A method for locating lesion positions in medical images based on single photon quantum

Abstract

The present invention discloses a method for locating the position of a lesion in a medical image based on a single-photon quantum, which belongs to the field of quantum optics and medical experiments. By obtaining a medical image with a known lesion position and marking the lesion position, a lesion position label is obtained and encoded; a single photon is used to irradiate the medical image, a corresponding relationship between each pixel point in the medical image and the quantum state of the photon is established, and the quantum state of the medical image is obtained; a unitary transformation is performed on the quantum state of the medical image, and converted into a single-photon row vector carrying information of all pixel points, and the lesion position label is located according to the single-photon row vector; the parameters of the unitary transformation are continuously updated to repeatedly locate the lesion position label, and the most accurate optimization model for positioning is obtained. The quantum field is combined with optical experiments, and the problems of high risk and low efficiency in the process of locating the lesion position in medical images are solved through reasonable quantum computing methods and reasonable application of lesion position features.

Description

Technical Field

[0001] The invention relates to the field of optics and medical experiments, and in particular to a method for locating the position of a lesion in a medical image based on a single-photon quantum. Background Art

[0002] In recent years, the forms and types of medical images in medical diagnosis have increased, such as X-ray imaging, magnetic resonance imaging, nuclear medicine imaging, and ultrasonic imaging. In the current medical image processing process, doctors usually judge whether there is a lesion in the entire medical image based on their own experience. With the influence of the environment, equipment, and data volume, this work is time-consuming and requires a lot of energy.

[0003] Therefore, the use of computers to locate lesions in medical images not only reduces the workload of staff, but also promotes the intelligence and informatization of the medical industry. Many methods have been proposed to deal with this work, including intracranial image hemorrhage detection and location based on deep learning technology and iris image location method based on regional Hough transform. These methods are all proposed on the basis of deep learning and can meet the accuracy requirements of medical image location under specific conditions.

[0004] Quantum computing, as a fast computing method in computing systems, has been widely studied in recent years. For different machine learning tasks, the addition of quantum systems not only brings faster computing speeds to the algorithm, but also plays a great role in practical applications. However, the quantum gates in quantum computing hardware are easily affected by the environment and cause interference, so it is difficult to implement in experiments. Optical experiments have found that single photons can make quantum gates less susceptible to environmental influences at room temperature. Therefore, the method of using single-photon quantum image positioning algorithms to locate lesions in medical images deserves attention. Summary of the invention

[0005] The purpose of the present invention is to combine the quantum field with optical experiments to solve the problems of high risk and low efficiency in the process of locating the lesion position in medical images, and to provide an innovative quantum positioning method based on single photons.

[0006] The objective of the present invention is achieved through the following technical solutions:

[0007] A method for locating the position of a lesion in a medical image based on single photon quantum is provided, the method comprising:

[0008] Acquire a medical image with a known lesion location, and mark the lesion location to obtain a lesion location label;

[0009] Using a single photon to illuminate the medical image, establishing a corresponding relationship between each pixel point in the medical image and the quantum state of the photon, and obtaining the quantum state of the medical image; and encoding the lesion location label;

[0010] Performing a unitary transformation on the quantum state of the medical image and converting it into a single photon row vector carrying information of all pixel points;

[0011] Establishing a positioning model for the lesion position, and inputting the single photon row vector and the encoded lesion position label into the positioning model to locate the lesion position label;

[0012] Continuously updating the parameters of the unitary transformation to repeatedly locate the lesion position label to obtain an optimization model with the most accurate positioning;

[0013] The medical image to be tested is input into the optimization model to locate the lesion.

[0014] Specifically, a single medical image is composed of pixels with grayscale values ​​of 0-255 arranged in a matrix format with a size of 512*512.

[0015] Specifically, the using a single photon to illuminate the medical image includes:

[0016] Single photons are used to illuminate the medical image in a highly collimated manner, with the quantum state of the photons being in the same phase when they arrive at each individual pixel.

[0017] Specifically, the establishing of the correspondence between each pixel point in the medical image and the quantum state of a photon to obtain the quantum state of the medical image includes:

[0018] After photon irradiation, the Hilbert space is selected as a basis, and the correlation space of the quantum state of the medical image can be described by the elements of the M*N complex vector space, as follows:

[0019]

[0020] Among them, the basis function Corresponding to the photon passing through the pixel (m, n), the basis function The coefficient a mn is a non-negative real number;

[0021] Coefficient a mn It is proportional to the square root of the brightness of the medical image, that is:

[0022]

[0023] And the coefficient a mn Perform normalization:

[0024]

[0025] Specifically, performing unitary transformation on the quantum state of the medical image includes:

[0026] Quantum states of medical images Function A U α , so that the original quantum state becomes At the same time, the U α Needs to be satisfied The unitary constraint is implemented by regularizing the loss function;

[0027] The U α It can be obtained by the following formula:

[0028]

[0029] Among them, jS α By adjusting the weights T of a random matrix T during training α To achieve, that is:

[0030]

[0031] The weight T α It is input as a parameter into the squared error loss function.

