optimizes the path based on the degree of damage to the dentin caused by instrument operation;

[0009] an intelligent drug control module, which locates the position of the drug delivery catheter according to the infection probability map, triggers drug release through periapical microenvironment parameters detected by a pH sensor and an enzyme activity sensor, and dynamically adjusts the drug delivery pressure based on a PID control algorithm;

[0010] The biosensor monitoring module detects the concentration of inflammatory factors in real time and predicts the risk of infection recurrence through a long-short-term memory network. When the predicted recurrence risk exceeds a preset threshold, the multimodal image fusion module is triggered to rescan the periapical area and update the debridement path of the reinforcement learning debridement navigation module based on the regenerated fusion image.

[0011] Preferably, in the multimodal image fusion module:

[0012] The radiation dose of the low-dose cone-beam CT image is 50% of the conventional dose;

[0013] The OCT image has a scanning accuracy of micrometer level and is used to detect the thickness of biofilm on the root canal surface;

[0014] The three-dimensional fusion image is obtained by weighting formula Generate, where α is the fusion weight coefficient, is the registered OCT data; V CBCT is the three-dimensional voxel data of low-dose cone-beam CT; V fused is the fused three-dimensional image data.

[0015] Preferably, the lightweight convolutional neural network is a MobileNet-V3 model, and the calculation formula for its output infection probability map is:

[0016]

[0017] Among them, DWConv k represents the kth depth-wise separable convolutional layer; σ(·) is the Sigmoid activation function; w k is the weight parameter of the k-th convolution kernel; P inf is the infection probability map of the periapical region; K is the total number of depthwise separable convolutional layers in the lightweight convolutional network.

[0018] Preferably, in the reinforcement learning debridement navigation module:

[0019] The reward function is designed using the Deep Deterministic Policy Gradient algorithm (DDPG):

[0020] r t =β·ΔB t -γ·D t -η·||a t -a t-1 || 2 ;

[0021] Where ΔB t is the change in biofilm thickness at time t; D t is the cumulative degree of dentin damage; a t is the action vector; β, γ, η are weight coefficients.

[0022] The reward function of the deep deterministic policy gradient algorithm is designed as:

[0023]

[0024] Where Δd bio is the reduction in biofilm thickness; Δh k is the depth of dentin damage; β and γ are weight coefficients; is the initial biofilm thickness; h th is the safe damage threshold of dentin; K1 is the total number of dentin partitions;

[0025] Policy network μ θ Update via gradient ascent:

[0026]

[0027] θ is the trainable parameter of the policy network, which is used to generate action a; φ is the trainable parameter of the critic network; s is the state vector; a is the action vector; J is the objective function of policy optimization; is the gradient of the Critic network for action aa; is the gradient of the policy network with respect to the parameter θ;

[0028] The debridement path is updated at intervals of Δt based on the latest biofilm thickness data, wherein Δt is a preset fixed time interval.

[0029] Preferably, the dentin damage depth Δh k The calculation is performed using real-time scanning data of OCT images, and the weight coefficients β=0.6 and γ=0.4.

[0030] Preferably, in the intelligent drug control module:

[0031] The trigger threshold of the pH sensor is pH ≤ 5.5, and the trigger threshold of the enzyme activity sensor is MMP-8 ≥ 50 U / L;

[0032] The pressure regulation formula of the PID control algorithm is:

[0033]

[0034] Where u(t) is the adjusted injection pressure, which is used to control the drug delivery rate; e(t) = r(t) - y(t), which represents the drug release error, r(t) is the target drug dose; y(t) is the real-time release amount; K p is the proportional coefficient; K i is the integral coefficient; K d is the differential coefficient; τ is the integral variable, representing any moment in the time interval [0, t]; It is the first-order derivative of the error signal e(t) with respect to time, and represents the error change rate.

[0035] Preferably, the position of the drug delivery catheter is determined by the centroid coordinates of the infection probability map:

[0036]

[0037] Among them, x, y, z represent the voxel coordinates in the 3D fusion image coordinate system; P inf (x, y, z) represents the probability value of the voxel point with coordinates (x, y, z) belonging to the infected area, and the value range is [0, 1]; Σ represents the traversal summation of all voxel points in the 3D fusion image; x c is the coordinate of the center of mass of the infected area in the x-axis direction, which is used to locate the target position of the drug delivery catheter.

