Medical space layout design optimization system based on artificial intelligence

Abstract

The invention belongs to the technical field of layout design optimization, and particularly relates to an artificial intelligence-based medical spatial layout design optimization system, which comprises a data acquisition module, a data preprocessing module, a spatial layout module and a layout optimization module, the initial spatial layout is converted into the current spatial layout through the spatial layout module, the spatial layout is continuously optimized, all areas and facilities in the medical space can be reasonably arranged, the action routes of patients and medical staff can be reasonably planned, and the overall use efficiency of the medical space is improved; a layout optimization model is constructed through a layout optimization module, the layout optimization model is used for analyzing context information, a back propagation method is adopted for introducing a momentum item to train the layout optimization model, a layout optimization result is generated, spatial features of the current spatial layout are combined, the context information of the spatial layout is continuously updated, and the requirement change of a hospital can be rapidly adapted; and powerful support is provided for medical services.

Description

Technical Field

[0001] The present invention belongs to the technical field of layout design optimization, and specifically refers to a medical space layout design optimization system based on artificial intelligence. Background Art

[0002] The design of medical space layout refers to the process of planning and arranging various functional areas, facilities and equipment within a medical site. A scientific and reasonable medical space layout can optimize the patient treatment process and also help medical staff carry out their work efficiently;

[0003] However, there are many deficiencies in the current medical space layout design:

[0004] In the current medical space layout design, the functional connections and actual needs of each department are not fully considered. The department distribution is chaotic, and the inspection areas are far apart, resulting in frequent back-and-forth trips between different departments for patients and medical staff, increasing unnecessary walking distances and reducing the treatment efficiency. The limited medical space is not fully utilized and cannot carry more medical service activities;

[0005] In the existing medical space layout design, it is difficult to reasonably optimize according to the patient flow and treatment processes of different departments. The layout of each department may not fully consider the coherence of patient treatment, and the distribution of each functional area is unreasonable. During the peak patient period, it is easy to cause congestion of the flow of people, making the medical service process not smooth and unable to meet the needs of patients for efficient treatment;

[0006] The current medical space layout design lacks good adaptability and iterative ability. When a new large medical device is introduced into the hospital, due to the lack of appropriate space reserved in the previous layout, there is no place to place the medical device, and large-scale renovation is required, which is costly and affects the normal operation of the hospital. When a new service item appears, the existing layout cannot be flexibly adjusted to adapt to the new treatment process, which hinders the continuous improvement of medical services and cannot provide strong support for the development of medical services. Summary of the Invention

[0007] To address the above problems, the present invention provides an optimization system for the design of medical space layout based on artificial intelligence. Through the space layout module, according to the preprocessed data, a space layout sequence and space layout constraints are generated, the initial space layout is converted into the current space layout, the space layout is continuously optimized, various areas and facilities in the medical space are reasonably arranged, the phenomenon of space waste is reduced, the action routes of patients and medical staff are reasonably planned, the overall utilization efficiency of the medical space is improved, and the limited medical space can carry more medical service activities; through the layout optimization module, a layout optimization model is constructed, the current space layout is projected onto the spatial feature dimension, context information is captured, the context information is analyzed using the layout optimization model to generate a label distribution, a momentum term is introduced using the backpropagation method, the layout optimization model is trained, and the context information is updated in combination with the spatial features, thereby generating a layout optimization result, and reasonable layout optimization is performed according to the patient flow and treatment process of different departments. The system can quickly adapt to the changing needs of the hospital and provide strong support for medical services.