[0032] Specifically, the quantum state after the unitary transformation Convert to a single grid form, where W and H are the height and width of a single grid respectively, N is the number of grids in each row or column, Ceils is the total number of grids, X is the Xth grid, and X∈[0, Ceils-1], Pixels represents the pixel points that the Xth grid should contain, and Pixels is a row vector, the size of the row vector is W*H;

[0033] The pixels of the Xth grid are calculated according to the following formula:

[0034] Pixels = [i × N + j]

[0035] i∈[X / / N×H,X / / N×H+H]

[0036] j∈[X%N×W,X%N×W+W],

[0037] Among them, i and j represent random values ​​according to pixel requirements.

[0038] Specifically, the calculation of the single photon row vector includes:

[0039] The obtained single grid of the medical image is converted into a W*H complex vector space, and the trace of each complex vector space is calculated to obtain the probability of the corresponding single grid;

[0040] The probabilities of all single grids are arranged from left to right and from top to bottom into a row vector, and the row vector corresponds to a lesion location label.

[0041] Specifically, the single grid is converted into a W*H matrix mode, where the trace of the matrix tr=e 11 +e 22 +...+e WH .

[0042] Specifically, the continuously updating the parameters of the unitary transformation to repeatedly locate the lesion position label to obtain the most accurate positioning optimization model includes:

[0043] After obtaining the probability of a single grid, the gradient descent algorithm is used to update the parameters T required by the square error loss function. α ;

[0044] By utilizing the principle of continuously decreasing square error loss function, we can obtain the most accurate optimization model for positioning.

[0045] Specifically, the lesion location positioning process is repeated, and the parameter T is continuously updated. α .

[0046] It should be further explained that the technical features corresponding to the above options can be combined or replaced with each other to form a new technical solution if there is no conflict.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] (1) The present invention mainly combines the quantum field with optical experiments, and through the reasonable application of reasonable quantum computing methods and lesion location characteristics, achieves the continuous improvement of the accuracy of lesion location positioning and the continuous acceleration of the positioning process simulation, thereby solving the problems of high risk and low efficiency in the process of lesion location positioning in medical images.

[0049] (2) The quantum state of the medical image is unitarily transformed and converted into a single-photon row vector that carries the information of all pixel points, thereby increasing the accuracy of positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0051] Figure 1 This is a flow chart of a method for locating the position of a lesion in a medical image based on single-photon quantum according to the present invention;

[0052] Figure 2 A schematic diagram of the conversion of the quantum state of a medical image into a single photon row vector according to the present invention;

[0053] Figure 3 It is a single grid matrix pattern of the present invention. DETAILED DESCRIPTION

[0054] The technical solution of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0055] In addition, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0056] The present invention mainly combines the quantum field with optical experiments, and through the reasonable application of reasonable quantum computing methods and lesion location characteristics, achieves the simulation of continuously improving the accuracy of lesion location and continuously accelerating the positioning process, thereby solving the problems of high risk and low efficiency in the process of lesion location positioning in medical images.

[0057] Before introducing the embodiments of the present application, some terms involved in the present application are first explained.

[0058] 1. Single photon: When light is extremely weak, it will be dispersed into individual photons, which are called single photons.

[0059] 2. Quantum state: a microscopic state determined by a set of quantum numbers.

[0060] 3. Pixel: refers to the smallest unit in an image represented by a digital sequence.

[0061] 4. Unitary transformation: In mathematics, a unitary transformation is a transformation that preserves the inner product. In quantum mechanics, a unitary transformation is a transformation from one representation to another.

[0062] 5. Hilbert space: a generalized complete inner product space.

[0063] 6. Gradient descent algorithm: It is an optimization algorithm that solves the minimum value along the direction of gradient descent.