[0038] Preferably, in the biosensor monitoring module:

[0039] The inflammatory factors include IL-6 and CRP, and their concentrations are detected in real time by a flexible biosensor;

[0040] The input of the long short-term memory network is time series data {C II-6 (t),C CRP (t),d bio (t)}, where C II-6 (t) is the interleukin-6 concentration detected at time t, in pg / mL; C CRP (t) is the C-reactive protein concentration detected at time t, in mg / L; d bio (t) is the biofilm thickness on the root canal surface measured in real time by optical coherence tomography (OCT) at time t, in μm, and the recurrence probability P is output. recThe calculation formula is:

[0041] P rec =σ(W p ·h t +b p );

[0042] Among them, h t is the LSTM hidden state; σ(·) is the Sigmoid function; W p is the weight matrix of the fully connected layer; b p is the bias vector of the fully connected layer.

[0043] Preferably, the preset threshold is P rec ≥0.8, and the interval between rescanned fusion images to update the reinforcement learning debridement pathway was consistent with the initial scan interval.

[0044] The present invention also provides an auxiliary treatment method for irreversible suppurative periapical periodontitis, comprising the following steps:

[0045] Low-dose cone-beam CT and optical coherence tomography were used to obtain periapical imaging data, generate three-dimensional fused images, and segment the infected area.

[0046] Based on fused images and real-time biofilm thickness data, a deep reinforcement learning algorithm is used to dynamically plan the debridement path and optimize the damage to dentin caused by instrument manipulation.

[0047] The drug catheter is positioned according to the infection probability map, the microenvironmental parameters are detected by sensors to trigger drug release, and the delivery pressure is dynamically adjusted;

[0048] Real-time monitoring of inflammatory factor concentrations can predict recurrence risk. When the risk exceeds the threshold, the image can be rescanned and the debridement pathway can be updated.

[0049] The present invention provides an auxiliary treatment system and method for irreversible suppurative periapical periodontitis. It has the following beneficial effects:

[0050] 1. By integrating cross-modal data from low-dose cone-beam CT and optical coherence tomography (OCT) with a lightweight convolutional neural network, we achieve precise segmentation and three-dimensional localization of periapical infection areas. This invention overcomes the limitations of a single imaging modality, effectively distinguishing tiny lesions from healthy tissue and providing a high-precision spatial navigation basis for debridement path planning and targeted drug delivery.

[0051] 2. Based on a deep reinforcement learning algorithm, biofilm thickness data and dentin damage feedback are integrated in real time to dynamically adjust the trajectory and operating parameters of the debridement instrument. This invention addresses the rigidity of traditional static planning, significantly reducing the risk of accidental dentin damage while improving debridement efficiency.

[0052] 3. Using infection probability maps to drive drug catheter positioning, combined with a dual triggering mechanism using pH and enzyme activity sensors, this technology enables on-demand release of antibacterial and anti-inflammatory drugs. This breakthrough overcomes the blindness of traditional empirical drug delivery, ensuring precise drug coverage of the core lesion area, inhibiting drug resistance and promoting tissue repair.

[0053] 4. Based on the dynamic analysis of inflammatory factors using long-short-term memory networks, a prediction model for infection recurrence probability is constructed. This invention uses the temporal correlation mining of multi-dimensional biomarkers to achieve early warning of recurrence risk, providing a time window for proactive intervention and avoiding secondary surgical trauma.

[0054] 5. Through real-time data interaction between imaging, debridement, medication, and monitoring modules, a closed-loop "diagnosis-treatment-feedback" system is formed. This invention breaks the fragmented nature of traditional segmented treatment, enabling dynamic, iterative optimization of periapical periodontitis treatment strategies and improving the therapeutic adaptability and clinical operability of complex cases. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 This is a module architecture diagram of the present invention;

[0056] Figure 2 is a flow chart of the method of the present invention;

[0057] Figure 3 It is a schematic diagram of the device structure of the present invention;

[0058] Figure 4 Schematic diagram of the structure of the catheter of the present invention.