[0008] An optimization system for the design of medical space layout based on artificial intelligence, comprising a data acquisition module, a data preprocessing module, a space layout module, and a layout optimization module;

[0009] The data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the space layout module, and the space layout module is connected to the layout optimization module;

[0010] The data acquisition module collects the layout design data of the medical space;

[0011] The data preprocessing module performs preprocessing operations on the layout design data to obtain preprocessed data, extracts spatial features from the preprocessed data, and integrates the preprocessed data to form a global dictionary;

[0012] The space layout module constructs an initial space layout, generates a space layout sequence and space layout constraints, executes the space layout sequence under the condition of observing the space layout constraints, converts the initial space layout into the current space layout, and generates feedback information of the space layout sequence;

[0013] The layout optimization module, according to the feedback information of the generated space layout sequence, constructs a layout optimization model, calculates the space layout planning loss, trains and optimizes the layout optimization model by minimizing the space layout planning loss, and generates a layout optimization result.

[0014] The process by which the data preprocessing module performs preprocessing operations on the layout design data to obtain preprocessed data and generates a global dictionary includes the following steps:

[0015] Step S1: Data supplementation: Check the layout design data and supplement the missing values of the layout design data;

[0016] Step S2: Data conversion: clarify the layout design requirements, and according to the layout design requirements, use one-hot encoding to convert the layout design data format to obtain preprocessed data;

[0017] Step S3: Feature extraction: extract spatial features from the preprocessed data. The spatial features include the area, shape, and positional relationship of each department, and count the usage features from the preprocessed data. The usage features include the personnel flow and equipment usage frequency in each area during different time periods;

[0018] Step S4: Generate a global dictionary: integrate the preprocessed data, sort out various elements and attributes of the preprocessed data, and assign unique identifiers to various elements and attributes to form a global dictionary.

[0019] The spatial layout module, in the process of constructing an initial spatial layout, generating a spatial layout sequence and spatial layout constraints, executing the spatial layout sequence under the condition of observing the spatial layout constraints, converting the initial spatial layout into the current spatial layout, and generating feedback information of the spatial layout sequence, includes the following steps:

[0020] Step A1: Define the spatial layout: according to the preprocessed data, construct a graph structure and a relationship set to represent the spatial layout plan;

[0021] Step A2: Generate the initial spatial layout: use a spatial layout generator, according to the graph structure of the planned spatial layout, receive the preprocessed data, relationship set, and global dictionary, and generate the initial spatial layout;

[0022] Step A3: Generate spatial layout constraints: according to the initial spatial layout, construct a spatial layout planner to generate a spatial layout sequence and spatial layout constraints;

[0023] The formula used to generate the spatial layout constraints is specifically as follows:

[0024] ;

[0025] Among them, i represents the layout object number, j represents traversing different layout areas, k represents the total number of layout areas, represents the constraint of the i-th layout object, represents the secondary layout object, represents the layout area of the layout object, represents other layout areas, represents the distance from the i-th layout object to the i-th layout area, represents the distance from the i-th layout object to the j-th layout area;

[0026] Step A4: Update the current spatial layout: Execute the spatial layout sequence under the condition of complying with the spatial layout constraints, convert the initial spatial layout into the current spatial layout, and generate feedback information of the spatial layout sequence;

[0027] Step A5: Iteratively update the spatial layout: Set the end flag, repeat Steps A2 - A4, and execute the spatial layout sequence until the end flag is obtained.

[0028] The layout optimization module constructs a layout optimization model, generates the label distribution of the current spatial layout, determines the layout optimization metrics, calculates the spatial layout planning loss, and trains and optimizes the layout optimization model by minimizing the spatial layout planning loss to generate the layout optimization result. The process includes the following steps:

[0029] Step L1: Project the spatial layout: Construct a probability encoder and a deterministic decoder, project the current spatial layout of the medical space into the dimension of spatial features through a multi - layer perceptron, and capture the spatial layout context information according to the feedback information of the generated spatial layout sequence;

[0030] Step L2: Generate the label distribution: Construct a layout optimization model, use the layout optimization model to analyze the spatial layout context information provided by the current spatial layout, and obtain the label distribution of the current spatial layout;

[0031] Step L3: Optimize the layout optimization model: Determine the layout optimization metrics according to the label distribution of the current spatial layout, calculate the spatial layout planning loss, train and optimize the layout optimization model, combine the spatial features of the current spatial layout, update the spatial layout context information, and generate the layout optimization result.