[0064] Example 1

[0065] Provide a method for locating the position of lesions in medical images based on single photon quantum, such as Figure 1 As shown, the method includes:

[0066] Acquire a medical image with a known lesion location, and mark the lesion location to obtain a lesion location label;

[0067] Using a single photon to illuminate the medical image, establishing a corresponding relationship between each pixel point in the medical image and the quantum state of the photon, and obtaining the quantum state of the medical image; and encoding the lesion location label;

[0068] Performing a unitary transformation on the quantum state of the medical image and converting it into a single photon row vector carrying information of all pixel points;

[0069] Establishing a positioning model for the lesion position, and inputting the single photon row vector and the encoded lesion position label into the positioning model to locate the lesion position label;

[0070] Continuously updating the parameters of the unitary transformation to repeatedly locate the lesion position label to obtain an optimization model with the most accurate positioning;

[0071] The medical image to be tested is input into the optimization model to locate the lesion.

[0072] Specifically, in modern society, medical images come from patients, and they are images that reflect the internal structure or internal function of the anatomical region. Most medical images are single-channel grayscale images. Features are the essential features that distinguish images at the lesion location from other normal location images. Lesion location features are the key to detecting whether there are lesions in medical images, and their main features include morphological features, color depth features, texture features, etc. Taking cancer cells as an example in medical images, the main morphology of cells can be expressed as the height, perimeter, width, area, etc. of the structural elements. Through large-scale comparison, we can know the main differences between lesion cells and normal cells. Starting from the differences, we can obtain the characteristics of lesion cells. Similarly, medical images also include intracranial hemorrhage CT, eye disease images, medical scan images, etc.

[0073] The lesion position is located by establishing an optimization model, which is divided into a training process and a testing process. The training process includes loading medical imaging data, medical imaging data preprocessing, parameter generation and updating, and optimizer optimization model; the testing process includes loading medical imaging data, medical imaging preprocessing, input into the medical imaging lesion position positioning model, and output of the lesion position.

[0074] First, medical images with known lesion locations are obtained as training sets, and the location of the abnormal image in a single medical image in the training set is annotated. This annotation is the label position of the lesion location, and the lesion location label is obtained.

[0075] Secondly, the image data is mapped to a quantum state. After being irradiated by a single photon with high collimation, the absorption coefficient of the location area of ​​the medical image is reflected by the grayscale value of the pixel. That is to say, the coefficient of the quantum state is proportional to the pixel value of the medical image. The darker the grayscale image, the higher the pixel value, and the brighter the grayscale image, the lower the pixel value. After photon irradiation, the corresponding relationship between each pixel point in the medical image and the quantum state of the photon is established to obtain the quantum state of the medical image.

[0076] At the same time, the lesion location label is encoded. If a lesion location label is (2, 3), each number is one-hot encoded, that is, encoded as (00100000000, 00010000000), that is, the number of the digits is 1, and the other bits are 0.

[0077] The quantum state of the medical image is then unitarily transformed. The unitary transformation will generate a parameter for model training, which is input into the localization model to locate the lesion location label. Specifically, after calculating the corresponding feature probability of each single grid, the row vector formed by these probabilities and the encoded label are simultaneously input into the optimizer of the localization model.

[0078] The parameters of the unitary transformation are continuously updated to repeatedly locate the lesion position label, which is equivalent to the process of gradient optimization, and the most accurate optimization model for positioning is obtained. A medical image test picture is input. After the image is loaded and preprocessed, the specific position of the lesion image in the entire medical image can be obtained through the above training set process, thereby completing and outputting the positioning result.

[0079] The present invention mainly combines the quantum field with optical experiments, and through the reasonable application of reasonable quantum computing methods and lesion location characteristics, achieves the simulation of continuously improving the accuracy of lesion location and continuously accelerating the positioning process, thereby solving the problems of high risk and low efficiency in the process of lesion location positioning in medical images.

[0080] Example 2

[0081] Based on Example 1, a method for locating the position of a lesion in a medical image based on single-photon quantum is provided. A single medical image is composed of pixels with grayscale values ​​of 0-255 arranged in a matrix form with a size of 512*512.

[0082] Further, the using single photons to illuminate the medical image comprises:

[0083] The medical image is highly collimated using a single photon, and the quantum state of the photon is in the same phase when it reaches each individual pixel. Specifically, a photon passes through a single-channel grayscale medical image from a distance, and the image size should be small enough or the illumination range of the single photon is large enough and the light source can provide highly collimated illumination.

