JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Machine learning prediction model for gray-level co-occurrence matrix features of synchronous liver metastasis in colorectal cancer

  • 0Department of Gastrointestinal Surgery, Renmin Hospital of Wuhan University, Wuhan 430030, Hubei Province, China.
World Journal of Gastrointestinal Surgery +

|

July 10, 2024

|

July 10, 2024

View abstract on PubMed

Summary

This summary is machine-generated.

This study developed a machine learning model using MRI-based radiomics to predict synchronous liver metastasis (SLM) in colorectal cancer (CRC) patients, aiding early clinical decisions.

Area Of Science

  • Oncology
  • Radiology
  • Medical Imaging

Background

  • Synchronous liver metastasis (SLM) significantly increases morbidity in colorectal cancer (CRC).
  • Current diagnostic methods lack effective algorithms to predict adverse SLM events in CRC patients.
  • Predicting SLM is crucial for timely and personalized patient management.

Purpose Of The Study

  • To identify risk factors associated with SLM in CRC.
  • To develop a visual prediction model for SLM using radiomic features from MRI.
  • To integrate Gray-Level Co-occurrence Matrix (GLCM) features for enhanced prediction accuracy.

Main Methods

  • Retrospective analysis of 392 CRC patients.
  • Inclusion of clinical parameters and GLCM features from MRI.
  • Development of prediction models using generalized linear regression, random forest (RFM), and artificial neural networks.
  • Evaluation of model performance using ROC and decision curves.

Main Results

  • 12.24% of patients presented with SLM.
  • Fourteen GLCM features were identified as significant predictors.
  • The RFM demonstrated high prediction efficiency with an AUC of 0.917 in the training set and 0.909 in the validation set.
  • Key features included inverse difference, sum entropy, and energy.

Conclusions

  • A predictive model integrating GLCM radiomic features and machine learning effectively predicts SLM in CRC.
  • This model offers a valuable tool for clinicians to support early and personalized treatment decisions.
  • The findings highlight the potential of radiomics in improving CRC patient outcomes.

Keywords: