CN-122025091-A - Method and system for quantifying distribution of chronic obstructive pulmonary emphysema based on multi-scale segmentation

CN122025091ACN 122025091 ACN122025091 ACN 122025091ACN-122025091-A

Abstract

The invention relates to the technical field of medical image processing, and discloses a slow pulmonary emphysema distribution quantification method and a slow pulmonary emphysema distribution quantification system based on multi-scale segmentation, wherein the method comprises the steps of performing anisotropic diffusion filtering noise reduction on a chest CT image; the method comprises the steps of segmenting lung fields and eliminating vascular bronchus structures by adopting a region growing algorithm, constructing a multi-scale image representation based on Gaussian pyramid, identifying emphysema candidate regions at a coarse scale layer, accurately outlining a boundary at a fine scale layer by adopting a self-adaptive threshold value, extracting local texture features to distinguish lobular center type emphysema from full lobular type emphysema, dividing severity level according to space aggregation characteristics and density gradient distribution, calculating emphysema volume ratio and distribution heterogeneity index, and visualizing and generating a structured quantitative report by three-dimensional pseudo-color volume rendering.

Inventors

LI ZONGYU
ZHU YAN
ZHANG TIANTIAN
FAN MENGDI

Assignees

树兰(杭州)医院有限公司

Dates

Publication Date: 20260512
Application Date: 20260128

Claims (10)

1. The slow pulmonary emphysema distribution quantification method based on multi-scale segmentation is characterized by comprising the following steps of: Acquiring a chest CT image sequence of a patient with slow lung resistance, performing anisotropic diffusion filtering treatment on the chest CT image sequence, and keeping the edge sharpness of a lung parenchyma region while inhibiting image noise to obtain a CT image after noise reduction; A lung field segmentation step, namely completing the whole segmentation of the lung field by adopting an area growth algorithm on the CT image after noise reduction, and removing a large vessel structure and a bronchus structure in the lung field by morphological treatment to obtain a lung field mask image; Constructing a multi-scale image representation of the lung field mask image based on a Gaussian pyramid, screening and identifying a large-scale low-density region through a density threshold value in a coarse-scale layer to determine an emphysema candidate region, and accurately outlining the boundary of the emphysema region by adopting an adaptive threshold value segmentation method in a fine-scale layer, wherein the adaptive threshold value is dynamically determined according to the local density statistical characteristics of a current analysis region; A step of classifying the emphysema subtype, which is to extract local texture features of an emphysema area from the fine scale layer and distinguish lobular center emphysema from whole lobular emphysema according to the spatial distribution mode of the local texture features; the severity grading and quantitative analysis step, namely automatically dividing the emphysema severity grade according to the spatial aggregation characteristic and the density gradient distribution characteristic of the low attenuation region, and calculating the emphysema volume ratio of each lung lobe region and the distribution heterogeneity index for representing the spatial distribution heterogeneity degree of emphysema; And a step of visualization and report generation, in which the spatial distribution of the emphysema is visually presented in a three-dimensional pseudo-color volume rendering form, and a structured quantitative report is generated, wherein the structured quantitative report comprises emphysema indexes of whole lung and each lung lobe, emphysema subtype composition ratio and longitudinal comparison data with a baseline examination.
2. The method according to claim 1, wherein the anisotropic diffusion filtering process uses a Perona-Malik diffusion model, a diffusion coefficient control parameter is set in a range of 20HU to 50HU, the number of filtering iterations is set in a range of 5 times to 15 times, and a filtering scale parameter is set in a range of 0.5mm to 2.0 mm.
3. The method according to claim 1, wherein the number of layers of the gaussian pyramid is set in a range of 3 layers to 5 layers, the downsampling factor between adjacent layers is 2, and the standard deviation of the gaussian filter kernel is set in a range of 0.8 to 1.5.
4. The method according to claim 1, wherein the density threshold of the coarse-scale layer is set in the range of-960 HU to-940 HU, and the adaptive threshold of the fine-scale layer is calculated from the mean value of the voxel density values in the local area and the standard differential state.
5. The method of claim 1, wherein the local texture features include contrast features, correlation features, energy features, and homogeneity features calculated based on a gray scale co-occurrence matrix, and short and long run dominance features calculated based on a gray scale run matrix.
6. The method of claim 1, wherein the method of distinguishing lobular emphysema from whole lobular emphysema comprises extracting morphological centrality features and density attenuation gradient features of the emphysema region, determining lobular emphysema when the centrality feature value is greater than a preset centrality threshold and the density attenuation gradient exhibits a tendency to gradually increase from center to outside, and otherwise determining whole lobular emphysema.
7. The method according to claim 1, wherein the method for calculating the distribution heterogeneity index comprises dividing the lung fields into a preset number of equal volume subareas, calculating the emphysema volume ratio of each subarea, and calculating the distribution heterogeneity index according to the variation coefficient of the emphysema volume ratio of each subarea.
8. The method of claim 1, wherein the severity level classification criteria includes a normal or light level when the whole emphysema volume fraction is less than 5%, a light level when the whole emphysema volume fraction is in the range of 5% to 15%, a medium level when the whole emphysema volume fraction is in the range of 15% to 30%, and a heavy level when the whole emphysema volume fraction is greater than 30%.
9. The method of claim 1, wherein the method of generating longitudinal contrast data comprises three-dimensionally registering the emphysema region of the current examination with the baseline examination, and calculating the volume change of emphysema, the extent of expansion of the emphysema, and the severity level change of the corresponding location after registration.
10. A slow pulmonary emphysema distribution quantification system based on multi-scale segmentation for implementing the method of any of claims 1-9, comprising: The image preprocessing module is used for acquiring a chest CT image sequence of a patient with slow lung resistance, and performing anisotropic diffusion filtering processing on the chest CT image sequence to obtain a CT image after noise reduction; The lung field segmentation module is used for completing the whole segmentation of the lung field and removing the large vessel structure and the bronchus structure by adopting an area growth algorithm to the CT image after noise reduction to obtain a lung field mask image; The multi-scale emphysema detection module is used for constructing multi-scale image representation based on Gaussian pyramid, identifying emphysema candidate areas at a coarse-scale layer, and accurately outlining the boundary of an emphysema area by adopting self-adaptive threshold segmentation at a fine-scale layer; The emphysema subtype classification module is used for extracting local texture characteristics of an emphysema area and distinguishing lobular center type emphysema from full lobular type emphysema according to the local texture characteristics; The severity grading and quantitative analysis module is used for grading severity grades according to the spatial aggregation characteristics and the density gradient distribution, and calculating the emphysema volume ratio and the distribution heterogeneity index of each lung lobe area; And the visualization and report generation module is used for visually presenting the emphysema spatial distribution in a three-dimensional pseudo-color volume rendering mode and generating a structured quantitative report containing emphysema indexes, subtype composition ratios and longitudinal comparison data.

Description

Method and system for quantifying distribution of chronic obstructive pulmonary emphysema based on multi-scale segmentation Technical Field The invention relates to the technical field of medical image processing and computer-aided diagnosis, in particular to a slow pulmonary emphysema distribution quantification method and system based on multi-scale segmentation. Background Chronic obstructive pulmonary disease, a respiratory disease that severely threatens human health, is characterized by emphysema, which is manifested by destruction of the alveolar walls and loss of elasticity of lung tissue. The accurate assessment of the distribution range, severity and subtype composition of emphysema has important clinical value for early diagnosis of chronic obstructive pulmonary disease, disease monitoring and treatment scheme formulation. With the development of computed tomography technology, quantitative analysis of emphysema based on CT images has become an important tool in clinical research and practice. Emphysema is typically represented in CT images as low density regions of the lung field scattered throughout, which represent abnormal air cavities formed after the alveolar walls are destroyed. From a pathological standpoint, emphysema can be divided into a variety of subtypes, leaflet center, full leaflet, paraseptal, and parascar, with leaflet center and full leaflet being the most common in the clinic and being closely related to the etiology and prognosis of the patient. In the prior art, china invention CN109035283A discloses a emphysema accurate detection and quantitative analysis method based on randomly selected partitions. The method comprises the steps of firstly completing CT image sequence input and image standardization pretreatment by a data input module, then automatically dividing a CT image by a lung parenchyma and trachea extraction module to extract lung parenchyma and extracting a trachea by a wavefront detection algorithm, and further extracting a lung tissue region of interest. The method divides the lung into a regional structure, wherein the right lung is divided into three regions and the left lung is divided into two regions, and then the lung parenchyma volume is randomly extracted on a CT image through a random region selection module so as to calculate the emphysema degree of the partial lung parenchyma. In a focus extraction link, a slow-resistance lung disease focus extraction module of the method calculates a CT value distribution model of focuses and healthy lung tissues according to the volume of each lung parenchyma, extracts focus areas by utilizing a condensation hierarchical clustering algorithm, and finally calculates focus characteristic indexes according to the extracted focus areas. The method also comprises the steps of displaying CT threshold values of all areas on the image through an output and display module, and performing Mongolian display on the classification result to display the emphysema area. However, the prior art has the following defects that firstly, the method adopts a fixed CT threshold value to carry out emphysema identification, and a-950 HU or a-910 HU is generally used as a classification threshold value, and the fixed threshold value method is difficult to adapt to the variation of CT value distribution under different patients, different scanning devices and different scanning parameters, so that the segmentation accuracy is limited in some cases. Secondly, the method cannot fully utilize multi-scale image information, and is difficult to accurately outline an emphysema region with blurred boundaries, particularly for a focus region with early emphysema lesions or gradually changed boundaries, and the accuracy of a segmentation result is required to be improved. Thirdly, the method lacks automatic distinguishing capability for the emphysema subtype, and can not effectively identify lobular center type emphysema and full lobular type emphysema, and the two subtypes have obvious differences in the aspects of etiology mechanism, disease progress and clinical prognosis, and accurate subtype classification has important significance for formulating a targeted treatment strategy. Fourth, the quantitative analysis index of the method is relatively single, is mainly concentrated on basic parameters such as emphysema percentage and CT threshold value, lacks deep quantitative analysis on the spatial distribution characteristic of emphysema, and is difficult to comprehensively reflect the spatial heterogeneity characteristic of lesions. Disclosure of Invention Aiming at the technical problems in the prior art, the invention provides a slow pulmonary emphysema distribution quantification method and a slow pulmonary emphysema distribution quantification system based on multi-scale segmentation. The invention provides a multi-scale segmentation-based slow pulmonary emphysema distribution quantification method, which comprises the following steps of imag