[0059] Among them, 2. external pipeline; 3. equipment body; 5. pus suction tube; 6. catheter; 8. first switch; 9. second switch; 10. third switch; 301. cloud system. DETAILED DESCRIPTION

[0060] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the specification of the present invention. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0061] Please see the attached Figure 1 The embodiment of the present invention provides an auxiliary treatment system for irreversible suppurative apical periodontitis, comprising:

[0062] A multimodal image fusion module is used to fuse low-dose cone-beam CT images with optical coherence tomography (OCT) images to generate a three-dimensional fused image of the periapical region. It then segments the infected area using a lightweight convolutional neural network and outputs an infection probability map.

[0063] In this embodiment, the multimodal image fusion module is used to integrate the complementary information of low-dose cone-beam CT images and optical coherence tomography (OCT) images to generate high-precision three-dimensional fused images, and realize automatic segmentation of the infected area through a lightweight convolutional neural network, providing a spatial positioning basis for subsequent debridement path planning and drug control.

[0064] Low-dose cone-beam CT imaging is designed with radiation dose reduction at its core. This is achieved by optimizing tube voltage, tube current, and exposure time, enabling radiation dose control while preserving bone tissue resolution and minimizing patient radiation exposure. Optical coherence tomography (OCT) imaging, based on near-infrared interferometry, provides micron-level high-resolution imaging of root canal surface biofilms and soft tissue, compensating for the lack of contrast in soft tissue structures typically associated with CT imaging. Spatial and resolution differences between the two modalities are addressed through a cross-scale registration algorithm, ensuring effective fusion of multi-source data.

[0065] In order to achieve accurate fusion of CT and OCT images, a feature point-based registration method is adopted. Specifically, bony landmarks in CT images (such as the apical foramen and alveolar ridge top) and biofilm-dentin interface feature points in OCT images are extracted, and feature matching is performed using the scale-invariant feature transform (SIFT) algorithm. The OCT data is mapped to the CT coordinate system using a rigid body transformation model. Preferably, the registration error is further optimized using the iterative closest point (ICP) algorithm to ensure that the three-dimensional spatial alignment accuracy meets clinical needs.

[0066] The fusion process uses an adaptive weighted formula Where α is the fusion weight coefficient, is the registered OCT data; V CBCT is the three-dimensional voxel data of low-dose cone-beam CT; V fused Preferably, the weight coefficient is optimized by cross-validation to adapt to the differences in imaging characteristics of periapical lesions in different patients.

[0067] The lightweight convolutional neural network uses MobileNet-V3 as its basic architecture and performs end-to-end training on 3D fusion images. The calculation formula for its output infection probability map is:

[0068]

[0069] Among them, DWConv krepresents the kth depth-wise separable convolutional layer; σ(·) is the Sigmoid activation function; w k is the weight parameter of the k-th convolution kernel; P inf is the infection probability map of the periapical region; K is the total number of depthwise separable convolutional layers in the lightweight convolutional network.

[0070] The network input is a two-dimensional slice sequence reconstructed from multiple planes. The local texture features are extracted through the depth-separable convolution layer, and the multi-scale context information is retained by combining the jump connection structure. The output layer is activated by the Sigmoid function to generate a pixel-by-pixel infection probability map P inf , whose value range is mapped to [0,1], indicating the confidence that each voxel belongs to the infected area. Preferably, the network training adopts a weighted cross entropy loss function to solve the problem of class imbalance between infected and non-infected areas.

[0071] The infection probability map serves as a core output, providing the intelligent drug control module with a basis for catheter positioning. Specifically, the coordinates of the drug delivery target are determined by centroid calculation of high-confidence infection areas within the probability map. When the biosensor monitoring module detects a recurrence risk exceeding a threshold, it triggers the module to re-execute the image acquisition, registration, and segmentation processes, ensuring the timeliness of the fused image and the completeness of lesion coverage.

