Intelligent interpretable lung CT (computed tomography) image diagnosis method and intelligent interpretable lung CT image diagnosis equipment

Abstract

The invention relates to the technical field of medical image intelligent diagnosis, and provides an interpretable lung CT image intelligent diagnosis method and device, and the device comprises an image calling module, an image processing module, an image analysis module and a result display module. According to the method, identity information of a patient is read to automatically verify and authorize, and an electronic CT image is called and stored in a data storage card through an encryption transmission channel. And the image processing module performs denoising, enhancement and interpretability analysis on the original image to generate a thermodynamic diagram to assist in decision making. The image analysis module completes lesion detection and classification, verifies the consistency between a model decision and doctor experience through a thermodynamic diagram, and optimizes network parameters in combination with feedback data. And the result display module displays a diagnosis result and a thermodynamic diagram in real time, and supports multi-terminal cooperation and data archiving. The device supports data sharing and remote consultation, the efficiency and convenience of diagnosis are improved, and the wide application prospect of artificial intelligence in the field of medical diagnosis is shown.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image intelligent diagnosis, and in particular to an interpretable lung CT image intelligent diagnosis method and device. Background Art

[0002] Lung cancer is one of the most lethal malignant tumors worldwide. Its early and accurate identification directly impacts patients' five-year survival rates. However, traditional histopathological analysis methods are time-consuming and rely on specialized personnel, limiting the feasibility of large-scale screening and hindering timely detection and treatment. Intelligent recognition and detection of lung CT images are crucial for the early diagnosis of lung cancer. Currently, medical cloud platforms are using lung CT images extracted from patient identity information for processing to identify abnormal areas and perform automated analysis. Deep learning algorithms, particularly convolutional neural networks, enable models to learn image features and improve detection accuracy. The application of these technologies not only improves diagnostic efficiency but also helps reduce human error, providing patients with more accurate medical diagnoses. With the further development of multi-omics data fusion technology and the integration of artificial intelligence and imaging genomics, the diagnosis and treatment of lung cancer will become more precise and personalized.

[0003] While existing deep learning-based computer vision technology can localize and classify lung nodules, it still faces significant challenges in clinical application. First, the model's decision-making process lacks transparency, making it difficult for doctors to trust AI predictions. Second, data labeling relies heavily on the subjective experience of radiologists, and differences in labeling between different medical institutions can lead to a decrease in model generalization. Furthermore, issues such as missed detection of small lesions and insufficient fusion of multi-resolution image features further hinder improvements in diagnostic accuracy. Summary of the Invention

[0004] The technical problem addressed by this invention is lung CT imaging diagnosis. The purpose of this invention is to provide an interpretable intelligent lung CT imaging diagnostic method and device, in order to assist physicians in improving diagnostic efficiency and accuracy. By retrieving and encrypting electronic lung CT images using patient identity information, the original images are preprocessed and feature extracted, enabling classified storage, rapid identification and detection, and visualization of the model's decision-making basis for lung CT images, while minimizing the impact on the testing environment and surrounding areas.

[0005] The present invention is an interpretable lung CT image intelligent diagnosis method and device, including the following modules:

[0006] S1, the image retrieval module, automatically verifies authorization by reading patient identity information or entering unique identification information, quickly retrieves electronic lung CT images, and stores them in the data storage card via an encrypted transmission channel.

[0007] S2, the image processing module, preprocesses, denoises, and enhances the original image. It uses a Gaussian filter algorithm to eliminate image noise and a dynamic histogram equalization technique to enhance the contrast of the lesion area. It then performs interpretability analysis based on the Gradient-weighted algorithm (Grad-CAM) and generates a heatmap visualization to provide decision-making support.

[0008] S3, the image analysis module, detects and classifies lesions based on processed image data, verifying the consistency of model decisions with physician experience through heatmaps. It then dynamically optimizes network parameters based on annotated feedback data, calculating sensitivity and specificity to evaluate the performance of the optimized model.

[0009] S4, the result display module, displays the diagnostic results and heat maps synchronously on the device display system, presenting analysis data in real time, supporting multi-terminal collaboration, and completing data archiving and system maintenance.

[0010] The specific steps of module S1 are as follows:

[0011] S11, identity information input and electronic image authorization.

