Identifying ureteral stent encrustation using machine learning based on CT radiomics features: a bicentric study.

Area of Science:

• Urology
• Medical Imaging
• Artificial Intelligence

Background:

• Ureteral stent encrustation is a common complication after stent placement.
• Accurate identification of encrusted stents is crucial for timely intervention and patient management.

Purpose of the Study:

• To develop and validate radiomics and machine learning models for identifying encrusted ureteral stents.
• To compare the recognition performance of these models using multiple metrics.

Main Methods:

• Radiomics features were extracted from CT images of 354 patients.
• Six machine learning models (LR, DT, SVM, RF, XGBoost, KNN) were trained using selected radiomics features.
• A combined model incorporating radiomics score and indwelling time was constructed and validated.

Main Results:

• 20 significant radiomics features were selected from 1,409 extracted features.
• The combined model demonstrated high accuracy, sensitivity, and specificity across training, internal, and external validation cohorts.
• The combined model achieved an AUC of 0.810 in the external validation cohort, with superior performance in the combined model (AUC 0.999 in training).

Conclusions:

• Machine learning models based on radiomics features can effectively identify ureteral stent encrustation.
• The developed models show high accuracy and favorable clinical utility for detecting encrusted stents.
• Radiomics and machine learning offer a promising approach for non-invasive assessment of ureteral stent complications.