[0084] Furthermore, the establishing of the correspondence between each pixel point in the medical image and the quantum state of the photon to obtain the quantum state of the medical image includes:

[0085] After photon irradiation, the Hilbert space is selected as a basis, and the correlation space of the quantum state of the medical image can be described by the elements of the M*N complex vector space, as follows:

[0086]

[0087] Among them, the basis function Corresponding to the photon passing through the pixel (m, n), the basis function The coefficient a mn is a non-negative real number;

[0088] Coefficient a mn It is proportional to the square root of the brightness of the medical image, that is:

[0089]

[0090] And the coefficient a mn Perform normalization and take the square root of each pixel value and normalize it:

[0091]

[0092] Furthermore, the quantum state of the medical image is unitarily transformed to increase the accuracy of positioning, and the unitary transformation includes:

[0093] Quantum states of medical images Function A U α , so that the original quantum state becomes At the same time, the U α Needs to be satisfied The unitary constraint is implemented by regularizing the loss function;

[0094] The U α It can be obtained by the following formula:

[0095]

[0096] Among them, jS α By adjusting the weights T of a random matrix T during trainingα To achieve, that is:

[0097]

[0098] The weight T α Input as a parameter into the squared error loss function. The squared error loss function is: loss = label 2 -out 2 , Loss represents the loss function, label represents the location label, and out represents the output location.

[0099] Further, the quantum state after the unitary transformation Convert to a single grid form, take W and H as the height and width of a single grid, N is the number of grids in each row or column, Ceils is the total number of grids, X is the Xth grid, and X∈[0, Ceils-1], Pixels represents the pixel points that the Xth grid should contain, and Pixels is a row vector, such as Figure 2 As shown, the size of the row vector is W*H;

[0100] The pixels of the Xth grid are calculated according to the following formula:

[0101] Pixels = [i × N + j]

[0102] i∈[X / / N×H,X / / N×H+H]

[0103] j∈[X%N×W,X%N×W+W],

[0104] Among them, i and j represent random values ​​according to pixel requirements, and i and j are calculated from X, N, H, and W.

[0105] Furthermore, the calculation of the single photon row vector includes:

[0106] Then the medical image grayscale image arranged in matrix form is converted into a row vector, and the value of each pixel is normalized after taking the square root. Then a unitary transformation is implemented on the medical image grayscale image, in which the quantum state information of a single grid is taken out to form a row vector.

[0107] Specifically, the obtained single grid of the medical image is converted into a W*H complex vector space, and the trace of each complex vector space is calculated to obtain the probability of the corresponding single grid;

[0108] The probabilities of all single grids are arranged from left to right and from top to bottom into a row vector, and the row vector corresponds to a lesion position label. If the corresponding position label is obtained, the specific location of the lesion is obtained.

[0109] Furthermore, if Figure 3 As shown, the single grid is converted into a W*H matrix mode, the trace of the matrix tr = e 11 +e 22 +...+e WH The trace of tr is the sum of the diagonal elements of the matrix. The data of a single grid can be converted into a matrix. Tr is the sum of the diagonal elements of the matrix. The sum of the diagonal elements here is the probability of a single grid. The probability obtained here corresponds to a positioning position in the position encoding. This positioning position and the label are then input into the loss function for loss calculation.

[0110] Specifically, H is a mathematical symbol representing the conjugate transpose of a matrix. WH Represents the value of the corresponding position in the matrix, such as e here WH is the value of row W and column H. The specific value needs to be calculated based on the pixels in a single grid.

[0111] Furthermore, the continuously updating the parameters of the unitary transformation to repeatedly locate the lesion position label to obtain the most accurate positioning optimization model includes:

[0112] After obtaining the probability of a single grid, the gradient descent algorithm is used to update the parameters T required by the square error loss function. α , repeat the lesion location process and continuously update the parameter T α Specifically, the parameter T is updated according to the calculated squared error loss function α .

[0113] By using the principle of decreasing square error loss function, we can get the most accurate optimization model. α Input it into the gradient optimizer of the squared error loss function, and after repeated iterations, a high accuracy positioning accuracy will be obtained. Finally, input our medical image test set data to test the quality of the model. Therefore, as long as we input a medical image, the model can calculate the specific location of the lesion very well.

[0114] Example 3

[0115] This embodiment has the same inventive concept as that of Embodiment 1. On the basis of Embodiment 1, a storage medium is provided, on which computer instructions are stored. When the computer instructions are executed, the steps of a method for locating the position of a lesion in a medical image based on a single photon quantum are executed in Embodiment 1.

