AI-based Prediction of Catheter-related Thrombosis Risk for Cancer Patients

Yaoqi Guo, Yun Ma, Zijian Shao, Weichen Bi, Yanfeng Wang

Abstract

Cancer patients face a heightened risk of venous thromboembolism (VTE), emerging as the second most prevalent cause of death within this population. Central venous catheterization (CVC), a routine procedure in cancer care, amplifies the VTE risk, leading to catheter-related thrombosis (CRT). Although traditional risk-assessment models and certain AI methods exist for VTE prediction, their capability and application in CRT risk prediciton for cancer patients remains limited.This paper addresses the shortcomings of current models (RAMs) by crafting a dedicated AI model to predict CRT risks for cancer patients. Leveraging a dataset encompassing 10,512 cancer patients undergoing catheterization over a decade, we meticulously select nine specific features for model construction, resulting in an impressive 0.794 AUROC in prediction, 54.9% higher than baseline. Furthermore, we estimate CRT-free probability using the Kaplan-Meier method. We also develop a WeChat Mini Program designed for efficient data collection and risk prediction, enhancing the efficiency of CRT risk detection for both doctors and patients.
BibTeX
@inproceedings{guo2024ai,
  author = {Yaoqi Guo and Yun Ma and Zijian Shao and Weichen Bi and Yanfeng Wang},
  title = {{AI-based Prediction of Catheter-related Thrombosis Risk for Cancer Patients}},
  booktitle = {The Web Conference},
  year = {2024},
  doi = {10.1145/3589335.3651452},
}

Loading related content...

Figure