A medical assistance method and system based on artificial intelligence

Abstract

The present invention relates to a medical assistance method and system based on artificial intelligence. The method includes: Step S1: Collect slice images of a target to obtain a slice image sequence; Step S2: Perform target segmentation on each image in the slice image sequence to obtain a target image sequence; Step S3: Convert each target image in the target image sequence to a grayscale color space, and determine a first grayscale feature vector of each target image based on the grayscale value of the target image; Step S4: Determine the grayscale histogram of the target image, and determine a second grayscale feature vector based on the grayscale histogram; Step S6: Obtain an auxiliary analysis result based on the first grayscale feature vector and the second grayscale feature. The information content of the present invention is complex but easy to obtain, has strong compatibility, and low computational complexity.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence, and particularly relates to a medical assistance method and system based on artificial intelligence.

Background Art

[0002] The application of artificial intelligence in the medical field has shown great potential. It can provide assistance and support in various aspects, from disease prediction and diagnosis to treatment plan formulation and monitoring, and even including patient management and medical resources. Artificial intelligence-assisted diagnosis and treatment is based on big data intelligence and needs to solve the problem of fragmented medical and health data, achieve the leap from data to knowledge and from knowledge to intelligence, break through data islands, establish a cross-domain medical knowledge center linking individuals and medical institutions, and form an open and interconnected medical information sharing mechanism. Artificial intelligence can help doctors identify and predict diseases at an early stage through image recognition and pattern analysis technologies. Artificial intelligence can analyze medical imaging data, provide automatic detection, segmentation, and classification of tumors, and help doctors determine whether there are abnormalities, thereby improving the accuracy and sensitivity of early diagnosis. In addition, based on historical data and machine learning algorithms, artificial intelligence can build disease prediction models to predict the possible disease risks of patients in the future. This can provide early warnings and personalized suggestions about disease risks to doctors and patients, so as to take corresponding preventive measures and adjust lifestyles. Artificial intelligence can analyze patients' medical record data, symptoms, and laboratory test results, and combine large-scale medical databases and clinical guidelines to provide personalized treatment suggestions and plans for doctors.

[0003] In the prior art, there are still certain difficulties in medical assistance based on artificial intelligence. On the one hand, medical assistance is often based on image analysis, but using images as the input of an artificial intelligence model will bring considerable complexity, and even make the training of the artificial intelligence model unachievable. On the other hand, the creation of artificial intelligence models cannot keep up with the pace of diverse target types, lacking a general and compatible medical assistance means. The present invention supplements the possible information loss caused by image slicing, achieving dimensionality reduction of the computational amount; the input vector can reflect the semantic and non-semantic features of the image, with complex information content but simple acquisition; the target type is replaceable, so it has strong compatibility;

Summary of the Invention

[0004] In order to solve the above problems in the prior art, the present invention proposes a medical assistance method and system based on artificial intelligence, and the method includes:

[0005] Step S1: Collect sliced images of the target to obtain a sequence of sliced images;

[0006] Step S2: Perform target segmentation on each image in the sliced image sequence to obtain a target image sequence;

[0007] Step S3: Convert each target image in the target image sequence to the grayscale color space, and determine the first grayscale feature vector of each target image based on the grayscale value of the target image; where: the first grayscale feature vector indicates the grayscale value quantization feature of the target image;

[0008] The specific steps of step S3 are as follows:

[0009] Step S31: Divide the target image into U1 regions in the first manner, calculate the average grayscale value of each region, and arrange them in a preset order as the 1st to U1st element values in the first grayscale feature vector;

[0010] Step S32: Divide the target image into U2 regions in the second manner, calculate the average grayscale value of each region, and arrange them in a preset order as the (U1 + 1)th to (U1 + U2)th element values in the first grayscale feature vector;

[0011] Step S33: Calculate the average grayscale value of the entire target image as the (U1 + U2 + 1)th element value in the first grayscale feature vector;

