Analysis of factors influencing hookwire dislodgement in CT-guided hookwire localization: a retrospective study using variable importance analysis with a random forest model

Abstract

Background: Video-assisted thoracoscopic surgery (VATS) is a minimally invasive and safe procedure. However, lung deflation during the operation causes anatomic landmark distortion, complicating small nodules detection. Computed tomography (CT)-guided hookwire localization promotes the success rates of VATS, but faces issues with hookwire dislodgement, potentially losing intraoperative tumor reference. This study was conducted to identify the relative importance ranking of potential factors influencing dislodgement in CT-guided hookwire localization.

Methods: This retrospective study reviewed 123 cases of CT-guided hookwire localization followed by VATS resection. Variables analyzed included sex, age, nodule size, emphysema, chest wall/muscle/total depth, distance from the nodule (DNP) or wire tip to the pleura (DWP), procedure time, nodule subtypes, multiple localization, post-procedural hemorrhage, pneumothorax, nodule penetration, and time intervals between completion of procedure to initiation of surgery (PS interval). Variables were compared using chi-square tests or Mann-Whitney tests. A random forest model, enhanced with the Synthetic Minority Over-sampling Technique (SMOTE) for oversampling, was employed to determine the relative importance of each variable. The relative importance of variables was presented using the mean decrease Gini and mean decrease accuracy metrics. For sensitivity analysis, relative variable importance was analyzed using extreme gradient boosting (XGBoost) model, and the relative importance of variables was presented using the gain metric.

Results: Among the 123 cases, dislodgement occurred in 15. In univariable analysis, only the PS interval was statistically significant (134.1 ± 73.1 vs. 104.1 ± 46.1 minutes in dislodgement or non-dislodgement, p = 0.031). The random forest and XGBoost model identified the top five important variables as the PS interval, DWP, DNP, total depth, and age. The top five factors demonstrated a distinct difference when compared to the other factors.

Conclusions: The study identified the PS interval as the most critical factor in hookwire dislodgement, along with DNP, DWP, total depth, and age. These results identified the presence of modifiable factors within the hospital and can assist practitioners and surgeons in recognizing the dislodgement risk of procedures based on various patient factors.