[0072] Through the complementary modalities of low-dose CT and OCT, this module reduces radiation hazards while improving the detection sensitivity of tiny infection foci; the closed-loop feedback mechanism effectively responds to changes in lesion morphology during treatment by dynamically updating fused images, laying a data foundation for adaptive control of the entire system.

[0073] The reinforcement learning debridement navigation module uses a deep deterministic policy gradient algorithm to dynamically plan the debridement path based on 3D fusion imaging and biofilm thickness data detected in real time by OCT, and optimizes the path based on the degree of dentin damage caused by instrument operation;

[0074] In this embodiment, the reinforcement learning debridement navigation module is based on a deep reinforcement learning algorithm, combines real-time imaging data with biofilm thickness to dynamically plan the debridement path, and optimizes the risk of damage to dentin caused by instrument operation to achieve precise removal of periapical lesions.

[0075] The state space input includes a three-dimensional fusion image generated by the multimodal image fusion module, real-time biofilm thickness data (measured by optical coherence tomography OCT), and the current position coordinates of the debridement instrument. The three-dimensional fusion image provides anatomical information of the periapical region, the real-time biofilm thickness data characterizes the progress of lesion removal, and the instrument coordinates are fed back in real time through the spatial encoder. The motion space output is the movement direction (including pitch angle and yaw angle) and cutting force parameters of the debridement instrument, which control the movement trajectory and operation intensity of the instrument. Preferably, the motion space adopts continuous value encoding to adapt to the output characteristics of the deep deterministic policy gradient algorithm.

[0076] The reward function of the deep deterministic policy gradient algorithm is designed as:

[0077]

[0078] Where Δd bio Indicates the reduction in biofilm thickness at the current moment, calculated using real-time OCT detection data, reflecting the effectiveness of the debridement operation; is the initial biofilm thickness, which is used to normalize the reward value to ensure comparability between different cases; Δh k is the cutting depth of the debridement instrument on the kth dentin partition, which is calculated by the real-time scanning data of the OCT image. Its value exceeds the safety threshold h th The penalty term is triggered when β and γ are the weight coefficients of the reward and penalty terms, respectively, which are determined through clinical experience and cross-validation optimization to balance the debridement efficiency and safety requirements; K1 is the total number of dentin partitions.

[0079] The debridement path is updated every preset time interval Δt, and the update basis includes the latest biofilm thickness data, instrument position and dentin damage risk assessment results. Preferably, the time interval Δt is dynamically adjusted according to the complexity of the lesion to ensure the real-time and computational efficiency of the path planning. During the path update process, the reinforcement learning agent regenerates the action sequence based on the current state and verifies the feasibility of the path through the simulation environment. If the updated path leads to a significant increase in the risk of dentin damage, such as The path rollback mechanism is triggered and the historical optimal path is re-called for execution.

[0080] The deep deterministic policy gradient algorithm was pre-trained in a simulation environment based on real-world case data, including multimodal imaging, biofilm thickness curves, and a dentin mechanical response model. During training, the policy and value networks employed a double-delayed update (TD3) mechanism to mitigate Q-value overestimation.

[0081] Preferably, the model adapts to the anatomical variations of different patients through transfer learning, specifically by freezing the pre-convolutional layers of the pre-trained network and only fine-tuning the parameters of the fully connected layers.

[0082] The coordinates of the end point of the debridement path serve as the trigger signal for the intelligent drug control module. When the device reaches the end point of the path, the drug delivery process is automatically started.

[0083] Preferably, based on Figure 3 and attached Figure 4 The device body 3 inside is connected to the cloud system 301 by wires and is controlled by the cloud system 301. When it is controlled by the system, it can control the first switch 8, the second switch 9 and the third switch 10 on the surface of the device body 3, wherein the first switch 8 is used to control the pus suction mechanism built into the device body 3. The pus suction structure is a prior art and will not be described in detail herein. When the pus suction structure is working, the pus suction tube 5 in the external pipe 2 fixed on the outside can absorb the pus at the path and the end point. The second switch 9 is used to control the catheter 6 in the external pipe 2 to deliver the drug through the intelligent drug control module. The third switch 10 is used to control the opening and closing of the device body 3.

[0084] The intelligent drug control module locates the drug delivery catheter according to the infection probability map, triggers drug release through periapical microenvironment parameters detected by pH sensors and enzyme activity sensors, and dynamically adjusts drug delivery pressure based on a PID control algorithm;

[0085] The intelligent drug control module dynamically triggers drug release and adjusts delivery pressure based on the probability distribution of the infected area and the periapical microenvironment parameters, achieving precise delivery and dosage control of local drugs in the lesion to inhibit infection recurrence and promote tissue repair.

[0086] The position of the drug delivery catheter is determined by the centroid coordinates of the infection probability map. The specific calculation formula is:

[0087]

[0088] Among them, x, y, z represent the voxel coordinates in the 3D fusion image coordinate system; P inf (x, y, z) represents the probability value of the voxel point with coordinates (x, y, z) belonging to the infected area, and the value range is [0, 1]; Σ represents the traversal summation of all voxel points in the 3D fusion image; x c is the coordinate of the center of mass of the infected area in the x-axis direction, which is used to locate the target position of the drug delivery catheter.

[0089] A microneedle array is configured at the end of the catheter, which achieves millimeter-level precision movement through a piezoelectric drive mechanism, ensuring that the drug release target coincides with the core area of ​​infection.

[0090] The triggering conditions for drug release are based on real-time monitoring data of the local microenvironment around the apex, including pH and enzyme activity indicators:

[0091] pH sensor: monitors the pH of periapical tissue in real time. When the pH value is ≤5.5, it is determined to be an acidic microenvironment caused by infection, triggering the release of antimicrobial drugs.

[0092] Enzyme activity sensor: detects the concentration of matrix metalloproteinase-8 (MMP-8). When MMP-8 ≥ 50 U / L, it is determined to be the active stage of tissue destruction, triggering the release of anti-inflammatory drugs.

[0093] Preferably, the sensor adopts a flexible electrode and microfluidic chip integrated design, and transmits the signal to the control unit via wireless signal to avoid interfering with the operation of the debridement instrument.

[0094] The release trigger conditions are:

[0095] pH≤τ pH AND Enzyme ≥ τ enzyme ;

[0096] pH is the pH value detected in real time in the periapical region, which is measured by a micro pH sensor and reflects the acidity of the infection microenvironment; τ pH It represents the pH threshold for drug release. A value below this value indicates that dentin demineralization has begun. Enzyme is the collagenase activity value in the periapical region, which is measured by an enzyme activity biosensor and reflects the metabolic activity of the biofilm. enzyme is the enzyme activity threshold that triggers drug release. A value above this value indicates that the infection is in the acute progressive stage; AND is the logical AND operator.

[0097] The drug delivery pressure is dynamically adjusted using a proportional-integral-derivative (PID) control algorithm to match the error between the target drug dose and the real-time release amount. The PID control formula is:

[0098]

[0099] Where u(t) is the adjusted injection pressure, which is used to control the drug delivery rate; e(t) = r(t) - y(t), which represents the drug release error, r(t) is the target drug dose; y(t) is the real-time release amount; K p is the proportional coefficient; K i is the integral coefficient; K d is the differential coefficient; τ is the integral variable, representing any moment in the time interval [0, t]; It is the first-order derivative of the error signal e(t) with respect to time, and represents the error change rate.

[0100] The PID controller adopts anti-integral windup design to prevent pressure overshoot caused by long-term error accumulation.

[0101] The biosensor monitoring module detects inflammatory factor concentrations in real time and predicts the risk of infection recurrence through a long-short-term memory network. When the predicted recurrence risk exceeds a preset threshold, the multimodal image fusion module is triggered to rescan the periapical area and update the debridement path of the reinforcement learning debridement navigation module based on the regenerated fused image.

[0102] In this embodiment, the biosensor monitoring module detects the concentration of inflammatory factors and changes in biofilm thickness in real time, combines the time series prediction model to evaluate the risk of infection recurrence, and drives the closed-loop feedback mechanism to update the treatment strategy, thereby realizing dynamic monitoring and adaptive adjustment of the apical periodontitis treatment process.

