CN-115862832-B - Optimal body position combination optimizing method and system for capsule endoscope

CN115862832BCN 115862832 BCN115862832 BCN 115862832BCN-115862832-B

Abstract

The invention discloses a capsule endoscope optimal body position combination optimizing method and a capsule endoscope optimal body position combination optimizing system, wherein the method is used for realizing track sequence probability modeling through original problem serialization, track sequence probability modeling, state transition probability modeling, body position combination evaluation modeling and body position combination optimizing through a state transition probability modeling mode, providing a value evaluation method for different body position combinations, realizing optimal body position combination with more universality under a quick searching probability space, particularly providing two state transition probability modeling methods, namely a statistical modeling method based on real capsule endoscope data and a probability quantification modeling method based on doctor experience space, on one hand, the method can greatly reduce dependence on doctor experience of a professional digestive tract inner diameter in a capsule endoscope inspection process, and on the other hand, the universality of a body position path optimized by the method also lays a certain foundation for realizing future capsule endoscope intelligent diagnosis and treatment modes.

Inventors

LIU HUAN
LIU SIDE
SUN JIAN
ZHANG JIE
LUO XIAOBEI
YANG HERAN
Gu Chuncai
WANG YUSI
WANG YUNZHONG

Assignees

西安交通大学
广州思德医疗科技有限公司

Dates

Publication Date: 20260508
Application Date: 20221228

Claims (9)

