Development of a machine learning model for predicting pneumothorax risk in coaxial core needle biopsy (≤3 cm).

PURPOSE: The aim is to devise a machine learning algorithm exploiting preoperative clinical data to forecast the hazard of pneumothorax post-coaxial needle lung biopsy (CCNB), thereby informing clinical decision-making and enhancing perioperative care. METHOD: This retrospective analysis aggregated clinical and imaging data from patients with lung nodules (≤3 cm) biopsies. Variable selection was done using univariate analysis and LASSO regression, with the dataset subsequently divided into training (80 %) and validation (20 %) subsets. Various machine learning (ML) classifiers were employed in a consolidated approach to ascertain the paramount model, which was followed by individualized risk profiling showcased through Shapley Additive eXplanations (SHAP). RESULTS: Out of the 325 patients included in the study, 19.6% (64/325) experienced postoperative pneumothorax. High-risk factors determined were Cancer, Lesion_type, GOLD, Size, and Depth. The Gaussian Naive Bayes (GNB) classifier demonstrated superior prediction with an Area Under the Curve (AUC) of 0.82 (95% CI 0.71-0.94), complemented by an accuracy rate of 0.8, sensitivity of 0.71, specificity of 0.84, and an F1 score of 0.61 in the test cohort. CONCLUSION: The formulated prognostic algorithm exhibited commendable efficacy in preoperatively prognosticating CCNB-induced pneumothorax, harboring the potential to refine personalized risk appraisals, steer clinical judgment, and ameliorate perioperative patient stewardship.

Development and validation of a prediction model of pneumothorax after CT-guided coaxial core needle lung biopsy.

Background: Pneumothorax is the most common complication of computed tomography-guided coaxial core needle biopsy (CCNB) and may be life-threatening. We aimed to evaluate the risk factors and develop a model for predicting pneumothorax in patients undergoing computed tomography-guided CCNB, and to further determine its clinical utility. Methods: Univariate and multivariate logistic regression analyses were conducted to identify independent risk factors for pneumothorax from 18 variables. A predictive model was established using multivariable logistic regression and presented as a nomogram based on a training cohort of 690 patients who underwent computed tomography-guided CCNB. The model was validated in 253 consecutive patients in the validation cohort and 250 patients in the test cohort. The area under the curve was used to determine the predictive accuracy of the proposed model. Results: The risk factors associated with pneumothorax after computed tomography-guided CCNB were sex, patient position, lung field, lesion contact with the pleura, lesion size, distance from the pleura to the lesion, presence of emphysema adjacent to the biopsy tract, and crossing fissures. The predictive model that incorporated these predictors showed good predictive performance in the training cohort [area under the curve, 0.71 (95% confidence interval: 0.67-0.75)], validation cohort [0.71 (0.64-0.78)], and internal test cohort [0.68 (0.60-0.75)]. The nomogram also provided excellent calibration and discrimination, and decision curve analysis (DCA) demonstrated its clinical utility. Conclusions: The predictive model showed good performance for pneumothorax after computed tomography-guided CCNB and may help improve individualized preoperative prediction.

Nomogram model to predict pneumothorax after computed tomography-guided coaxial core needle lung biopsy.

PURPOSE: To develop a predictive model to determine risk factors of pneumothorax in patients undergoing the computed tomography (CT)1-guided coaxial core needle lung biopsy (CCNB). METHODS: A total of 489 patients who underwent CCNBs with an 18-gauge coaxial core needle were retrospectively included. Patient characteristics, primary pulmonary disease, target lesion image characteristics and biopsy-related variables were evaluated as potential risk factors of pneumothorax which was determined on the chest X-ray and CT scans. Univariate and multivariate logistic regressions were used to identify the independent risk factors of pneumothorax and establish the predictive model, which was presented in the form of a nomogram. The discrimination and calibration of the model were evaluated as well. RESULTS: The incidence of pneumothorax was 32.91 % and 31.42 % in the development and validation groups, respectively. Age, emphysema, pleural thickening, lesion location, lobulation sign, and size grade were identified independent risk factors of pneumothorax at the multivariate logistic regression model. The forming model produced an area under the curve of 0.718 (95 % CI = 0.660-0.776) and 0.722 (95 % CI = 0.638-0.805) in development and validation group, respectively. The calibration curve showed good agreement between predicted and actual probability. CONCLUSIONS: The predictive model for pneumothorax after CCNBs had good discrimination and calibration, which could help in clinical practice.

Development of a predictive model of pulmonary hemorrhage in patients after CT-guided coaxial core needle lung biopsy / 西安交通大学学报(医学版)

【Objective】 To establish a predictive model for patients with hemorrhage after CT-guided coaxial core needle lung biopsy （CCNB） based on logistic regression. 【Methods】 A total of 489 patients who had undergone CCNB were retrospectively recruited. The potential risk factors of hemorrhage after lung biopsy were analyzed by univariate and multivariate logistic regression， through which we screened the independent risk factors and established a prediction model for hemorrhage. We evaluated the discrimination， calibration and clinical usefulness of the model. 【Results】 There were 141 cases （42.6%） of hemorrhage in the development group and 66 cases （41.8%） of hemorrhage in the validation group； there was no case of severe hemorrhage or hemothorax. Multivariate logistic regression analysis showed that fibrinogen degradation products， pulmonary interstitial fibrosis， largest diameter and puncture depth were independent predictive factors of hemorrhage. Hemorrhage prediction model was established and presented in the form of a nomogram. Discrimination of the model： the AUC was 0.837 in the development group and 0.777 in the validation group. The calibration curve showed good agreement between predicted probability and actual probability of hemorrhage. The unreliability test yielded a P value of 0.849 in the development group and 0.147 in the validation group. The DCA curve showed that the hemorrhage predictive model could increase the benefit of patients. 【Conclusion】 The predictive model of hemorrhage in patients after CCNB based on logistic regression can be used in clinical practice.