[0103] The module integrates a flexible biosensor array for continuous monitoring of specific inflammatory markers in the periapical microenvironment. Interleukin-6 (IL-6) concentration is detected based on electrochemical impedance spectroscopy, which generates impedance changes through antigen-antibody specific binding reactions and converts them into concentration readings via a signal amplification circuit. C-reactive protein (CRP) concentration is detected using surface plasmon resonance (SPR) technology, which uses CRP molecules to bind to a sensor surface modified with gold nanoparticles to cause a resonance angle shift. This shift angle is quantified and the concentration value is inverted using an optical system. Preferably, the sensor array is encapsulated in a biocompatible material, and data is uploaded to the processing unit in real time via a wireless transmission module.

[0104] The input of the long short-term memory network (LSTM) is multi-dimensional time series data, including IL-6 concentration C II-6 (t), CRP concentration C CRP (t) and real-time biofilm thickness d bio (t). IL-6 and CRP concentration data are sampled at a frequency of minutes, and biofilm thickness data are updated at preset intervals using optical coherence tomography (OCT). Preferably, the input data is normalized to eliminate dimensional differences, and a sliding time window (e.g., a 6-hour window) is used to construct a time series to capture the dynamic changes in inflammatory factors.

[0105] The output of the LSTM network is the infection recurrence probability P rec , the calculation formula is:

[0106] P rec =σ(W p ·h t +b p );

[0107] Among them, h t is the LSTM hidden state; σ(·) is the Sigmoid function; W p is the weight matrix of the fully connected layer; b p is the bias vector of the fully connected layer.

[0108] Preferably, the LSTM network contains a two-layer hidden structure, the number of hidden units is adaptively adjusted according to the input dimension, and Dropout regularization is used during training to alleviate overfitting.

[0109] When the predicted recurrence probability P rec When the preset threshold (e.g. ≥0.8) is exceeded, the following closed-loop control process is triggered:

[0110] Image rescan command: Sends a signal to the multimodal image fusion module to start a combined low-dose cone-beam CT and OCT scan and update the three-dimensional fusion image of the periapical area;

[0111] Path update instructions: Input the updated imaging data into the reinforcement learning debridement navigation module to replan the debridement path, focusing on removing new or residual infection lesions;

[0112] Drug parameter reset: Notify the intelligent drug control module to adjust the PID coefficient to adapt to the drug delivery needs of the newly infected area.

[0113] LSTM unit state update formula:

[0114]

[0115] Among them, W f , W i , W C , W o are the weight matrices of the forget gate, input gate, cell state, and output gate respectively; b f , b i , b C , b o are the bias items of the corresponding gates, initialized to zero vector; h t-1 is the hidden state of the previous moment, encoding the historical trend of inflammatory factors; x t is the input feature vector at the current moment; σ is the Sigmoid activation function; tanh is the hyperbolic tangent activation function; ⊙ is the element-by-element multiplication; f t Forget gate output, which determines how much historical cell state C is retained t-1 ;i t Output of the input gate, controlling the new candidate state Update ratio; C is the candidate cell state, storing the new information at the current moment; t The updated cell state combines historical and current information; t is the output gate, which determines the current hidden state h t Output ratio; h t Hide status for the current moment.

[0116] The prediction loss function uses weighted cross entropy:

[0117]

[0118] Among them, y i is the true label of the i-th sample; For the The predicted recurrence probability of each sample is output by the LSTM model; For the The class weight coefficient of each sample is used to alleviate the class imbalance problem; Σ represents the sum of all training samples.

[0119] The robustness of recurrence risk assessment can be improved through the multimodal data fusion of inflammatory factors and biofilm thickness; the time series modeling capability of the LSTM network effectively captures the dynamic laws of inflammatory progression and compensates for the lag of the traditional threshold alarm mechanism; the closed-loop feedback mechanism opens up the "monitoring-diagnosis-treatment" data flow, ensuring the system response speed and the timeliness of the treatment strategy.