[0116] Based on this understanding, the technical solution of this embodiment, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0117] Example 4

[0118] This embodiment also provides a terminal, which has the same inventive concept as that of embodiment 1, and includes a memory and a processor, wherein the memory stores computer instructions that can be run on the processor, and when the processor runs the computer instructions, the steps of a method for locating a lesion position in a medical image based on a single photon quantum are executed in embodiment 1. The processor may be a single-core or multi-core central processing unit or a specific integrated circuit, or one or more integrated circuits configured to implement the present invention.

[0119] Each functional unit in the embodiment provided by the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0120] The above specific implementation methods are detailed descriptions of the present invention. It cannot be determined that the specific implementation methods of the present invention are limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions and substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A method for locating the position of a lesion in a medical image based on single-photon quantum, characterized in that: The method comprises: Acquire a medical image with a known lesion location, and mark the lesion location to obtain a lesion location label; Using a single photon to illuminate the medical image, establishing a corresponding relationship between each pixel point in the medical image and the quantum state of the photon, and obtaining the quantum state of the medical image; and encoding the lesion location label; Performing a unitary transformation on the quantum state of the medical image and converting it into a single photon row vector carrying information of all pixel points; The unitary transformation of the quantum state of the medical image and converting it into a single photon row vector carrying all pixel point information includes: Quantum states of medical images Function A U α , so that the original quantum state becomes At the same time, the U α Needs to be satisfied The unitary constraint is implemented by regularizing the loss function; The U α It can be obtained by the following formula: Among them, jS α By adjusting the weights T of a random matrix T during training α To achieve, that is: The weight T α Input as a parameter into the squared error loss function; The quantum state after the unitary transformation Convert to a single grid form, where W and H are the height and width of a single grid respectively, N is the number of grids in each row or column, Ceils is the total number of grids, X is the Xth grid, and X∈[0, Ceils-1], Pixels represents the pixel points that the Xth grid should contain, and Pixels is a row vector, the size of the row vector is W*H; The pixels of the Xth grid are calculated according to the following formula: Pixels = [i × N + j] i∈[X / / N×H,X / / N×H+H] j∈[X%N×W,X%N×W+W], Among them, i and j represent random values ​​according to pixel requirements; The obtained single grid of the medical image is converted into a W*H complex vector space, and the trace of each complex vector space is calculated to obtain the probability of the corresponding single grid; Arrange the probabilities of all single grids from left to right and from top to bottom into a row vector, where the row vector corresponds to a lesion location label; Establishing a positioning model for the lesion position, and inputting the single photon row vector and the encoded lesion position label into the positioning model to locate the lesion position label; Continuously updating the parameters of the unitary transformation to repeatedly locate the lesion position label to obtain an optimization model with the most accurate positioning; The medical image to be tested is input into the optimization model to locate the lesion.

2. The method for locating the position of a lesion in a medical image based on single photon quantum according to claim 1, characterized in that: A single medical image is composed of pixels with grayscale values ​​of 0-255 arranged in a matrix format with a size of 512*512.

3. The method for locating the position of a lesion in a medical image based on single photon quantum according to claim 1, characterized in that: The method of irradiating the medical image with a single photon comprises: Single photons are used to illuminate the medical image in a highly collimated manner, with the quantum state of the photons being in the same phase when they arrive at each individual pixel.

4. The method for locating the position of a lesion in a medical image based on single photon quantum according to claim 1, characterized in that: The step of establishing a correspondence between each pixel point in the medical image and the quantum state of a photon to obtain the quantum state of the medical image includes: After photon irradiation, the Hilbert space is selected as a basis, and the correlation space of the quantum state of the medical image can be described by the elements of the M*N complex vector space, as follows: Among them, the basis function Corresponding to the photon passing through the pixel (m, n), the basis function The coefficient a mn is a non-negative real number; Coefficient a mn It is proportional to the square root of the brightness of the medical image, that is: And the coefficient a mn Perform normalization:

5. The method for locating the position of a lesion in a medical image based on single photon quantum according to claim 1, characterized in that: Convert the single grid into a W*H matrix pattern, where the trace of the matrix is ​​tr=e 11 +e 22 +…+e WH .

6. The method for locating the position of a lesion in a medical image based on single photon quantum according to claim 1, characterized in that: The method of continuously updating the parameters of the unitary transformation to repeatedly locate the lesion position label to obtain the most accurate optimization model includes: After obtaining the probability of a single grid, the gradient descent algorithm is used to update the parameters T required by the square error loss function. α ; By utilizing the principle of continuously decreasing square error loss function, we can obtain the most accurate optimization model for positioning.

7. The method for locating the position of a lesion in a medical image based on single photon quantum according to claim 6, characterized in that: Repeat the lesion location process and continuously update the parameter T α .

Images