[0032] Furthermore, the layout optimization model includes a prior encoder, a posterior encoder, a decoder, a re - encoder, a convolutional encoder, and a layout optimization planner; the prior encoder, posterior encoder, decoder, and re - encoder all include a transformer layer, a repeated self - attention block, a cross - attention block, and a multi - layer perceptron; the layout optimization planner includes two cross - attention blocks.

[0033] Furthermore, Step L2 includes the following steps:

[0034] Step L21: Calculate the parameters of the posterior distribution: Use the posterior encoder to calculate the parameters of the posterior distribution and output the spatial layout information through the cross - attention block;

[0035] Step L22: Calculate the parameters of the prior distribution: Use the transformer layer in the prior encoder, set the initial hypothesis, iteratively adjust the spatial layout context information using the initial hypothesis for noise detection, estimate the mean vector and diagonal covariance matrix of the prior distribution, and calculate the parameters of the prior distribution to describe the spatial layout information error;

[0036] Step L23: Feature information fusion: Use the decoder to generate a set of bounding boxes representing the spatial layout information error. Through repeated self-attention blocks and cross-attention blocks, perform feature information fusion on the initial hypothesis and the spatial layout information error to form fused features.

[0037] Step L24: Generate label distribution: Use the re-encoder to re-encode the fused features to form encoded features. At the same time, obtain the spatial features of the current spatial layout from the convolutional encoder. Use the two cross-attention blocks in the layout optimization planner to focus on the encoded features and spatial features respectively to obtain the label distribution of the current spatial layout.

[0038] Further, step L3 includes the following steps:

[0039] Step L31: Calculate the spatial layout planning loss: Use the K-means clustering algorithm to aggregate the label distribution of the current spatial layout. Determine the layout optimization metrics based on the label distribution of the current spatial layout. Combine the layout optimization metrics with the prior distribution to calculate the spatial layout planning loss.

[0040] The formula for calculating the spatial layout planning loss is specifically as follows:

[0041] ;

[0042] Where, represents the spatial layout planning loss, p represents the channel point of the spatial layout, P represents the set of channel points, represents the boundary of the i-th layout object, represents the bounding box of the i-th layout object, represents the distance from the channel point p to the boundary of the layout object i;

[0043] Step L32: Optimize the layout optimization model: Use the backpropagation method to introduce a momentum term to train the layout optimization model. Combine the spatial features of the current spatial layout to update the spatial layout context information and generate the layout optimization result.

[0044] The formula for using the backpropagation method to introduce a momentum term to train the layout optimization model is specifically as follows:

[0045] ;

[0046] Where, m represents the momentum coefficient, v represents the speed of the layout optimization model, represents the learning rate, represents the gradient of the spatial layout planning loss with respect to the latent variable at time t.

[0047] The beneficial effects achieved by the present invention are as follows:

[0048] (1) Through the spatial layout module, the system constructs a graph structure and a relationship set based on the preprocessed data, generates an initial spatial layout, a spatial layout sequence, and spatial layout constraints, making the layout of each area and facility in the medical space more scientific, reducing space waste, and at the same time being able to reasonably plan the movement routes of patients and medical staff, reducing unnecessary walking distances, improving the overall utilization efficiency of the medical space, and enabling the limited medical space to carry more medical service activities;

[0049] (2) Through the layout optimization module, the system constructs a layout optimization model, which can perform reasonable layout optimization according to the patient flow and treatment processes of different departments, effectively shorten the medical treatment time of patients, make the medical service process smoother, and improve the medical experience and satisfaction of patients;

[0050] (3) The system has good adaptability and iterative ability. When the spatial layout module updates the spatial layout, it will generate feedback information of the spatial layout sequence. The layout optimization module combines the spatial characteristics of the current spatial layout, updates the spatial layout context information, and continuously optimizes the layout optimization model, which can adapt to new demand changes and provide strong support for the continuous improvement of medical services. Brief Description of the Drawings

[0051] Figure 1 It is the method flowchart of the spatial layout module proposed by the present invention;

[0052] Figure 2 It is the system architecture diagram of the layout optimization model proposed by the present invention. Detailed Embodiments

[0053] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention.

