JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

A radiomics-boosted deep-learning for risk assessment of synchronous peritoneal metastasis in colorectal cancer

  • 0Medical School of Nantong University, Nantong, JiangSu, China.
Insights Into Imaging +

|

June 17, 2024

|

June 17, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

A novel radiomics-boosted deep learning model using PET/CT images accurately predicts synchronous colorectal cancer peritoneal metastasis (CRPM). This tool aids in personalized treatment and follow-up for patients with poor prognoses.

Area Of Science

  • Medical Imaging
  • Artificial Intelligence
  • Oncology

Background

  • Synchronous colorectal cancer peritoneal metastasis (CRPM) presents a significant challenge due to its poor prognosis.
  • Early and accurate risk assessment of CRPM is crucial for effective patient management.

Purpose Of The Study

  • To develop and validate a radiomics-boosted deep learning model for predicting synchronous CRPM using PET/CT imaging.
  • To evaluate the model's performance in identifying patients at high risk for CRPM.

Main Methods

  • A deep learning model based on ResNet50 was trained using PET/CT image patches and radiomic feature maps (RFMs).
  • The model incorporated radiomic features extracted from CT and PET images to enhance predictive capabilities.
  • Model performance was assessed using the area under the curves (AUC) across training and validation datasets.

Main Results

  • The radiomics-boosted deep learning model achieved high AUC values, ranging from 0.885 to 0.926 across different datasets.
  • The model demonstrated consistent calibration and was identified as the most predictive tool.
  • The study explored the contribution of the peritumoral region to CRPM assessment.

Conclusions

  • The developed radiomics-boosted deep learning model shows superior performance in the preoperative prediction of synchronous CRPM.
  • This AI-driven approach offers potential for personalized treatment strategies and optimized follow-up plans for colorectal cancer patients.
  • The synergy of radiomics and deep learning provides a valuable tool for improving outcomes in CRPM management.

Keywords: