CN-122023339-A - Quantitative analysis method, system and medium for ultrasonic contrast time intensity curve

CN122023339ACN 122023339 ACN122023339 ACN 122023339ACN-122023339-A

Abstract

The invention discloses a quantitative analysis method, a quantitative analysis system and a quantitative analysis medium for an ultrasonic contrast time intensity curve, and relates to the technical field of medical imaging ultrasonic contrast. The method comprises the steps of obtaining an ultrasonic contrast dynamic image sequence of a tumor focus, defining a focus region of interest (ROI) of the dynamic image sequence, drawing an ultrasonic contrast time intensity curve by taking time as an abscissa and a gray value of each frame of image in the focus region of the ROI as an ordinate, extracting characteristics of the ultrasonic contrast time intensity curve by using an encoder, reconstructing the characteristics of the ultrasonic contrast time intensity curve into a smooth time intensity curve by using a decoder, identifying key time nodes of the smooth time intensity curve by using an intelligent inflection point detection algorithm, obtaining intensity parameters of the smooth time intensity curve by using an adaptive segmentation integration technology, quantitatively analyzing the ultrasonic contrast time intensity curve based on the key time nodes and the intensity parameters of the corresponding tumor focus, and improving efficiency and accuracy of quantitative analysis of the time intensity curve.

Inventors

ZHANG BO
CHENG YONG
MA JIAOJIAO

Assignees

北京化工大学

Dates

Publication Date: 20260512
Application Date: 20260130

Claims (7)

1. A method for quantitatively analyzing an ultrasound contrast time intensity curve, the method comprising: The method comprises the steps of obtaining ultrasonic contrast dynamic image sequences of different tumor focuses, defining focus interested areas ROI for the dynamic image sequences, drawing ultrasonic contrast time intensity curves by taking time as an abscissa and gray values of each frame of image in focus ROI areas as an ordinate; Reconstructing the ultrasonic contrast time intensity curve through a pre-trained reconstruction model to obtain a smooth time intensity curve, wherein the reconstruction model is of a two-way structure of an encoder and a decoder, the encoder extracts time sequence characteristics of the ultrasonic contrast time intensity curve, and the decoder reconstructs the time sequence characteristics of the ultrasonic contrast time intensity curve into the smooth time intensity curve; The key time nodes of the smooth time intensity curve are identified through an intelligent inflection point detection algorithm, and the curve integrals of the ascending section, the peak section and the descending section of the smooth time intensity curve are respectively calculated through a self-adaptive segmentation integration technology, so that the intensity parameters of the smooth time intensity curve are obtained; And constructing a scoring model according to the tumor focus threshold value, the corresponding key time node and the corresponding intensity parameter, and quantitatively analyzing an ultrasonic contrast time intensity curve.
2. The quantitative analysis method of the ultrasonic contrast time intensity curve according to claim 1, wherein the reconstruction model is of a two-way structure of an encoder and a decoder, the encoder comprises three convolution layers containing ReLU activation functions and two self-attention layers, each convolution layer integrates a residual block, the residual block comprises two convolution layers and a batch normalization layer, the decoder comprises the convolution layers and the self-attention layers which are symmetrical to the encoder, and the outputs of all the layers of the encoder are added with the inputs of corresponding layers of the decoder through residual connection and then are connected with a full connection layer.
3. The quantitative analysis method of ultrasound contrast time intensity curve according to claim 1, wherein the reconstruction model loss function includes a main loss term, a smoothness constraint term and a physical constraint term, the main loss term is an improved mean square error loss and introduces a time weight factor, the smoothness constraint term is a constraint term based on curve derivative, and the physical constraint term is used for ensuring that the reconstructed smooth time intensity curve accords with hemodynamic characteristics.
4. The method of claim 1, wherein the key time nodes of the time intensity profile comprise peak arrival time, rise time and enhancement duration, and wherein the intensity parameters comprise peak intensity, baseline intensity and maximum enhancement intensity.
5. The method of quantitative analysis of ultrasound contrast time intensity curves according to claim 1, wherein said calculating the key time node and the intensity parameter score for a tumor lesion comprises: The random forest algorithm is used for establishing a scoring model of key time nodes and intensity parameters corresponding to tumor focus threshold values by learning key time nodes and intensity parameter data corresponding to different tumor focuses, calculating scores of the key time nodes and the intensity parameters corresponding to the tumor focuses by using the trained random forest algorithm, and quantitatively analyzing an ultrasonic contrast time intensity curve.
6. A computer system comprising at least one processor and a memory communicatively coupled to the at least one processor, wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the ultrasound contrast time intensity profile quantitative analysis method of any one of claims 1 to 5.
7. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program which, when being executed by a processor, is capable of performing the ultrasound contrast time intensity profile quantitative analysis method as defined in any one of claims 1 to 5.

Description

Quantitative analysis method, system and medium for ultrasonic contrast time intensity curve Technical Field The invention relates to the technical field of medical image ultrasonic radiography, in particular to an ultrasonic radiography time intensity curve quantitative analysis method, an ultrasonic radiography time intensity curve quantitative analysis system and a medium. Background The ultrasonic radiography technology is safe, noninvasive and high in real-time performance, and is widely applied to clinical diagnosis of thyroid cancer, breast cancer and other tumor diseases. The contrast agent containing microbubbles is injected intravenously, the blood flow echo signal is enhanced by utilizing nonlinear vibration of the microbubbles, focus blood flow perfusion condition is clearly presented, and a key basis is provided for judging pathological change properties. The traditional ultrasonic contrast analysis relies on qualitative judgment of subjective experience of doctors, and has the problems of poor diagnosis consistency, lack of standardized quantitative indexes, low efficiency, poor repeatability and the like. Time Intensity Curve (TIC) analysis is used as a mainstream quantitative method, although objective parameters such as peak reaching time and peak intensity can be extracted, the traditional mathematical model fitting complex curve has insufficient precision, is easily interfered by noise and artifacts, and parameter calculation has no unified standard and needs manual operation, so that clinical requirements are difficult to meet. The deep learning technology provides a new thought for solving the problems, but the research on the aspects of intelligent TIC reconstruction, accurate parameter extraction and standardized analysis is still in a preliminary exploration stage, and a solution of a mature system is lacking. Therefore, the development of an automatic and accurate TIC reconstruction and quantitative analysis method has important clinical value for improving the accuracy and normalization of tumor ultrasonic contrast diagnosis. Disclosure of Invention The invention aims to provide a quantitative analysis method, a quantitative analysis system and a quantitative analysis medium for an ultrasonic contrast time intensity curve, which improve the efficiency and the accuracy of quantitative analysis of the time intensity curve. In order to solve the technical problems, the embodiment of the invention provides an ultrasonic contrast time intensity curve quantitative analysis method, which comprises the following steps: The method comprises the steps of obtaining ultrasonic contrast dynamic image sequences of different tumor focuses, defining focus interested areas ROI for the dynamic image sequences, drawing ultrasonic contrast time intensity curves by taking time as an abscissa and gray values of each frame of image in focus ROI areas as an ordinate; Reconstructing the ultrasonic contrast time intensity curve through a pre-trained reconstruction model to obtain a smooth time intensity curve, wherein the reconstruction model is of a two-way structure of an encoder and a decoder, the encoder extracts time sequence characteristics of the ultrasonic contrast time intensity curve, and the decoder reconstructs the time sequence characteristics of the ultrasonic contrast time intensity curve into the smooth time intensity curve; The key time nodes of the smooth time intensity curve are identified through an intelligent inflection point detection algorithm, and the curve integrals of the ascending section, the peak section and the descending section of the smooth time intensity curve are respectively calculated through a self-adaptive segmentation integration technology, so that the intensity parameters of the smooth time intensity curve are obtained; And constructing a scoring model according to the tumor focus threshold value, the corresponding key time node and the corresponding intensity parameter, and quantitatively analyzing an ultrasonic contrast time intensity curve. In some alternative embodiments, the reconstruction model is an encoder-decoder dual-path structure, the encoder comprises three convolution layers containing ReLU activation functions and two self-attention layers, each convolution layer integrates a residual block, the residual block comprises two convolution layers and a batch normalization layer, the decoder comprises the convolution layers and the self-attention layers which are symmetrical to the encoder, and the output of each layer of the encoder is added with the input of the corresponding layer of the decoder through residual connection, and then the full connection layer is connected. In some alternative embodiments, the reconstruction model loss function includes a main loss term that is an improved mean square error loss and incorporates a time weight factor, a smoothness constraint term that is a constraint term based on curve derivatives, and a physical constra