CN-121414757-B - Cardiovascular disease detection method based on image analysis

CN121414757BCN 121414757 BCN121414757 BCN 121414757BCN-121414757-B

Abstract

The invention relates to the technical field of medical image analysis and discloses a cardiovascular disease detection method based on image analysis. The method acquires a medical image sequence of a target patient, extracts a blood vessel region outline and generates an initial blood vessel topological graph. And registering the initial topological graph with a standard cardiovascular model, calculating the curvature deviation degree and the pipe diameter variation coefficient of each blood vessel branch, and generating a blood vessel elastic characteristic matrix by combining abnormal displacement nodes. And carrying out multi-scale fusion on the curvature deviation degree, the pipe diameter variation coefficient and the elastic characteristic matrix, outputting a vascular structure abnormality index, and generating a hemodynamic parameter set. Dividing a risk area according to gradient distribution of the parameter set on the three-dimensional model, analyzing a texture symbiotic matrix of a pixel cluster in the high risk area, and extracting texture fingerprints of calcified plaque and lipid deposition. The type and severity of cardiovascular disease is determined from its peak profile. The invention realizes accurate and objective detection and risk assessment of cardiovascular diseases.

Inventors

HU XIN
ZHAO ZHILI
SONG CHAOQUN
WANG WEIJUN
LI YU

Assignees

中国人民解放军总医院第一医学中心

Dates

Publication Date: 20260512
Application Date: 20251229

Claims (9)

1. A cardiovascular disease detection method based on image analysis, comprising: Acquiring a medical image sequence of a target patient, extracting a blood vessel region outline in each frame of image and generating an initial blood vessel topological graph; constructing a vascular motion track according to the displacement vector and gray level difference of vascular regions in adjacent frame images, and marking abnormal displacement nodes; Carrying out spatial registration on the initial vessel topological graph and a standard cardiovascular model, and calculating the curvature deviation degree and the pipe diameter variation coefficient of each vessel branch; Generating a vascular elastic characteristic matrix based on the distribution density of the abnormal displacement nodes and the mechanical parameters of vascular branches; Carrying out multi-scale fusion on the curvature deviation degree, the pipe diameter variation coefficient and the vascular elastic characteristic matrix, and outputting a vascular structure abnormality index; Extracting time domain fluctuation characteristics of blood flow signals in the image sequence, and generating a hemodynamic parameter set by combining vascular structure abnormality indexes; Dividing a high-risk lesion area and a low-risk normal area according to gradient distribution of a hemodynamic parameter set on a three-dimensional blood vessel model; Performing texture co-occurrence matrix analysis on pixel clusters in a high-risk lesion area, and extracting texture fingerprints of calcified plaques and lipid deposition; Constructing a lesion probability prediction matrix based on the coupling relation between the texture fingerprint and the vascular structure abnormality index; Determining the type and severity level of cardiovascular diseases according to the peak distribution of the lesion probability prediction matrix; The extracting texture fingerprints of calcified plaque and lipid deposits comprises: extracting a plurality of groups of pixel gray sequences along the normal direction of the blood vessel wall in the high-risk lesion area; performing gabor filter group processing on each group of pixel gray sequences, and extracting texture response spectrums of different frequency bands; and generating texture fingerprint codes of calcified plaque and lipid deposition after the dimension reduction by principal component analysis.
2. The method for detecting cardiovascular disease based on image analysis according to claim 1, wherein the extracting the contour of the blood vessel region in each frame of image and generating the initial blood vessel topological graph comprises: extracting a vascular region binarization mask in each frame of image by adopting a self-adaptive threshold segmentation algorithm; Performing morphological corrosion operation on the vascular region binarization mask of the adjacent frames, and generating a connected domain marker graph after eliminating artifact interference; And extracting the space coordinates of all the vessel center lines in the connected domain marker graph, and constructing an initial vessel topological graph through cubic spline interpolation.
3. The method for detecting cardiovascular diseases based on image analysis according to claim 2, wherein the constructing a vascular motion trajectory according to the displacement vector and the gray scale difference of the vascular region in the adjacent frame images comprises: calculating Euclidean distance and direction included angle of the central line of the matched blood vessel in the images of the adjacent frames; When the Euclidean distance exceeds the preset multiple of the diameter of the blood vessel and the direction included angle is larger than a critical threshold value, marking the node as an abnormal displacement node; and performing space-time clustering on all abnormal displacement nodes to generate a fracture region map of the vascular motion trail.
4. The image analysis-based cardiovascular disease detection method according to claim 3, wherein spatially registering the initial vessel topology map with the standard cardiovascular model comprises: Performing feature point matching on a main branch of the initial vessel topological graph and an anatomical structure of a standard cardiovascular model; correcting the spatial deformation of the vascular branches through a thin plate spline transformation algorithm, and outputting a registered vascular curvature field; the curvature deviation of each branch central axis from the standard model is calculated in the registered vascular curvature field.
5. The method for detecting cardiovascular disease based on image analysis according to claim 4, wherein generating a vascular elastic feature matrix based on the distribution density of abnormal displacement nodes and the mechanical parameters of vascular branches comprises: measuring the periodic contraction amplitude and the phase delay of the blood vessel wall around the abnormal displacement node; Deriving a local Young's modulus according to a linear relation between the contraction amplitude and the intravascular pressure; And integrating Young modulus data of all abnormal displacement nodes to generate a vascular elastic characteristic matrix.
6. The method for detecting cardiovascular disease based on image analysis according to claim 5, wherein the multi-scale fusion of curvature deviation, coefficient of pipe diameter variation and vascular elasticity feature matrix comprises: carrying out wavelet decomposition on the vascular structure abnormality index to obtain a high-frequency detail component and a low-frequency contour component; Carrying out convolution operation on the high-frequency detail components and the vascular elastic feature matrix to enhance the local lesion feature response; The weight distribution of the low frequency contour components and the hemodynamic parameter set is optimized by a back propagation algorithm.
7. The image analysis-based cardiovascular disease detection method according to claim 6, wherein dividing the risk area comprises: calculating Laplacian of the hemodynamic parameter set at each vertex of the three-dimensional blood vessel model surface; when the Laplacian operator exceeds a dynamic threshold value and the vascular structure abnormality index is continuously increased, judging that the vascular structure abnormality index is a high-risk lesion area; the region growing algorithm is expanded on the high-risk lesion region until the fluid shear force mutation point at the bifurcation of the blood vessel is encountered.
8. The method for detecting cardiovascular disease based on image analysis according to claim 7, wherein the constructing a lesion probability prediction matrix comprises: Establishing a joint probability density function of texture fingerprint codes and vascular structure abnormality indexes; simulating spatial distribution modes of different lesion types by adopting a Monte Carlo sampling method; And generating a lesion probability prediction matrix according to the aggregation degree of various lesions in the sampling result.
9. The image analysis-based cardiovascular disease detection method according to claim 8, wherein the determining the type and severity level of cardiovascular disease from the peak distribution of the lesion probability prediction matrix comprises: Identifying the space geometric features of the continuous peak areas in the lesion probability prediction matrix; when the peak area presents annular distribution and the texture fingerprint code accords with calcification characteristics, judging atherosclerosis; and when the peak area is in linear distribution and the numerical value of the vascular elastic characteristic matrix is abnormal, judging that the vascular fibrosis lesion is formed.