[0012] Step S4: Determine the grayscale histogram of the target image, determine the second grayscale feature vector based on the grayscale histogram, and determine the grayscale histogram of the target image; the second grayscale feature vector indicates the distribution and deviation characteristics of the grayscale histogram frequency values;

[0013] The specific steps of step S4 are as follows:

[0014] Step S41: Obtain the frequency mean, maximum value, and minimum value of the grayscale histogram;

[0015] Step S42: Determine the first partitioning method based on the frequency mean, maximum value, and minimum value;

[0016] Step S43: Divide the gray levels of the grayscale histogram into L1 first intervals based on the first partitioning method; the gray level is the X-axis coordinate of the histogram;

[0017] Step S44: Calculate the frequency mean in each first interval, and arrange the frequency means in the order of the first intervals to form the 1st to L1st element values in the second grayscale feature vector; where: l1 is the first interval number;

[0018] Step S45: Calculate the frequency deviation value df of each first interval based on the following formula (1) l1 ; where: e1 is the e1th element in the l1th first interval; fre1,l1 is the frequency value of the e1-th element; E1 is the number of elements in the l1-th first interval;

[0019]

[0020] Step S46: Calculate the first interval deviation difference sbdf based on the following formula (2) l1 ; Arrange the first interval deviation difference sbdf l1 in the order of intervals to form the (L1 + 1)-th to 2L1-th element values in the second gray-scale feature vector;

[0021]

[0022] Step S47: Arrange the frequency mean, maximum value, and minimum value in order to form the (2L1 + 1)-th to (2L1 + 3)-th element values in the second gray-scale feature vector;

[0023] Step S5: Construct the input vector of the auxiliary analysis model based on the first gray-scale feature vector and the second gray-scale feature vector; Input the input vectors of each target image into the auxiliary analysis model in sequence to obtain multiple output vectors for each slice image; Determine the medical auxiliary analysis result based on the multiple output vectors.

[0024] Further, the first partitioning method is to evenly partition the gray levels into L1 first intervals, where L1 is a preset value.

[0025] Further, the first method is grid partitioning.

[0026] Further, U1 = U2 = 64.

[0027] Further, L1 = 9 or 16.

[0028] Further, the auxiliary analysis model is an artificial intelligence model.

[0029] Further, the auxiliary analysis model is a classification model based on a neural network.

[0030] A medical assistance platform based on artificial intelligence, including a processor, the processor is coupled with a memory, and the memory stores program instructions. When the program instructions stored in the memory are executed by the processor, the described medical assistance method based on artificial intelligence is implemented.

[0031] A computer-readable storage medium, including a program, when it runs on a computer, causes the computer to execute the described medical assistance method based on artificial intelligence.

[0032] A medical assistance system based on artificial intelligence, the system being configured to execute the medical assistance method based on artificial intelligence described above.

[0033] The beneficial effects of the present invention include:

[0034] (1) Through the multi-dimensional expression of image gray-scale features, compared with full-image analysis, the analysis complexity is greatly reduced; at the same time, image slices are used to supplement the possible information loss, thus achieving dimensionality reduction of the computational amount; in addition, target segmentation based on a preset shape is introduced to increase the diversity of the input, enhance the robustness of the artificial intelligence model, and overcome the differences in acquisition devices and acquisition environments.

[0035] (2) By means of image two-dimensional area division, gray-level histogram division, and target division considering target types, the dimension of image information acquisition is increased; and calculation methods such as deviation difference and frequency deviation value based on the division method are simple and do not involve iterative operations, which can reflect the semantic and non-semantic features of the image, making the input vector information complex but easy to obtain.

[0036] In particular, the third gray-scale feature vector that interweaves gray-scale color space and gray-scale histogram information has rich information expression ability; it can be independently used for obtaining auxiliary information; the target type is replaceable, so it has strong compatibility.

[0037] (3) Through the iterative merging of multi-slice image output vectors, the output vector merging that conforms to the auxiliary logic is realized, and further, through the weighted merging based on element units, the auxiliary requirements for different target features are considered, increasing the personalization and usability of the auxiliary results.

BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings described herein are used to provide a further understanding of the present invention, form a part of this application, but do not constitute an improper limitation to the present invention. In the drawings:

[0039] Figure 1 is a schematic diagram of the medical assistance method based on artificial intelligence provided by the present invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] The present invention will be described in detail below in conjunction with the drawings and specific embodiments, in which the illustrative embodiments and descriptions are only used to explain the present invention, but do not limit the present invention.

[0041] The present invention provides a medical assistance method and system based on artificial intelligence. As shown in the attached Figure 1 figures, the method includes the following steps:

[0042] Step S1: Collect sliced images of the target to obtain a sequence of sliced images; align the positions and interval points of the sectional images so that the targets corresponding to all the sliced images in the sequence of sliced images are aligned in three-dimensional space, and the i-th image and the i-th interval point in the sequence of sliced images correspond; the number of interval points corresponding to different patients is the same;

[0043] Through the alignment in three-dimensional space, the placement positions of the targets of different users in three-dimensional space are fixed; for example: when the target is the heart, the longest diameter and the shortest diameter of the target coincide with two three-dimensional coordinate axes respectively, so that the targets of all users can be aligned in three-dimensional space; of course, for other targets, similar or different alignment methods can also be used, which depends on the inherent shape of the target;

[0044] Step S2: Segment each image in the sequence of sliced images to obtain a sequence of target images; after segmentation, the target images still remain aligned; common image segmentation methods can be used for target segmentation;

[0045] Alternatively: segment the image with a preset shape; make each side or a preset proportion of the sides of the preset shape tangent to the target; further: the preset shape is a square; that is to say, the preset shape is fixed for segmenting each sliced image in the same sequence of sliced images, but its side length changes with the size of the target in different slices; through the preset shape, the diversity of the input of the subsequent artificial intelligence model can be increased, thereby bringing more information related to the sliced images; increasing the robustness of the artificial intelligence model and overcoming the differences in acquisition equipment and acquisition environment;

[0046] The present invention greatly reduces the analysis complexity compared with full-image analysis through the multi-dimensional expression of the image gray-scale features; at the same time, supplements the possible information loss caused by image slicing, thereby realizing the dimensionality reduction of the calculation amount; in addition, introducing the target segmentation based on the preset shape increases the input diversity, increases the robustness of the artificial intelligence model, and overcomes the differences in acquisition equipment and acquisition environment;

[0047] Step S3: Convert each target image in the sequence of target images to the gray-scale color space, and determine the first gray-scale feature vector of each target image based on the gray-scale value of the target image; where: the first gray-scale feature vector indicates the gray-scale value quantization feature of the target image;

[0048] The step S3 specifically includes the following steps:

[0049] Step S31: Divide the target image into U1 regions according to the first method, calculate the average gray-scale value of each region and arrange them in a preset order as the 1st to U1th element values in the first gray-scale feature vector;

[0050] Preferably, the value of U1 is related to the amount of calculation, and the larger the value of U1, the greater the amount of calculation;

[0051] Preferably, the first method is grid division;

[0052] Step S32: Divide the target image into U2 regions according to the second method, calculate the average gray value of each region, and arrange them in a preset order as the (U1 + 1)-th to (U1 + U2)-th element values in the first gray feature vector;

[0053] Preferably, the first method is raster division;

[0054] Preferably, U1 and U2 are preset values, and U1 = U2 = 64;

[0055] Step S33: Calculate the average gray value of the entire target image as the (U1 + U2 + 1)-th element value in the first gray feature vector;

[0056] Step S4: Determine the gray histogram of the target image, determine the second gray feature vector based on the gray histogram, and determine the gray histogram of the target image; the second gray feature vector indicates the distribution and deviation characteristics of the gray histogram frequency values;

[0057] The specific steps of step S4 are as follows:

[0058] Step S41: Obtain the frequency mean, maximum value, and minimum value of the gray histogram;

