A Method for Automatic Tracheal Tree Extraction from Chest CT Images

Abstract

The invention belongs to the technical field of image processing based on medical images, and in particular relates to a method for automatically extracting the tracheal tree from chest CT images. Obtain the main trachea and main bronchi from chest CT images; establish an adaptive threshold 3D region growth segmentation model and an adaptive threshold leakage model based on the 3D region growth segmentation method and the acquired main trachea and main bronchi information; use the adaptive threshold 3D region The growth segmentation model and the adaptive threshold leakage model extract the second type of tracheal branches from the chest CT image; based on the extracted intermediate information of the second type of tracheal branches, adjust the parameters of the adaptive threshold 3D regional growth model and the adaptive threshold leakage model, Then, the third type of tracheal branches of the chest CT image are extracted; terminal tracheal branches are extracted based on the acquired tracheal tree topology, and the tracheal tree of the chest CT image is obtained. The method provided by the invention improves the accuracy of tracheal segmentation for extracting tracheal trees from CT images, while reducing the extraction time.

Description

technical field [0001] The invention belongs to the technical field of image processing based on medical images, in particular to a method for automatically extracting a trachea tree from a chest CT image. Background technique [0002] Obtaining accurate lung and tracheal tree structures from CT images is of great significance in the medical and computer application circles. Clinicians can conduct pathological analysis and follow-up research on common respiratory diseases such as chronic obstructive pulmonary disease and bronchiectasis through tracheal parameters and grading information; they can also perform non-invasive virtual bronchoscopy on patients; in addition, the tracheal tree and The anatomical one-to-one correspondence of sub-level lung structures such as lung lobes and lung segments is also an important basis for image segmentation and analysis of related structures. Therefore, lung and trachea segmentation based on CT images has always been a hot spot for resea...

AI Technical Summary

Problems solved by technology

[0004] Due to the special topological structure and gray texture features of the pulmonary tracheal tree, as the tracheal tree divides step by step, the lumen becomes thinner and the tube wall becomes thinner, and the gray value of the tracheal wall gradually decreases on the CT image. , some segmentation methods routinely used for lung parenchyma, liver, and brain are not well suited for the segmentation of lung and trachea

Tracheal segmentation based on traditional threshold growth tends to spread across the tracheal wall into the lung parenchyma, forming a large-scale over-segmentation, that is, the "leakage" phenomenon, and it is difficult to effectively identify subtle or diseased branches, leaving a lot of detailed information, which is difficult to obtain accurate pulmonary tracheal tree

[0005] Among the currently disclosed tracheal tree processing methods, patent CN201210423958.2 uses multi-scale gray scale reconstruction to enhance the lumen area; on this basis, patent CN201510009239.X uses multi-scale tubular structure features to extract tracheal tree, both methods consume A lot of time; the patent CN201110405950.9 uses a fixed threshold to extract the bronchi below the segment level, and the designed leakage processing model is only for the segmental branches, it is difficult to obtain deeper bronchi and fine trachea, and there is no specific leakage processing method designed for these branches; Patent CN201510224781.7 utilizes energy function to reconstruct subtle and peripheral trachea, but still needs a fixed threshold to judge trachea membership and leakage

The above methods require strict manual parameter setting, or require complex preprocessing and enhancement operations on CT images, or need to extract tubular structural features and other auxiliary tracheal segmentation, which greatly increases the processing time and restricts the accuracy of tracheal segmentation, making it difficult to complete multiple The segmentation task of CT images under various imaging conditions seriously restricts the reliability and applicability in clinical applications

Images