İnce Okan, Önder Hakan, Gençtürk Mehmet, Golzarian Jafar, Young Shamar
Department of Radiology, Rush University Medical College, Chicago, Illinois.
Department of Radiology, Health Sciences University, Prof Dr Cemil Tascioglu City Hospital, Istanbul, Turkey.
J Vasc Interv Radiol. 2025 Jan;36(1):99-105.e1. doi: 10.1016/j.jvir.2024.09.022. Epub 2024 Oct 9.
To evaluate the potential of interpretable machine learning (ML) models to predict ascites improvement in patients undergoing transjugular intrahepatic portosystemic shunt (TIPS) placement for refractory ascites.
In this retrospective study, 218 patients with refractory ascites who underwent TIPS placement were analyzed. Data on 29 demographic, clinical, and procedural features were collected. Ascites improvement was defined as reduction in the need of paracentesis by 50% or more at the 1-month follow-up. Univariate statistical analysis was performed. Data were split into train and test sets. Feature selection was performed using a wrapper-based sequential feature selection algorithm. Two ML models were built using support vector machine (SVM) and CatBoost algorithms. Shapley additive explanations values were calculated to assess interpretability of ML models. Performance metrics were calculated using the test set.
Refractory ascites improved in 168 (77%) patients. Higher sodium (Na; 136 mEq/L vs 134 mEq/L; P = .001) and albumin (2.91 g/dL vs 2.68 g/dL; P = .03) levels, lower creatinine levels (1.01 mg/dL vs 1.17 mg/dL; P = .04), and lower Model for End-stage Liver Disease (MELD) (13 vs 15; P = .01) and MELD-Na (15 vs 17.5, P = .002) scores were associated with significant improvement, whereas main portal vein puncture was associated with a lower improvement rate (P = .02). SVM and CatBoost models had accuracy ratios of 83% and 87%, with area under the curve values of 0.83 and 0.87, respectively. No statistically significant difference was found between performances of the models in DeLong test (P = .3).
ML models may have potential in patient selection for TIPS placement by predicting the improvement in refractory ascites.
评估可解释机器学习（ML）模型预测经颈静脉肝内门体分流术（TIPS）治疗顽固性腹水患者腹水改善情况的潜力。
在这项回顾性研究中，分析了218例行TIPS治疗的顽固性腹水患者。收集了29项人口统计学、临床和手术特征数据。腹水改善定义为在1个月随访时腹腔穿刺需求减少50%或更多。进行单因素统计分析。数据分为训练集和测试集。使用基于包装器的顺序特征选择算法进行特征选择。使用支持向量机（SVM）和CatBoost算法构建两个ML模型。计算Shapley加法解释值以评估ML模型的可解释性。使用测试集计算性能指标。
168例（77%）患者顽固性腹水得到改善。较高的钠（Na；136 mEq/L对134 mEq/L；P = .001）和白蛋白（2.91 g/dL对2.68 g/dL；P = .03）水平、较低的肌酐水平（1.01 mg/dL对1.17 mg/dL；P = .04）以及较低的终末期肝病模型（MELD）（13对15；P = .01）和MELD-Na（15对17.5，P = .002）评分与显著改善相关，而主要门静脉穿刺与较低的改善率相关（P = .02）。SVM和CatBoost模型的准确率分别为83%和87%，曲线下面积值分别为0.83和0.87。在DeLong检验中，模型性能之间未发现统计学显著差异（P = .3）。
ML模型通过预测顽固性腹水的改善情况，可能在TIPS治疗患者选择方面具有潜力。