[0059] Step S42: Determine the first division method based on the frequency mean, maximum value, and minimum value;

[0060] Preferably, the corresponding relationship between the frequency mean, maximum value, minimum value, and the first division method is preset, and the first division method is determined by looking up the corresponding relationship;

[0061] Preferably, the first division method is to evenly divide the gray levels into L1 first intervals;

[0062] Step S43: Divide the gray levels of the gray histogram into L1 first intervals based on the first division method; the gray level is the X-axis coordinate of the histogram;

[0063] Step S44: Calculate the frequency mean in each first interval, and arrange the frequency means in the order of the first intervals to form the 1st to L1-th element values in the second gray feature vector; where: l1 is the first interval number;

[0064] Preferably, L1 is a preset value; for example: L1 = 9, 16;

[0065] Step S45: Calculate the frequency deviation value df of each first interval based on the following formula (1) l1 ; where: e1 is the e1-th element in the l1-th first interval; fr e1,l1 is the frequency value of the e1-th element; E1 is the number of elements in the l1-th first interval;

[0066]

[0067] Step S46: Calculate the first interval deviation difference sbdf based on the following formula (2) l1 ; Arrange the first interval deviation difference sbdf l1 in the order of intervals to form the (L1 + 1)-th to the 2L1-th element values in the second gray feature vector;

[0068]

[0069] Alternatively: Calculate the first interval deviation difference based on the following formula (3); This calculation method can reduce the training amount to a certain extent but will cause a certain amount of information loss;

[0070]

[0071] Step S47: Arrange the frequency mean, maximum value, and minimum value in order to form the (2L1 + 1)-th to the (2L1 + 3)-th element values in the second gray feature vector;

[0072] Preferably: The method further includes Step S5;

[0073] Step S5: Perform region division on the target image based on the target type to obtain multiple sub-target regions, and determine the third gray feature vector based on the gray color space and gray histogram; The third gray feature vector indicates the inherent gray and frequency characteristics between regions of the sub-target regions based on the target type;

[0074] The specific steps of Step S5 are as follows:

[0075] Step S51: Determine the first target division method corresponding to the target type based on the target type, and divide the target image into T1 first sub-target regions; The target division method corresponds to a specific gray segmentation corresponding to each first sub-target region; And this specific gray segmentation can show the gray characteristics of the region to a certain extent; When the segmentation is large, the gray characteristics are blurred. Similarly, by reducing the segmentation, the gray characteristics can be accurately characterized; It can be set according to needs;

[0076] Step S52: Sequentially obtain an unprocessed first sub-target region Ar t1; where: t1 is the first sub-target area number;

[0077] Step S53: Determine the unprocessed first sub-target area Ar t1 in its corresponding specific gray level segment [gL t1 , gH t1 , the proportion gP of the number of elements in the total number of elements ET1 in the first sub-target area; t1 Specifically: The following formulas (4) and (5) are used to calculate the proportion gP t1 ; where: gr et1,t1 is the gray level value corresponding to the et1-th element in the first sub-target area Ar t1 ;

[0078] gP t1 = ∑Tp t1 / ET1 (4);

[0079]

[0080] Step S54: Arrange the proportion gP t1 in the order of the first sub-target area Ar t1 to form the 1st to T1st element values in the third gray level feature vector;

[0081] Step S55: Based on the target type, determine the second target division method corresponding to the target type, and divide the target image into T2 second sub-target areas; The second target division method corresponds to the specific frequency band segment corresponding to each second sub-target area;

[0082] Preferably: The first target division method and the second target division method are the same;

[0083] Step S56: Sequentially obtain an unprocessed second sub-target area Ar t2 ; where: t2 is the second sub-target area number;

[0084] Step S57: Determine the unprocessed second sub-target area Ar t2 in its corresponding specific gray level segment [fL t2 , fH t2 , the frequency mean value Specifically: The following formula (6) is used to calculate the frequency mean value; where: fr et1,t2 is the frequency value corresponding to the et2-th element in the sub-target area Ar t2 ;

