CT Lilun yu yingyong yanjiu (Jun 2022)

A Preliminary Study on the Value of Texture Analysis in Predicting Bleeding Complications of CT-guided Percutaneous Lung Biopsy

  • Jingci XU,
  • Zilai PAN,
  • Kemin CHEN,
  • Xiaoyong CHEN,
  • Xiao LIU

DOI
https://doi.org/10.15953/j.ctta.2021.016
Journal volume & issue
Vol. 31, no. 3
pp. 383 – 391

Abstract

Read online

Objective: In this paper we intends to explore the value of texture analysis based on CT image in predicting the complications of percutaneous pulmonary puncture hemorrhage under the guidance of CT. Methods: The preoperative plain scan images of 130 patients who underwent CT-guided percutaneous lung biopsy were analyzed retrospectively. The patients with pulmonary hemorrhage greater than or equal to grade 2 in the operative area were assined into the bleeding group while the patients with grade 0 or grade 1 pulmonary hemorrhage were assined into the no / small bleeding group. 100 cases were randomly selected as the training group, and the lung field around the pre-puncture path on the plain scan lung window image was manually drawn as the region of interest (ROI) by using MaZda software. The most valuable texture features were selected by methods of Fisher coefficient, classification error probability joint average correlation coefficient (POE+ACC) and interactive information (MI) to distinguish between bleeding group and no / small amount of bleeding group. Then, the four following feature classification statistical methods; raw data analysis (RDA), principal component analysis (PCA), linear classification analysis (LDA) and nonlinear classification analysis (NDA), were used for judgement, and the results were shown by way of error rate. Finally, the other 30 cases were verified according to the optimal texture parameters and feature classification method. Results: The lowest error rate was 11.00% (11/100) when the lung field around the puncture path was used as ROI. The error rates were respectively 13.33% (4/30) and 16.67% (5/30), when the feature selection method was POE+ACC or MI, and the feature classification statistical method was NDA, there was no significant difference between the two groups. Conclusion: The analysis of the texture characteristics of the lung field around the puncture path is helpful in predicting the risk of pulmonary puncture complicated with hemorrhage under the guidance of CT, and can provide certain basis for selecting a suitable puncture path to reduce the complications of pulmonary hemorrhage.

Keywords