[0012] The user starts the device, automatically verifies authorization by reading the patient's ID card or entering unique identification information, quickly retrieves electronic lung CT images, parses metadata to achieve full-second data access, and provides standardized, high-fidelity image input.

[0013] S12, Data Encryption Transmission and Dynamic Storage Management. Image data is stored via a transmission channel to a data storage card, providing standardized input for subsequent processing. The data storage card utilizes an encrypted storage architecture to prevent the leakage of patient privacy data.

[0014] S13, user-defined function interactive control. The user can use the device switch (102) to start or stop specific functions of the device at any time as needed to meet different usage scenarios and needs.

[0015] The specific steps of module S2 are as follows:

[0016] S21, Image preprocessing and adaptive enhancement.

[0017] S211, Image input and standardization: After the electronic lung CT image is encrypted and stored, the image is grayscale normalized through the computing unit, and the pixel value is adjusted to the preset range to eliminate the brightness deviation caused by equipment differences.

[0018] S212, Image Denoising and Enhancement: Gaussian filtering algorithm is used to eliminate image noise, and dynamic histogram equalization technology is used to enhance the contrast of the lesion area, providing standardized input for subsequent feature extraction and interpretable heat map generation.

[0019] S22, Deep Feature Extraction and Probabilistic Classification. This system extracts deep features from lung CT images using a convolutional neural network. The softmax function outputs classification probabilities, and a gradient-weighted algorithm is used to generate heat maps. This allows for visualization of lung disease classification and decision-making.

[0020] Specifically, during the forward propagation process, a multi-layer feature map is extracted through a pre-trained convolutional neural network (CNN), which contains multiple convolutional layers and pooling layers, and the last fully connected layer outputs the category score. .

[0021] Furthermore, the Softmax function is used to calculate the probability and score each category. Converted into probability distribution, the specific formula is:

[0022]

[0023] Where C is the total number of pathological classifications of lung lesions (such as malignant, benign, and non-neoplastic lesions), is the target category (e.g. stands for "benign lesions"), is the input image Belong to category The probability of It is The scores of target categories.

[0024] S23, hierarchical interpretability heatmap generation. The Gradient-weighted Class Activation Mapping (Grad-CAM) algorithm is executed as follows:

[0025] S231, Gradient Backpropagation: For target categories , calculate the category score for the last convolutional layer feature map The gradient value of , the specific formula is:

[0026]

[0027] in It is feature maps for target categories The weight of is the total number of pixels in the feature map, Indicates the Feature maps at position The activation value of is the target category Score For the first Feature maps gradient.

[0028] These modules have deep learning capabilities and can gradually improve the accuracy and efficiency of detection through continuous training and optimization.

[0029] S232, heat map generation: linearly superimpose the gradient-weighted feature map and retain the positive contribution area through the ReLU activation function. The specific formula is:

[0030]

[0031] in is a category Heat map, ReLU is the activation function used to retain the area that has a positive contribution to the prediction result. It is A feature map.

[0032] The generated heat map is upsampled to the original image resolution through bilinear interpolation and superimposed on the CT image. The highlighted areas (such as the outline of lung nodules) indicate the basis for the model's decision.

[0033] The specific steps of module S3 are as follows:

[0034] S31, determine the probability threshold of the lesion. Set the classification threshold , if the highest probability of Softmax output , it is determined to be the corresponding category and marked appropriately.

[0035] S32, Thermal Map Verification and Model Credibility Assessment. The thermal map and the original lung CT image are displayed synchronously on the display system. Doctors confirm the credibility of the model's decision by comparing the consistency of the thermal map's marked area with clinical experience. If the thermal map deviates from the actual lesion area, the model optimization process is triggered.

[0036] S33, model self-learning and intelligent optimization. Specific steps include:

[0037] S331, data iteration update: import the corrected data annotated by the doctor into the data storage card to expand the training data set.

[0038] S332, Parameter Fine-tuning: Dynamically adjust model parameters through the backpropagation algorithm to optimize lesion classification accuracy and heatmap credibility. Comparing model heatmaps with real areas of interest annotated by physicians, we balance the classification task with interpretability, continuously improving lung lesion identification accuracy.

[0039] S333, Performance Evaluation: Validate the optimized model's performance on an independent test set by calculating sensitivity and specificity, ensuring a minimum 10% improvement in the detection rate of small lesions. Sensitivity focuses on reducing the risk of missed diagnoses, while specificity reduces misdiagnoses. Evaluation results must meet the criteria on an independent test set, balancing these two criteria to achieve the optimal balance between clinical reliability and patient safety.

[0040] The specific steps of module S4 are as follows:

[0041] S41, Visualization of Diagnostic Imaging Results. The diagnostic conclusion and thermal map are displayed synchronously on the device display system, and an external monitor can be connected via a high-definition signal interface to facilitate team collaboration and expert consultation.

[0042] S42, medical data export and compliance sharing. Use a data cable interface to export analysis results to external devices. A memory card also supports rapid retrieval of historical cases, improving the efficiency of longitudinal comparisons.

[0043] S43, Intelligent Imaging Diagnosis System Maintenance. The device's charging port supports multiple fast-charging protocols to ensure battery life and compatibility. The main switch has an integrated safety lock and remote control to prevent data loss caused by misoperation and supports remote activation and deactivation. The device also features intelligent energy-saving adaptive switching, automatically switching to low-power standby mode or safe shutdown after completing the diagnostic process, optimizing energy consumption.

[0044] The beneficial effects of the present invention are:

[0045] 1. This invention effectively improves the automation and security of lung CT image data acquisition by constructing an intelligent retrieval module based on patient identity verification, providing efficient and secure technical support for smart medical image data analysis.

[0046] 2. An image processing method that combines deep learning algorithms with Gaussian filtering, dynamic histogram enhancement, gradient-weighted class activation mapping (Grad-CAM) and other technologies has improved the feature extraction accuracy of lung CT images and the visualization effect of lesion areas.

[0047] 3. This invention achieves transparency and enhanced credibility of the decision-making process of artificial intelligence models by constructing an interpretable heat map verification closed-loop system, effectively alleviating the problem of clinicians' distrust of AI diagnostic results.

[0048] 4. Combining the convolutional neural network (CNN) with the classification optimization algorithm, and adopting a dynamic parameter fine-tuning mechanism and feedback learning method based on doctor-labeled data, we can achieve continuous optimization of model performance and in-depth mining of medical data. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1It is a schematic diagram of the overall structure of the device of the first embodiment of the present invention when the screen is open.

[0050] Figure 2 It is a schematic diagram of the back structure of the device of the first embodiment of the present invention when the screen is opened.

[0051] Figure 3 It is a schematic diagram of the left side structure of the device of the first embodiment of the present invention when the screen is opened.

[0052] Figure 4 It is a schematic diagram of the overall structure of the device of the first embodiment of the present invention when the screen is in a closed state.

[0053] Figure 5 It is a schematic diagram of the back structure of the device of the first embodiment of the present invention when the screen is in a closed state.

[0054] Figure 6 It is a front cross-sectional view of the device according to the first embodiment of the present invention when the screen is closed.

[0055] Figure 7 This is a rear cross-sectional view of the device according to the first embodiment of the present invention when the screen is closed.

[0056] Figure 8 This is a bottom cross-sectional view of the device according to the first embodiment of the present invention when the screen is closed.

[0057] Figure 9 This is a flow chart of a method for implementing interpretable intelligent diagnosis of lung CT images using the device described in the present invention.

[0058] In the figure: 10-control system: 101-main switch, 102-device switch, 103-power indicator light; 20-transmission system: 201-charging port, 202-card reading area, 203-data line interface, 204-HD signal interface; 205-data transmission light, 206-data storage card; 30-display system. DETAILED DESCRIPTION

[0059] In order to make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the technical solutions of the present invention will be further described in detail below with reference to the accompanying drawings.

[0060] An interpretable lung CT image intelligent diagnosis method and device, such as Figure 7 As shown, its specific operating mechanism is as follows:

[0061] S1, image retrieval module.

[0062] By reading the patient's identity information or entering unique identification information to automatically verify authorization, electronic lung CT images can be quickly retrieved and stored in the data storage card through an encrypted transmission channel. The specific steps include:

[0063] S11, identity information input and electronic image authorization.

[0064] The user activates the main switch (101), and the device is powered on and in standby mode. The power indicator light (103) lights up. The user places the patient's second-generation ID card in the card reader area (202). The RFID module reads the encrypted chip information and electronic medical records, or enters the patient's unique identification information (such as ID card number, medical insurance number), and automatically connects to the medical cloud platform to complete identity verification and data authorization, and retrieve the patient's electronic lung CT image. The system analyzes image metadata (such as scanning parameters and pixel pitch) to provide standardized, high-fidelity image input, ensuring compliance with privacy regulations and compressing the traditional manual review process from minutes to seconds.

[0065] S12, data encryption transmission and dynamic storage management. Image data is stored in the data storage card (206) via the transmission channel, providing standardized input for subsequent processing. During the scanning process, the data transmission light (205) will indicate the transmission status in real time. The data storage card (206) adopts AES-256 encryption storage architecture to prevent the leakage of patient privacy data.

[0066] S13, user-defined function interactive control. When using the specific functions of the device (such as video recording, data transmission), the user can use the device switch (102) to start or stop the device function at any time as needed to meet different usage scenarios.

[0067] S2, image processing module.

[0068] The original image is preprocessed, denoised, and enhanced. The Gaussian filter algorithm is used to eliminate image noise, and the dynamic histogram equalization technique is used to enhance the contrast of the lesion area. Subsequently, interpretability analysis is performed based on the gradient weighted algorithm (Grad-CAM), and a heat map visualization decision-making basis is generated. The specific steps include:

[0069] S21, Image preprocessing and adaptive enhancement.

[0070] S211, image input and standardization: The electronic lung CT image is transmitted to the data storage card (206) via the data transmission channel, and the image is grayscale normalized by the calculation unit to adjust the pixel value to a preset range (such as 0-255) to eliminate the brightness deviation caused by equipment differences.

[0071] S212, Image Denoising and Enhancement: Gaussian filtering is used to eliminate image noise (such as motion artifacts and quantum noise), and dynamic histogram equalization technology is used to enhance the contrast of diseased areas (such as lung nodules and tumors), providing standardized input with high signal-to-noise ratio and high contrast for subsequent feature extraction and interpretable heat map generation.

[0072] S22, Deep Feature Extraction and Probabilistic Classification. This system extracts deep features from lung CT images using a convolutional neural network. The softmax function outputs classification probabilities, and a gradient-weighted algorithm is used to generate heat maps. This allows for visualization of lung disease type and decision-making, ensuring a transparent and traceable diagnostic process.

[0073] Specifically, during the forward propagation process, a multi-layer feature map is extracted through a pre-trained convolutional neural network (CNN), which contains multiple convolutional layers and pooling layers, and the last fully connected layer outputs the category score. .

[0074] Furthermore, the Softmax function is used to calculate the probability and the category score output by the fully connected layer is It is converted into a probability distribution with a sum of 1, thereby determining the lung disease category to which the input lung CT image most likely belongs. The specific formula is:

[0075]

[0076] Where C is the total number of pathological classifications of lung lesions (such as malignant, benign, and non-neoplastic lesions), is the target category (e.g. stands for "benign lesions"), is the input image Belong to category The probability of It is The scores of target categories.

[0077] S23, hierarchical interpretability heatmap generation. The Gradient-weighted Class Activation Mapping (Grad-CAM) algorithm is executed as follows:

[0078] S231, Gradient Backpropagation: For target categories , calculate the category score for the last convolutional layer feature map The gradient value of , the specific formula is:

[0079]

[0080] in It is feature maps for target categories The weight of is the total number of pixels in the feature map, Indicates the Feature maps at position The activation value of is the target category Score For the first Feature maps gradient.

[0081] These modules have deep learning capabilities and can gradually improve the accuracy and efficiency of detection through continuous training and optimization.

[0082] S232, heat map generation: linearly superimpose the gradient-weighted feature map and retain the positive contribution area through the ReLU activation function. The specific formula is:

[0083]

[0084] in is a category Heat map, ReLU is the activation function used to retain the area that has a positive contribution to the prediction result. It is A feature map.

[0085] The generated heat map is upsampled to the original image resolution through bilinear interpolation and superimposed on the CT image. The highlighted areas (such as the outline of lung nodules) indicate the basis for the model's decision.

[0086] S3, image analysis module.

[0087] Lesion detection and classification are completed based on processed image data, and the consistency of model decisions and doctor experience is verified through heat maps. Then, network parameters are dynamically optimized based on labeled feedback data, and sensitivity and specificity are calculated to evaluate the performance of the optimized model. The specific steps include:

[0088] S31, determine the probability threshold of the lesion. Set the classification threshold (like ), if the highest probability of Softmax output , it is judged as the corresponding category (such as "malignant nodule"), otherwise it is marked as "needs manual review".

[0089] S32, thermal map verification and model credibility assessment. The thermal map and the original lung CT image are displayed synchronously on the display system (30). The doctor confirms the credibility of the model decision by comparing the consistency of the thermal map marked area with clinical experience (such as nodule morphology and burr sign). If the thermal map deviates from the actual lesion area (such as focusing on blood vessels instead of nodules), the model optimization process is triggered.

[0090] S33, model self-learning and intelligent optimization. Specific steps include:

[0091] S331, data iterative update: import the corrected data (such as heat map error cases) annotated by the doctor into the data storage card (206) through the data line interface (203) to expand the training data set.

[0092] S332, Parameter Fine-tuning: Dynamically adjust model parameters through the backpropagation algorithm, optimizing both lesion classification accuracy and heatmap credibility. The system compares the model-generated heatmaps with real-world areas of interest annotated by physicians. By balancing the weights of the classification task with interpretability, the system continuously improves the model's recognition accuracy for lung lesions.

[0093] S333, Performance Evaluation: Validate the optimized model's performance on an independent test set by calculating sensitivity and specificity, ensuring a minimum 10% improvement in the detection rate of small lesions. Sensitivity focuses on reducing the risk of missed diagnoses and avoiding delayed treatment; specificity reduces misdiagnoses and prevents over-treatment (e.g., unnecessary biopsies). Evaluation results must meet the criteria on the independent test set, balancing these two criteria to achieve an optimal balance between clinical reliability and patient safety.

[0094] S4, result display module.

[0095] The diagnostic results and heat maps are displayed synchronously on the device display system, showing the analysis data in real time, supporting multi-terminal collaboration, and completing data archiving and system maintenance. The specific steps include:

[0096] S41, visualization of imaging diagnostic results. The diagnostic conclusion (e.g., "high-risk lung nodules, biopsy recommended") is displayed synchronously with the thermal map on the device display system (30), and connected to an external monitor via a high-definition signal interface (204), facilitating team collaboration and expert consultation.

[0097] S42, export and share medical data in a compliant manner. Use the data line interface (203) to export analysis results to an external device (such as a hospital PACS system). At the same time, the memory card (206) supports rapid retrieval of historical cases, improving the efficiency of longitudinal comparison.

[0098] S43, maintenance of the imaging intelligent diagnosis system. The device charging port (201) supports multiple fast charging protocols to ensure device endurance and compatibility; the main switch (101) integrates a safety lock and remote control to prevent data loss caused by misoperation and supports remote opening and closing; the device has an intelligent energy-saving adaptive switching function. After the device completes the diagnostic process, it automatically switches to a low-power standby mode or a safe shutdown state to achieve energy consumption optimization and extend the device life. At the same time, it supports millisecond-level wake-up response to ensure continuous operation requirements.

Claims

1. An interpretable lung CT image intelligent diagnosis method, characterized by: The following steps are involved: Step S1, image retrieval: by reading the patient's identity information or inputting unique identification information to automatically verify the authorization, quickly retrieve the electronic lung CT image, and store it in the data storage card through the encrypted transmission channel; Step S2, image processing: The original image is preprocessed, denoised, and enhanced. A Gaussian filter algorithm is used to remove image noise, and dynamic histogram equalization is used to enhance the contrast of the lesion area. Subsequently, an interpretability analysis is performed based on the gradient-weighted class activation mapping (Grad-CAM) algorithm, and a heat map is generated to visualize the basis for the model's decision making. Step S3, image analysis: lesion detection and classification are completed based on the processed image data. The consistency of the model decision and the doctor's experience is verified through heat maps. The network parameters are then dynamically optimized based on the annotated feedback data. Sensitivity and specificity are calculated to evaluate the performance of the optimized model. Step S4, result display: The diagnostic results and the heat map are synchronously displayed on the display system, the analysis data is displayed in real time, and multi-terminal collaboration is supported to complete data archiving and system maintenance.

2. The method according to claim 1, wherein The image acquisition step S1 specifically includes: Step S11, identity information input and electronic image authorization: read the patient's ID card or enter unique identification information to automatically verify authorization and retrieve the electronic version of the lung CT image; Step S12, data encryption transmission and dynamic storage management: the image data is stored in a data storage card via an encrypted transmission channel, and the data storage card adopts an encrypted storage architecture; Step S13, user-defined function interactive control: the user uses the device switch to start or stop the device function as needed.

3. The method according to claim 1, wherein The image processing step S2 specifically includes: Step S21, image preprocessing and adaptive enhancement: grayscale normalization is performed on the image, denoising is performed using a Gaussian filter algorithm, and contrast is enhanced using a dynamic histogram equalization technique; Step S22, deep feature extraction and probabilistic classification: extract deep features of lung CT images based on convolutional neural networks, and output classification probabilities through the Softmax function; Step S23, hierarchical interpretable heat map generation: a heat map is generated based on the Grad-CAM algorithm, upsampled to the original image resolution through bilinear interpolation, and superimposed on the CT image.

4. The method according to claim 1, wherein The image analysis step S3 specifically includes: Step S31, lesion probability threshold determination: set the classification threshold, and determine the lesion category based on the highest probability output by the Softmax function; Step S32, thermal map verification and model credibility assessment: The thermal map is displayed synchronously with the original image, and the doctor compares and confirms the credibility of the model decision. If the thermal map deviates from the actual lesion area, the model optimization is triggered; Step S33, model self-learning and intelligent optimization: import the corrected data annotated by doctors to expand the training set, dynamically adjust the model parameters through the back propagation algorithm, and calculate the sensitivity and specificity to evaluate the model performance.

5. The method according to claim 1, wherein The result display step S4 specifically includes: Step S41, visualization of imaging diagnostic results: synchronously displaying the diagnostic conclusion and the thermal map on the device display system, and connecting to an external display via a high-definition signal interface; Step S42, exporting and sharing medical data in compliance: using a data line interface to export analysis results to an external device, while also supporting rapid retrieval of historical cases; Step S43, maintenance of the intelligent imaging diagnosis system: the device is charged through the charging port, and the device is turned on and off through the main switch integrated with the safety lock and remote control. The device has an intelligent energy-saving adaptive switching function.

6. An interpretable lung CT image intelligent diagnostic device, used to implement the method according to any one of claims 1 to 5, characterized in that: Includes the following modules: The image retrieval module is used to automatically verify authorization by reading patient identity information or entering unique identification information, quickly retrieve electronic lung CT images, and store them in the data storage card through an encrypted transmission channel; The image processing module is used to preprocess, denoise, and enhance the original image. It uses a Gaussian filter algorithm to eliminate image noise and a dynamic histogram equalization technique to enhance the contrast of the lesion area. It then performs interpretability analysis based on the gradient-weighted class activation mapping (Grad-CAM) algorithm and generates a heat map. The image analysis module is used to detect and classify lesions based on processed image data. It verifies the consistency of model decisions with physician experience through heat maps. It then dynamically optimizes network parameters based on annotated feedback data and calculates sensitivity and specificity to evaluate the performance of the optimized model. The result display module is used to synchronously display the diagnostic results and thermal maps on the device display system, present the analysis data in real time, support multi-terminal collaboration, and complete data archiving and system maintenance.

7. The device according to claim 6, characterized in that The image retrieval module comprises a card reading area (202) and a data storage card (206), wherein the card reading area (202) is used to read the patient's ID card information, and the data storage card (206) adopts an AES-256 encryption storage architecture.

8. The device according to claim 6, characterized in that The image processing module and the image analysis module are implemented by a computing unit, and the computing unit has a built-in pre-trained convolutional neural network (CNN).

9. The device according to claim 6, characterized in that The result display module includes a device display system (30), a high-definition signal interface (204), a data line interface (203), a charging port (201) and a main switch (101); the device display system (30) is used to synchronously display the diagnosis conclusion and the thermal map; the high-definition signal interface (204) supports connection to an external display; the data line interface (203) supports exporting the analysis results to an external device; the charging port (201) supplies power to the device; and the main switch (101) integrates a safety lock and a remote control function for controlling the opening and closing of the device.

Images