[0085]

[0086] Step S57: The frequency mean value Arrange in the order of the sub-goal area Ar t2 to form the (T1 + 1)-th to (T1 + T2)-th element values in the third grayscale feature vector;

[0087] The present invention increases the acquisition dimension of image information through the methods of image two-dimensional area division, grayscale-level histogram division, and target division considering the target type; and the deviation difference and frequency deviation value df l1 and other calculation methods are simple and do not involve iterative operations, which can reflect the semantic and non-semantic features of the image, making the input vector information complex but easy to obtain;

[0088] Step S6: Construct an input vector of the auxiliary analysis model based on the first grayscale feature vector and the second grayscale feature vector, and / or the third grayscale feature vector; sequentially input the input vectors of each target image into the auxiliary analysis model to obtain multiple output vectors for each slice image; determine the medical auxiliary analysis result based on the multiple output vectors;

[0089] Preferably: The auxiliary analysis model is an artificial intelligence model; the input of the auxiliary analysis model is the input vector, and the output is the medical auxiliary result; for example: one or more classifications and their corresponding probability values; or a probability value vector; where each element in the vector respectively indicates a classification;

[0090] Preferably: Obtain historical diagnosis data for constructing sample data including the input vector and the auxiliary analysis result, and use the sample data to train and verify the artificial intelligence model; the auxiliary analysis result includes one or more diagnostic feature results and their occurrence probabilities;

[0091] Alternatively: The auxiliary analysis model is an artificial intelligence model;

[0092] The determining of the medical auxiliary analysis result based on the multiple output vectors; specifically: preprocess the multiple output vectors respectively for screening the output vectors; then make pairwise comparisons of the output vectors, and merge the output vectors based on the comparison results, and finally obtain the medical auxiliary analysis result;

[0093] Preferably: The determining of the medical auxiliary analysis result based on the multiple output vectors; specifically includes the following steps;

[0094] Step S6X1: Calculate the similarity between any two output vectors;

[0095] Preferably: The similarity is measured by the Euclidean distance between the output vectors. The shorter the Euclidean distance, the greater the similarity, and vice versa;

[0096] Step S6X2: Judge any two output vectors. If there exists an output vector whose similarity with other output vectors is less than the similarity threshold, and the number of merging times ub participated by the output vector is less than the preset number of merging times ub1, then delete the output vector;

[0097] Preferably: the number of merging times ub1 = 1;

[0098] Step S6X3: Judge the output vector pair formed by any two output vectors. If the similarity is greater than the similarity threshold, put the two output vectors into the set of candidate output vector pairs; if the set of candidate output vector pairs is empty, go to Step S6X6; otherwise, go to the next step;

[0099] Step S6X4: Put the output vector pair with the maximum similarity in the set of candidate output vector pairs into the set of output vector pairs; and delete the output vector pairs in the set of candidate output vector pairs that involve any output vector in the output vector pair with the maximum similarity; repeat this step until the set of candidate output vector pairs is empty;

[0100] Step S6X5: Merge the two output vectors involved in each output vector pair in the set of output vector pairs; the merging method is to set the value of the merged element equal to the average value of the corresponding element values of the two output vectors;

[0101] Alternatively: the merging method is to set the value of the merged element equal to the maximum value of the corresponding element values of the two output vectors;

[0102] Step S6X6: Weightedly merge the remaining output vectors V c =<v c,o > to obtain the auxiliary analysis result VS = <vs o >;

[0103] For the said Step S6X6, specifically: use the following formulas (7)(8) to determine the element value of the o-th element in the auxiliary analysis result; where: w s ∈w1~w S is the s-th preset weight value, w S is the S-th preset weight value; c ∈ 1~C is the remaining output vector number, C is the number of the remaining output vectors; Seq(v c,o ,"v 1,o ,…,v c,o ,…,v C,o ") is a sorting function, and its output is the permutation order number of v c,o in "v 1,o ,…,v c,o ,…,v C,o ";