[0054] Embodiment 1: Refer to Figure 1 , this embodiment provides a medical space layout design and optimization system based on artificial intelligence, including a data acquisition module, a data preprocessing module, a spatial layout module, and a layout optimization module;

[0055] The data acquisition module is connected to the data preprocessing module, the data preprocessing module is connected to the spatial layout module, and the spatial layout module is connected to the layout optimization module;

[0056] The data acquisition module collects the layout design data of the medical space;

[0057] The data preprocessing module performs preprocessing operations on the layout design data, obtains preprocessed data, extracts spatial features from the preprocessed data, and integrates the preprocessed data to form a global dictionary;

[0058] The spatial layout module constructs an initial spatial layout, generates a spatial layout sequence and spatial layout constraints, executes the spatial layout sequence while observing the spatial layout constraints, converts the initial spatial layout into the current spatial layout, and generates feedback information on the spatial layout sequence;

[0059] The layout optimization module constructs a layout optimization model based on the feedback information of the generated spatial layout sequence, calculates the spatial layout planning loss, trains and optimizes the layout optimization model by minimizing the spatial layout planning loss, and generates a layout optimization result.

[0060] Example 2: This example is based on Example 1. The data preprocessing module performs preprocessing operations on the layout design data to obtain preprocessed data. The process of generating the global dictionary includes the following steps:

[0061] Step S1: Data supplementation: Check the layout design data and supplement the missing values in the layout design data;

[0062] Step S2: Data conversion: Clarify the layout design requirements, and according to the layout design requirements, use one-hot encoding to convert the layout design data format to obtain preprocessed data;

[0063] Step S3: Feature extraction: Extract spatial features from the preprocessed data. The spatial features include the area, shape, and positional relationship of each department. Statistical usage features are extracted from the preprocessed data. The usage features include the personnel flow and equipment usage frequency in each area during different time periods;

[0064] Step S4: Generate the global dictionary: Integrate the preprocessed data, sort out various elements and attributes of the preprocessed data, and assign unique identifiers to various elements and attributes to form a global dictionary.

[0065] Example 3: This example is based on Example 2. The process by which the spatial layout module constructs an initial spatial layout, generates a spatial layout sequence and spatial layout constraints, executes the spatial layout sequence while observing the spatial layout constraints, converts the initial spatial layout into the current spatial layout, and generates feedback information on the spatial layout sequence includes the following steps:

[0066] Step A1: Define the spatial layout: According to the preprocessed data, construct a graph structure and a relationship set to represent the spatial layout plan;

[0067] Step A2: Generate the initial spatial layout: Use a spatial layout generator. According to the graph structure of the planned spatial layout, receive the preprocessed data, the relationship set, and the global dictionary to generate the initial spatial layout;

[0068] Step A3: Generate spatial layout constraints: According to the initial spatial layout, construct a spatial layout planner to generate a spatial layout sequence and spatial layout constraints;

[0069] Generate spatial layout constraints, and the specific formula used is as follows:

[0070] ;

[0071] where i represents the layout object number, j represents traversing different layout regions, and k represents the total number of layout regions, represents the constraint of the i-th layout object, represents the secondary layout object, represents the layout region of the layout object, represents other layout regions, represents the distance from the i-th layout object to the i-th layout region, represents the distance from the i-th layout object to the j-th layout region;

[0072] Step A4: Update the current spatial layout: Execute the spatial layout sequence under the condition of observing the spatial layout constraints, convert the initial spatial layout into the current spatial layout, and generate feedback information of the spatial layout sequence,

[0073] Step A5: Iteratively update the spatial layout: Set an end marker, repeat steps A2 - A4, and execute the spatial layout sequence until the end marker is obtained.