[0120] When the reinforcement learning debridement navigation module completes the debridement path, the coordinates of the debridement path's endpoint serve as a trigger signal, initiating the drug release process in the intelligent drug control module. Simultaneously, drug delivery pressure data is fed back to the biosensor monitoring module in real time. If uneven drug distribution is detected (e.g., pressure fluctuations exceeding a threshold), the multimodal image fusion module is triggered to rescan, update the infection probability map, and recalculate the center of mass coordinates, forming a closed-loop control system of "debridement-drug administration-monitoring."

[0121] The auxiliary treatment method for irreversible suppurative apical periodontitis described below and the auxiliary treatment system for irreversible suppurative apical periodontitis described above can be referred to in correspondence with each other.

[0122] Please see the attached Figure 2 The present invention also provides an auxiliary treatment method for irreversible suppurative periapical periodontitis, comprising the following steps:

[0123] Low-dose cone-beam CT and optical coherence tomography were used to obtain periapical imaging data, generate three-dimensional fused images, and segment the infected area.

[0124] Based on fused images and real-time biofilm thickness data, a deep reinforcement learning algorithm is used to dynamically plan the debridement path and optimize the damage to dentin caused by instrument manipulation.

[0125] The drug catheter is positioned according to the infection probability map, the microenvironmental parameters are detected by sensors to trigger drug release, and the delivery pressure is dynamically adjusted;

[0126] Real-time monitoring of inflammatory factor concentrations can predict recurrence risk. When the risk exceeds the threshold, the image can be rescanned and the debridement pathway can be updated.

[0127] The method of this embodiment can be used to execute the above system embodiment. Its principles and technical effects are similar and will not be described in detail here.

[0128] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. An auxiliary treatment system for irreversible suppurative apical periodontitis, characterized in that: include: A multimodal image fusion module is used to fuse low-dose cone-beam CT images with optical coherence tomography (OCT) images to generate a three-dimensional fused image of the periapical region. It then segments the infected area using a lightweight convolutional neural network and outputs an infection probability map. A reinforcement learning debridement navigation module uses a deep deterministic policy gradient algorithm to dynamically plan the debridement path based on the three-dimensional fusion image and biofilm thickness data detected by real-time OCT, and optimizes the path based on the degree of damage to the dentin caused by instrument operation; an intelligent drug control module, which locates the position of the drug delivery catheter according to the infection probability map, triggers drug release through periapical microenvironment parameters detected by a pH sensor and an enzyme activity sensor, and dynamically adjusts the drug delivery pressure based on a PID control algorithm; The biosensor monitoring module detects the concentration of inflammatory factors in real time and predicts the risk of infection recurrence through a long-short-term memory network. When the predicted recurrence risk exceeds a preset threshold, the multimodal image fusion module is triggered to rescan the periapical area and update the debridement path of the reinforcement learning debridement navigation module based on the regenerated fusion image.

2. The auxiliary treatment system for irreversible suppurative periapical periodontitis according to claim 1, characterized in that: In the multimodal image fusion module: The radiation dose of the low-dose cone-beam CT image is 50% of the conventional dose; The OCT image has a scanning accuracy of micrometer level and is used to detect the thickness of biofilm on the root canal surface; The three-dimensional fusion image is obtained by weighting formula Generate, where α is the fusion weight coefficient, is the registered OCT data; V CBCT is the three-dimensional voxel data of low-dose cone-beam CT; V fused is the fused three-dimensional image data.

3. The auxiliary treatment system for irreversible suppurative periapical periodontitis according to claim 2, characterized in that: The lightweight convolutional neural network is a MobileNet-V3 model, and the calculation formula for its output infection probability map is: Among them, DWConv k represents the kth depth-wise separable convolutional layer; σ(·) is the Sigmoid activation function; w k is the weight parameter of the k-th convolution kernel; P inf is the infection probability map of the periapical region; K is the total number of depthwise separable convolutional layers in the lightweight convolutional network.