[0104] vs o = w s × v c,o (7);

[0105] s = Seq(v c,o ,"v 1,o ,…,v c,o ,…,v C,o ") (8);

[0106] It can be seen that at this time, for each vector, its weight value is not fixed, but different based on each element value; when merging any element of the vector, the larger the element value, the higher its weight, and vice versa; precisely because each element is for auxiliary prediction of different target features, its meaning is often different; therefore, multiple w s levels can be set; the smaller the s value, the smaller the number of v c,o , the higher its ranking, and the greater the corresponding weight value;

[0107] Through the iterative merging of the output vectors of multi-slice images, the merging of the output vectors that conforms to the auxiliary logic is realized, and further, through the weighted merging based on the element units, the auxiliary requirements for different target features are considered, and the personalization and usability of the auxiliary results are increased;

[0108] Based on the same inventive concept, the present invention also provides a medical assistance system based on artificial intelligence, and the system includes: a medical terminal and a medical assistance server; the system is used to implement the above-mentioned medical assistance method based on artificial intelligence;

[0109] Preferably: there are multiple medical terminals, and communication connections are maintained between the multiple medical terminals and the server;

[0110] Preferably: the server is a distributed server;

[0111] A computer program (also known as a program, software, software application, script, or code) can be written in any form of programming language, including assembly or interpreted languages, declarative or procedural languages, and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, object, or other unit suitable for use in a computing environment. A computer program may or may not correspond to a file in a file system. The program can be stored as part of a file that holds other programs or data (such as one or more scripts in a markup language document), in a single file dedicated to the program, or in multiple cooperating files (such as files that store one or more modules, subroutines, or code portions). A computer program can be deployed to execute on one computer or on multiple computers located at one site or distributed across multiple sites and interconnected by a communication network.

[0112] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) that contain computer-usable program code.

[0113] The present invention is described with reference to the flowcharts and / or block diagrams of methods, apparatuses (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowchart and / or block diagram, as well as the combination of flows and / or blocks in the flowchart and / or block diagram, can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate means for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0114] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing devices to work in a specific manner, such that the instructions stored in the computer-readable memory generate a manufactured article including instruction means that realizes the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0115] These computer program instructions can also be loaded onto a computer or other programmable data processing devices, such that a series of operation steps are executed on the computer or other programmable devices to generate a computer-implemented process, so that the instructions executed on the computer or other programmable devices provide steps for realizing the functions specified in Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: still can modify the specific implementation manners of the present invention or make equivalent replacements, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the protection scope of the claims of the present invention.