[0074] Embodiment 4: This embodiment is based on Embodiment 2. The process of the spatial layout module constructing an initial spatial layout, generating a spatial layout sequence and spatial layout constraints, executing the spatial layout sequence under the condition of observing the spatial layout constraints, converting the initial spatial layout into the current spatial layout, and generating feedback information of the spatial layout sequence includes the following steps:

[0075] Step A1: Define the spatial layout structure: According to the preprocessed data, construct a graph structure and a relationship set for planning the spatial layout;

[0076] Step A2: Generate the initial spatial layout: Use a spatial layout generator to generate an initial spatial layout according to the graph structure of the planned spatial layout, receiving the preprocessed data, the relationship set, and the global dictionary;

[0077] Step A3: Generate spatial layout constraints: According to the initial spatial layout, construct a spatial layout planner to generate a spatial layout sequence and spatial layout constraints;

[0078] Generate spatial layout constraints, and the specific formula used is as follows:

[0079] ;

[0080] where i represents the layout object number, j represents traversing different layout regions, and k represents the total number of layout regions, Represents the constraint of the i-th layout object. Represents a secondary layout object. Represents the layout area of the layout object. Represents other layout areas. Represents the distance from the i-th layout object to the i-th layout area. Represents the distance from the i-th layout object to the j-th layout area.

[0081] Step A4: Update the current spatial layout: Execute the spatial layout sequence under the condition of observing the spatial layout constraints, convert the initial spatial layout into the current spatial layout, and generate feedback information of the spatial layout sequence.

[0082] Example 5: Refer to Figure 2 , this example is based on Example 3, and the layout optimization model includes a prior encoder, a posterior encoder, a decoder, a re-encoder, a convolutional encoder, and a layout optimization planner; the prior encoder, the posterior encoder, the decoder, and the re-encoder all include a transformer layer, a repeated self-attention block, a cross-attention block, and a multi-layer perceptron; the layout optimization planner includes two cross-attention blocks.

[0083] Example 6: This example is based on Example 4, and the layout optimization model includes a prior encoder, a posterior encoder, a decoder, a re-encoder, a convolutional encoder, a probability encoder, and a deterministic decoder.

[0084] Example 7: This example is based on Example 5. The process of constructing a layout optimization model, generating a label distribution of the current spatial layout, determining layout optimization metrics, calculating the spatial layout planning loss, and training and optimizing the layout optimization model by minimizing the spatial layout planning loss to generate a layout optimization result includes the following steps:

[0085] Step L1: Project the spatial layout: Construct a probability encoder and a deterministic decoder, project the current spatial layout of the medical space into the dimension of spatial features through a multi-layer perceptron, and capture spatial layout context information according to the feedback information of the generated spatial layout sequence;

[0086] Step L2: Generate a label distribution: Construct a layout optimization model, use the layout optimization model to analyze the spatial layout context information provided by the current spatial layout, and obtain the label distribution of the current spatial layout;

[0087] Step L2 includes the following steps:

[0088] Step L21: Calculate the parameters of the posterior distribution: Use the posterior encoder to calculate the parameters of the posterior distribution, and output spatial layout information through the cross-attention block;

[0089] Step L22: Calculate the prior distribution parameters: Use the transformer layer in the prior encoder, set the initial hypothesis, and iteratively adjust the spatial layout context information using the initial hypothesis for noise detection. Estimate the mean vector and diagonal covariance matrix of the prior distribution, and calculate the parameters of the prior distribution to describe the spatial layout information error.

[0090] Step L23: Feature information fusion: Use the decoder to generate a set of bounding boxes representing the spatial layout information error. Through repeated self-attention blocks and cross-attention blocks, perform feature information fusion on the initial hypothesis and the spatial layout information error to form fused features.