1. The method for optimizing the optimal body position combination of the capsule endoscope is characterized by comprising the following steps of: Serializing the original problem, namely describing the body bit combination and the anatomical structure by using serialization parameters; track sequence probability modeling for a given body bit composition Is observed in the case of an anatomical sequence of Modeling the track sequence probability into a cumulative multiplication of the state transition probability by using Markov property and probability product property; obtaining the combined sequence of the actions Next, note that the anatomical sequence was observed Probability of (2) The following are provided: Wherein: Is that The probability of a state transition in a round, For the size of the combination of selectable positions, State transition probability modeling, namely obtaining a state transition probability model through a statistical modeling method based on real data of the capsule endoscope and a probability quantification modeling method based on doctor experience space respectively I.e. the last anatomical structure in the sequence of observed anatomical structures is The body position is By changing the position to So that the observed anatomical sequence is State transition probabilities of (2); the probability quantitative modeling method based on doctor experience space comprises converting abundant experience of an endoscope doctor into specific probability, realizing modeling of state transition probability, 1) Doctor experience table is a table designed for a professional doctor, and the doctor marks according to own experience that under the current body position and the current observed anatomical structure, the doctor reaches the arrival possibility and the possibility level of the next anatomical structure through the next body position action, wherein the arrival possibility is 0 and 1, the possibility is 1, and the 1-4 correspond to different levels with respect to the possibility level; 2) Two empirical sequences, 1-2-1-2-1-2..1 means reciprocating between the anatomical structures labeled 1-2 and finally stabilizing at the anatomical structure labeled 1, 1-2-1-2-3-4-3-4..3 means reciprocating first over the anatomical structure labeled 1-2 and then transiting to the anatomical structure labeled 3-4 and finally stabilizing at the anatomical structure labeled 3, by physician experience, converting the labeling into two observable anatomical structure sequences according to the probability level 1-4; 3) The probability quantification is carried out, namely, different probability grades 1-4 are marked according to the experience of a doctor, and the probability is converted into specific probability, specifically, 1 corresponds to probability 1.0,2 corresponds to probability 0.95,3 corresponds to probability 0.5,4 corresponds to probability 0.45, and the quantification probability is dynamically optimized according to the improvement of the experience of the doctor; 4) According to the experience table marked by doctors, combining the experience sequence and the quantization probability, realizing the transition probability to any different states Modeling and calculation of (2); the body position combination evaluation, namely acquiring an anatomical structure track sequence set which can be acquired by sampling under a given body position combination, and acquiring an evaluation index capable of reflecting universality by combining the state transition probability and coverage rate of a certain anatomical structure sequence; and (3) body bit composition optimizing, namely quickly acquiring different body bit composition modes through a random sampling mode, and outputting an optimal body bit composition sequence through calculating the evaluation index.
2. The capsule endoscopic optimal body position composition optimizing method as defined in claim 1, wherein said original questions are serialized, body position composition and anatomy are described by serialization parameters by defining a selectable body position set, = { Left side position, high hip head low left side position,..degree}, defining a set of target anatomy, = { Esophagus, fundus, cardia. }, defining an anatomical sequence, , Represent the first Anatomical sequence captured by the turn defines Wherein Representing the th shot at the t-th round Anatomical structures, and Defining a sequence of body bits, ), Represent the first A body position selected in a round, wherein Defining the coverage of the anatomical structure, which is defined as the ratio of the number of observed non-repeated anatomical structures to the number of all anatomical structures, i.e = Wherein: Representation extraction Length represents the number of acquisition set elements.
3. The capsule endoscope optimal body composition optimizing method according to claim 1, wherein when the state transition probability modeling is performed, the last anatomical structure of the observed anatomical structure sequence is The body position is By changing the position to So that the observed anatomical sequence is The statistical modeling method based on the real data of the capsule endoscope obtains the state transition probability by carrying out target data truncation, track serialization and state transition frequency statistics on the data, and the specific calculation formula is as follows: Wherein, the Is the first in the data Seed state transition The frequency of the occurrence of this, Is the frequency with which all state transitions in the data occur.
4. The capsule endoscope optimal body composition optimizing method according to claim 1, wherein the body composition evaluation obtains an evaluation index capable of reflecting universality, in particular, a given body composition is obtained by sampling Down-sampable acquisition of a set of anatomical track sequences In combination with a sequence of anatomical structures Probability of (2) And coverage rate The specific formula is as follows: Wherein the method comprises the steps of For the size of the set of all track sequences, The normalization is indicated by the fact that, The state transition probability can be calculated by a track sequence probability formula, The larger the group of body position combinations, the better the group, with a greater likelihood of covering as much of the anatomy as possible.
5. The capsule endoscope optimal body composition optimizing method according to claim 1, wherein the body composition optimizing comprises the following specific steps: S1, modeling state transition probability, namely combining a statistical modeling method based on real data of a capsule endoscope with a probability quantification method based on a doctor experience space according to a certain weight to obtain state transition probability, wherein the weight is adjusted according to the magnitude condition of the real data; S2, randomly sampling according to the state transition probability, and storing the generated random track data into a random sampling track library; s3, calculating the coverage rate of each track in the random sampling track library, and storing the body bit combination corresponding to the random sampling track with the coverage rate reaching 100% into a duplicate-removed body bit combination library to be selected; s4, for any body bit combination in the candidate body bit combination library All non-repeated track sequences are obtained by traversing state transitions with state transition probability not 0, and the probability value of each track sequence is calculated by a track sequence probability modeling method Storing the position-all track libraries into the position-all track libraries; S5, by calculating any body bit combination Is (are) estimated value of (a) And according to the evaluation value The different body bit combinations are ordered in a descending order, and the ordering result is stored in a body bit combination evaluation library; s6, selecting the top-ranked optimal body bit combination from the body bit combination evaluation library to be provided for an endoscopist, and providing the optimized body bit combination path with universality for the endoscopist.
6. The capsule endoscope optimal body composition optimizing method of claim 5, wherein the entire process is processed in a parallel or distributed framework.
7. The capsule endoscope optimal body position combination optimizing system based on state transition probability modeling is characterized by comprising an original problem serialization module, a track sequence probability modeling module, a state transition probability modeling module, a body position combination evaluation module and a body position combination optimizing module; The original problem serialization module describes the positioning body bit combination and the anatomical structure by using serialization parameters; trajectory sequence probability modeling module gives body bit combinations Is observed in the case of an anatomical sequence of Modeling the track sequence probability into a cumulative multiplication of the state transition probability by using Markov property and probability product property; the state transition probability modeling module obtains a state transition probability model through a statistical modeling method based on real data of the capsule endoscope and a probability quantification modeling method based on a doctor experience space respectively; the body position combination evaluation module acquires an anatomical structure track sequence set which can be acquired by sampling under a given body position combination, and acquires an evaluation index capable of reflecting universality by combining the state transition probability and coverage rate of a certain anatomical structure sequence; the body bit composition optimizing module rapidly acquires different body bit composition modes through a random sampling mode, and outputs an optimal body bit composition sequence through calculating the evaluation index.
8. A computer device comprising a processor and a memory, the memory storing a computer executable program, the processor reading the computer executable program from the memory and executing the computer executable program, the processor executing the computer executable program to implement the method for optimizing the body position of the capsule endoscope according to any one of claims 1 to 6.
9. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the method for optimizing the combination of optimal body bits of the capsule endoscope according to any one of claims 1 to 6 can be realized.