4. The auxiliary treatment system for irreversible suppurative periapical periodontitis according to claim 1, characterized in that: In the reinforcement learning debridement navigation module: The reward function of the deep deterministic policy gradient algorithm is designed as: Where Δd bio is the reduction in biofilm thickness; Δh k is the depth of dentin damage; β and γ are weight coefficients; is the initial biofilm thickness; h th is the safe damage threshold of dentin; K1 is the total number of dentin partitions; The debridement path is updated at intervals of Δt based on the latest biofilm thickness data, wherein Δt is a preset fixed time interval.

5. The auxiliary treatment system for irreversible suppurative periapical periodontitis according to claim 4, characterized in that: The dentin damage depth Δh k The calculation is performed using real-time scanning data of OCT images, and the weight coefficients β=0.6 and γ=0.

4.

6. The auxiliary treatment system for irreversible suppurative periapical periodontitis according to claim 1, characterized in that: In the intelligent drug control module: The trigger threshold of the pH sensor is pH ≤ 5.5, and the trigger threshold of the enzyme activity sensor is MMP-8 ≥ 50 U / L; The pressure regulation formula of the PID control algorithm is: Where u(t) is the adjusted injection pressure, which is used to control the drug delivery rate; e(t) = r(t) - y(t), which represents the drug release error, r(t) is the target drug dose; y(t) is the real-time release amount; K p is the proportional coefficient; K i is the integral coefficient; K d is the differential coefficient; τ is the integral variable, representing any moment in the time interval [0, t]; It is the first-order derivative of the error signal e(t) with respect to time, and represents the error change rate.

7. The auxiliary treatment system for irreversible suppurative periapical periodontitis according to claim 6, characterized in that: The position of the drug delivery catheter is determined by the centroid coordinates of the infection probability map: Among them, x, y, z represent the voxel coordinates in the 3D fusion image coordinate system; P inf (x, y, z) represents the probability value of the voxel point with coordinates (x, y, z) belonging to the infected area, and the value range is [0, 1]; ∑ represents the traversal summation of all voxel points in the 3D fusion image; x c is the coordinate of the center of mass of the infected area in the x-axis direction, which is used to locate the target position of the drug delivery catheter.

8. The auxiliary treatment system for irreversible suppurative periapical periodontitis according to claim 1, characterized in that: In the biosensor monitoring module: The inflammatory factors include IL-6 and CRP, and their concentrations are detected in real time by a flexible biosensor; The input of the long short-term memory network is time series data {C II-6 (t),C CRP (t),d bio (t)}, where C II-6 (t) is the interleukin-6 concentration detected at time t, in pg / mL; C CRP (t) is the C-reactive protein concentration detected at time t, in mg / L; d bio (t) is the biofilm thickness on the root canal surface measured in real time by optical coherence tomography (OCT) at time t, in μm, and the recurrence probability P is output. rec The calculation formula is: P rec =σ(W p ·h t +b p ); Among them, h t is the LSTM hidden state; σ(·) is the Sigmoid function; W p is the weight matrix of the fully connected layer; b p is the bias vector of the fully connected layer.

9. The auxiliary treatment system for irreversible suppurative periapical periodontitis according to claim 8, characterized in that: The preset threshold is P rec ≥0.8, and the interval between rescanned fusion images to update the reinforcement learning debridement pathway was consistent with the initial scan interval.

10. An auxiliary treatment method for irreversible suppurative apical periodontitis, characterized in that: The auxiliary treatment system for irreversible suppurative periapical periodontitis according to any one of claims 1 to 9 comprises the following steps: Low-dose cone-beam CT and optical coherence tomography were used to obtain periapical imaging data, generate three-dimensional fused images, and segment the infected area. Based on fused images and real-time biofilm thickness data, a deep reinforcement learning algorithm is used to dynamically plan the debridement path and optimize the damage to dentin caused by instrument manipulation. The drug catheter is positioned according to the infection probability map, the microenvironmental parameters are detected by sensors to trigger drug release, and the delivery pressure is dynamically adjusted; Real-time monitoring of inflammatory factor concentrations can predict recurrence risk. When the risk exceeds the threshold, the image can be rescanned and the debridement pathway can be updated.

Images