[0091] Step L24: Generate the label distribution: Use the re-encoder to re-encode the fused features to form encoded features. At the same time, obtain the spatial features of the current spatial layout from the convolutional encoder. Use the two cross-attention blocks in the layout optimization planner to focus on the encoded features and spatial features respectively to obtain the label distribution of the current spatial layout.

[0092] Step L3: Optimize the layout optimization model: According to the label distribution of the current spatial layout, determine the layout optimization metrics, calculate the spatial layout planning loss, train and optimize the layout optimization model, combine the spatial features of the current spatial layout, update the spatial layout context information, and generate the layout optimization result.

[0093] Step L3 includes the following steps:

[0094] Step L31: Calculate the spatial layout planning loss: Use the K-means clustering algorithm to aggregate the label distribution of the current spatial layout. Determine the layout optimization metrics according to the label distribution of the current spatial layout. Based on the layout optimization metrics and combined with the prior distribution, calculate the spatial layout planning loss.

[0095] The formula for calculating the spatial layout planning loss is specifically as follows:

[0096] ;

[0097] where, represents the spatial layout planning loss, p represents the channel point of the spatial layout, P represents the set of channel points, represents the boundary of the i-th layout object, represents the bounding box of the i-th layout object, represents the distance from the channel point p to the boundary of the layout object i;

[0098] Step L32: Optimize the layout optimization model: Use the backpropagation method to introduce a momentum term to train the layout optimization model, combine the spatial features of the current spatial layout, update the spatial layout context information, and generate the layout optimization result.

[0099] The backpropagation method is used to introduce a momentum term to train the layout optimization model. The specific formula used is as follows:

[0100] ;

[0101] where m represents the momentum coefficient, v represents the velocity of the layout optimization model, represents the learning rate, represents the gradient of the spatial layout planning loss with respect to the latent variables at time t.

[0102] Example 8: This example is based on Example 6. The process of the layout optimization module constructing a layout optimization model, generating the label distribution of the current spatial layout, determining the layout optimization index, calculating the spatial layout planning loss, and training and optimizing the layout optimization model to generate the layout optimization result includes the following steps:

[0103] Step L1: Project the spatial layout: Construct a probabilistic encoder and a deterministic decoder, project the current spatial layout of the medical space into the dimension of spatial features through a multi-layer perceptron, and capture the spatial layout context information according to the feedback information of the generated spatial layout sequence;

[0104] Step L2: Generate the label distribution: Construct a layout optimization model, use the layout optimization model to analyze the spatial layout context information provided by the current spatial layout, and obtain the label distribution of the current spatial layout;

[0105] Step L3: Optimize the layout optimization model: Determine the layout optimization index according to the label distribution of the current spatial layout, further calculate the spatial layout planning loss, train and optimize the layout optimization model, combine the spatial features of the current spatial layout, update the spatial layout context information, and generate the layout optimization result.

[0106] The above describes the present invention and its implementation manners. Such a description is not restrictive. If those of ordinary skill in the art are inspired by it and without departing from the gist of the present invention, they design similar structural manners and embodiments to this technical solution without creative efforts, which shall fall within the protection scope of the present invention.