Description

Optimal body position combination optimizing method and system for capsule endoscope Technical Field The invention relates to the technical field of capsule endoscopy, in particular to a capsule endoscope optimal body position combination optimizing method and a system. Background At present, gastrointestinal endoscopy is the best means for early diagnosis and prevention and control of digestive tract tumors, but a digestive endoscopist is seriously insufficient, and the traditional gastrointestinal endoscopy is painful, high in risk and poor in compliance, and cannot meet the civil demands of prevention and control of digestive tract tumors. The capsule endoscope is an innovative medical technology capable of comfortably completing the whole examination of the digestive tract, but the traditional gastrointestinal endoscopy with high risk cannot be replaced at present, and the capsule endoscope is characterized in that each examination of the capsule endoscope generates a huge amount of pictures of more than 5 ten thousand sheets, so that the workload of a reader is greatly increased, the capsule endoscope is difficult to provide enough coverage and accurate position information for certain parts of the digestive tract due to the control problem of the capsule endoscope, thus the missed examination of lesions and the misjudgment of anatomical parts are easy to occur, and particularly, the capsule endoscope without a magnetic control module can avoid the discomfort brought to a tested person by a magnetic control mode, but also limits the control mode of the capsule in the digestive tract. Generally, such capsules often require the subject to adjust for appropriate changes in the motion of the capsule in the alimentary canal by making corresponding changes in body position under the direction of a specialist doctor and automatically record the anatomy taken during the procedure. Especially in the stomach structure of space relatively complicacy, to realize the more comprehensive of structure of shooing, often the whole process all needs the digestion endoscopist that has abundant professional experience, and the real-time picture according to the transmission of taking by the capsule is judged the approximate position and the orientation of capsule to instruct the person of being tested to make the targeted position change, guide the motion of capsule endoscope in whole stomach space. In a future-oriented large-volume capsule endoscopy intelligent diagnosis and treatment model, the cost of digestive endoscopist and time required for this procedure is very large and unsatisfied. Considering that the human stomach has a certain physical structure, how to summarize the existing examination data and summarize the expert experience of the professional doctor, a group of body position combination paths with universality are found, and the picture shot by the capsule can fully cover the anatomical structure in the alimentary canal as much as possible under the guidance of the group of body positions, so that the method has very important practical value. Disclosure of Invention The invention aims to provide an optimal body position combination optimizing method of a capsule endoscope, which realizes track sequence probability modeling in a state transition probability modeling mode, provides a value evaluation method for different body position combinations, and realizes fast searching of an optimal body position combination with more universality under a probability space. The invention aims at realizing the optimal body position combination optimizing method of the capsule endoscope, which comprises the following steps: Serializing the original problem, namely describing the body bit combination and the anatomical structure by using serialization parameters; modeling the track sequence probability, namely observing the probability of the anatomical structure sequence Q N under the condition of a given body bit composition S N, and modeling the track sequence probability into the cumulative multiplication of the state transition probability by applying Markov property and probability product property; The state transition probability modeling is that a state transition probability model P (q next|snext,(qnow)last,snow) is obtained through a statistical modeling method based on the real data of the capsule endoscope and a probability quantification modeling method based on the experience space of a doctor respectively, namely when the last anatomical structure of the observed anatomical structure sequence is (q now)last, the body position is s now, and the state transition probability of the observed anatomical structure sequence is q next is obtained through changing the body position to be s next; the body position combination evaluation, namely acquiring an anatomical structure track sequence set which can be acquired by sampling under a given body position combination, and acquiring an evaluation index capable