Claims

1. A medical assistance method based on artificial intelligence, characterized in that The method includes: Step S1: Collect slice images of the target to obtain a sequence of slice images; Step S2: Segment the target in each image in the sequence of slice images to obtain a sequence of target images; Step S3: Convert each target image in the sequence of target images to the grayscale color space, and determine the first grayscale feature vector of each target image based on the grayscale value of the target image; wherein: the first grayscale feature vector indicates the grayscale value quantization feature of the target image; The specific steps of step S3 include the following steps: Step S31: Divide the target image into U1 regions according to the grid division method, calculate the average grayscale value of each region, and arrange them in a preset order as the 1st to U1th element values in the first grayscale feature vector; Step S32: Divide the target image into U2 regions according to the raster division method, calculate the average grayscale value of each region, and arrange them in a preset order as the (U1 + 1)th to (U1 + U2)th element values in the first grayscale feature vector; Step S33: Calculate the average grayscale value of the whole target image as the (U1 + U2 + 1)th element value in the first grayscale feature vector; Step S4: Determine the grayscale histogram of the target image, and determine the second grayscale feature vector based on the grayscale histogram; the second grayscale feature vector indicates the distribution and deviation characteristics of the grayscale histogram frequency values of the target image; The specific steps of step S4 include the following steps: Step S41: Obtain the frequency average, maximum value, and minimum value of the grayscale histogram; Step S42: Determine the first division method based on the frequency average, maximum value, and minimum value; Step S43: Divide the gray levels of the grayscale histogram into L1 first intervals based on the first division method; the gray level is the X-axis coordinate of the histogram; L1 is a preset value; Step S44: Calculate the frequency mean in each first interval, and arrange the frequency means in the order of the first intervals to form the 1st to L1th element values in the second grayscale feature vector; where: is the first interval number; Step S45: Calculate the frequency deviation value of each first interval based on the following formula (1) ; where: is the e1-th element in the -th first interval; is the frequency value of the e1-th element; is the number of elements in the -th first interval; （1）； Step S46: Calculate the first interval deviation difference based on the following formula (2) ; Arrange the first interval deviation difference in the order of intervals to form the (L1 + 1)-th to 2L1-th element values in the second gray feature vector; （2）； Step S47: Arrange the frequency average, maximum value, and minimum value in order to form the (2L1 + 1)th to (2L1 + 3)th element values in the second grayscale feature vector; Step S5: Construct the input vector of the auxiliary analysis model based on the first grayscale feature vector and the second grayscale feature vector; sequentially input the input vectors of each target image into the auxiliary analysis model to obtain multiple output vectors for each slice image; determine the medical auxiliary analysis result based on the multiple output vectors; the input of the auxiliary analysis model is the input vector, and the output is the medical auxiliary result, and the auxiliary result is one or more classifications and their corresponding probability values; The determination of the medical auxiliary analysis result based on the multiple output vectors; specifically includes the following steps; Step S5X1: Calculate the similarity between any two output vectors; Step S5X2: Judge the any two output vectors. If there is an output vector whose similarity with other output vectors is less than the similarity threshold, and the number of merging times ub participated by the output vector is less than the preset number of merging times ub1, then delete the one output vector; ub1 = 1; Step S5X3: Determine the output vector pairs formed by any two of the output vectors. If the similarity is greater than the similarity threshold, put the two output vectors into the set of candidate output vector pairs; if the set of candidate output vector pairs is empty, go to Step S5X6; otherwise, go to the next step; Step S5X4: Put the output vector pair with the highest similarity in the set of candidate output vector pairs into the set of output vector pairs; and delete the output vector pairs in the set of candidate output vector pairs that involve any one of the output vectors in the output vector pair with the highest similarity; repeat this step until the set of candidate output vector pairs is empty; Step S5X5: Merge the two output vectors involved in each output vector pair in the set of output vector pairs; the merging method is to set the value of the merged element equal to the average of the corresponding element values of the two output vectors; Step S5X6: Weighted combination of the remaining output vectors to obtain the auxiliary analysis result ; The specific steps of step S5X6 are as follows: the element value of the o-th element in the auxiliary analysis result is determined by the following formulas (3) and (4); where: is the s-th preset weight value, is the S-th preset weight value; is the remaining output vector number, and C is the number of the remaining output vectors; is the sorting function, and its output is in the permutation order number; （3）； （4）。 2. The medical assistance method based on artificial intelligence according to claim 1, wherein U1 = U2 = 64.

3. The medical assistance method based on artificial intelligence according to claim 2, wherein, L1 = 9 or 16.

4. The medical assistance method based on artificial intelligence according to claim 3, characterized in that The auxiliary analysis model is an artificial intelligence model.

5. A medical assistance platform based on artificial intelligence, characterized in that, It includes a processor, the processor is coupled with a memory, and the memory stores program instructions. When the program instructions stored in the memory are executed by the processor, the medical assistance method based on artificial intelligence described in any one of claims 1-4 is implemented.

6. A computer-readable storage medium, characterized in that, It includes a program, which when running on a computer causes the computer to execute the medical assistance method based on artificial intelligence described in any one of claims 1-4.

7. A medical assistance system based on artificial intelligence, characterized in that, The system is configured to execute the medical assistance method based on artificial intelligence described in any one of claims 1-4.

Images