Description

Cardiovascular disease detection method based on image analysis Technical Field The invention relates to the technical field of medical image analysis, in particular to a cardiovascular disease detection method based on image analysis. Background The high incidence of cardiovascular disease has led to the need for long-term medical treatment and care for a large number of patients, which not only results in a serious decline in the personal quality of life of the patients, but also places great pressure on the home both economically and manually. Patients may need frequent medical visits, long-term medication, and even surgical treatments, which all place a heavy economic burden on the home. In addition, the family members also need to spend a great deal of time and effort to care for the patient, affecting the normal life and work of the family. The widespread prevalence of cardiovascular disease is, from a social perspective, a tremendous drain on medical resources. To meet the therapeutic needs of patients with cardiovascular disease, hospitals require a large amount of medical equipment, medicines, and specialized healthcare workers. In some large hospitals, cardiovascular medicine and cardiovascular surgery are often fully ill, the hospital bed is tense, and the workload of medical staff is heavy. The excessive concentration of medical resources in the cardiovascular disease treatment field also affects the medical service supply of other diseases, and causes imbalance of medical resource allocation. Meanwhile, a large number of patients lose the working capacity due to cardiovascular diseases, which also has negative influence on the social and economic development, reduces labor force supply, reduces production efficiency and prevents the continuous growth of economy. In face of serious challenges of cardiovascular diseases, timely and accurate detection is important for effective treatment, prevention and control of the diseases. However, the existing conventional detection methods have a plurality of limitations, and are difficult to meet the requirements of clinical practice. Electrocardiography is a common method for detecting cardiovascular diseases, and the principle is to record the change of the heart electrical activity by placing electrodes on the body surface. When the heart cells are excited, potential changes can be generated, the potential changes can be recorded through the electrodes, and the waveform of the electrocardiogram reflects the potential changes of the myocardial cells at different parts of the heart. In diagnosing arrhythmias and conduction blocks, the electrocardiogram has positive value, and characteristic electrocardiogram changes and evolutions are also reliable and practical methods for diagnosing myocardial infarction. However, there are significant drawbacks to electrocardiography. It has limited ability in detecting fine lesions and early diseases, and for hidden coronary heart disease, the electrocardiogram may not capture obvious abnormal changes due to the light degree of myocardial ischemia, which is easy to cause missed diagnosis. For microvascular dysfunction, it is also difficult for an electrocardiogram to provide accurate diagnostic information, as microvascular lesions typically do not cause significant changes in the macroscopic electrical activity of the heart. The heart color ultrasound, also called an echocardiogram, utilizes the physical characteristics of ultrasonic waves, transmits high-frequency sound waves through a probe, reflects the sound waves when the sound waves are transmitted in human tissues and meet different tissue interfaces, receives the reflected sound waves by the probe and converts the sound waves into electric signals, and then processes the electric signals through a computer to form images of the heart, so that the structure, the function and the blood flow state of the heart can be displayed in real time, an important basis is provided for clinical diagnosis and treatment, and the device can be used for detecting abnormal heart structure, evaluating heart functions and blood flow dynamics, diagnosing congenital heart diseases and monitoring the disease change of patients with heart diseases in a follow-up mode. However, it also has limitations, and is insufficient in judging the change of cardiovascular over time, so that it is difficult to comprehensively and accurately monitor the dynamic change of cardiovascular. For some functional lesions, such as myocardial function changes caused by early myocardial ischemia, cardiac color ultrasound may not be found in time, because the structural morphology of the heart may not be changed obviously, and it is difficult to accurately judge fine abnormalities of myocardial function only by using an ultrasonic image. In addition, the heart color Doppler ultrasound diagnosis result is greatly influenced by experience and skill of doctors, subjectivity is high, different doctors can possib