Claims

1. A medical space layout design optimization system based on artificial intelligence, comprising a data acquisition module and a data preprocessing module, wherein the data acquisition module collects layout design data of the medical space; characterized in that: The system also includes a spatial layout module and a layout optimization module; The data preprocessing module performs preprocessing operations on the layout design data to obtain preprocessed data, extracts spatial features from the preprocessed data, and integrates the preprocessed data to form a global dictionary; The spatial layout module generates feedback information of the spatial layout sequence; The layout optimization module constructs a layout optimization model according to the feedback information of generating the spatial layout sequence, calculates the spatial layout planning loss, trains and optimizes the layout optimization model by minimizing the spatial layout planning loss, and generates a layout optimization result; The process of generating feedback information of the spatial layout sequence by the spatial layout module includes the following steps: Step A1: Define spatial layout: Based on the preprocessed data, construct a graph structure and a set of relationships to represent the spatial layout plan; Step A2: Generate spatial layout: Use the spatial layout generator to receive preprocessed data, relationship set and global dictionary according to the graph structure of the planned spatial layout, and generate the initial spatial layout; Step A3: Define the planning task: Based on the initial spatial layout, build a spatial layout planner to generate spatial layout sequences and spatial layout constraints; Step A4: updating the spatial layout: executing the spatial layout sequence while complying with the spatial layout constraints, converting the initial spatial layout into the current spatial layout, and generating feedback information of the spatial layout sequence; Step A5: Perform iterative planning: set an end mark, repeat steps A2-A4, and execute the spatial layout sequence until the end mark is obtained.

2. The medical space layout design optimization system based on artificial intelligence according to claim 1, characterized in that: The process of generating the layout optimization result by the layout optimization module includes the following steps: Step L1: Projection spatial layout: construct a probabilistic encoder and a deterministic decoder, project the current spatial layout of the medical space to the dimension of spatial features through a multi-layer perceptron, and capture the spatial layout context information based on the feedback information of the generated spatial layout sequence; Step L2: Generate label distribution: Build a layout optimization model, use the layout optimization model to analyze the spatial layout context information provided by the current spatial layout, and obtain the label distribution of the current spatial layout; Step L3: Optimize the layout optimization model: According to the label distribution of the current spatial layout, determine the layout optimization index, calculate the spatial layout planning loss, train the optimized layout optimization model, combine the spatial features of the current spatial layout, update the spatial layout context information, and generate the layout optimization result.

3. The medical space layout design optimization system based on artificial intelligence according to claim 2, characterized in that: The layout optimization model includes a priori encoder, a posteriori encoder, a decoder, a re-encoder, a convolutional encoder and a layout optimization planner; the priori encoder, the posteriori encoder, the decoder and the re-encoder all include a transformer layer, a repeated self-attention block, a cross-attention block and a multi-layer perceptron; the layout optimization planner includes two cross-attention blocks.

4. The medical space layout design optimization system based on artificial intelligence according to claim 3 is characterized by: Step L2 includes the following steps: Step L21: Calculate the parameters of the posterior distribution: Use the posterior encoder to calculate the parameters of the posterior distribution and output the spatial layout information through the cross-attention block; Step L22: Calculate the prior distribution parameters: Use the transformer layer in the prior encoder to set the initial hypothesis, use the initial hypothesis to iteratively adjust the spatial layout context information for noise detection, estimate the mean vector and diagonal covariance matrix of the prior distribution, calculate the parameters of the prior distribution, and describe the spatial layout information error; Step L23: Feature information fusion: Use the decoder to generate a set of bounding boxes to represent the spatial layout information error. By repeating the self-attention block and the cross-attention block, the feature information of the initial hypothesis and the spatial layout information error is fused to form a fused feature. Step L24: Generate label distribution: Use the re-encoder to re-encode the fused features to form encoded features, and at the same time obtain the spatial features of the current spatial layout from the convolutional encoder. Use the two cross-attention blocks in the layout optimization planner to focus on the encoded features and spatial features respectively to obtain the label distribution of the current spatial layout.

5. The medical space layout design optimization system based on artificial intelligence according to claim 4 is characterized in that: Step L3 includes the following steps: Step L31: Calculate the spatial layout planning loss: Use the K-means clustering algorithm to aggregate the label distribution of the current spatial layout, determine the layout optimization index, and calculate the spatial layout planning loss based on the layout optimization index combined with the prior distribution; Step L32: Optimize the layout optimization model: Use the back-propagation method to introduce the momentum term to train the layout optimization model, combine the spatial features of the current spatial layout, update the spatial layout context information, and generate the layout optimization result.

Images