Multiparametric MRI-Based Integrated Analysis of Clinical, Radiomics, Deep Learning, and Machine Learning for...

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence in Medicine

Background:

• Locally advanced rectal cancer (LARC) requires accurate prediction of tumor cell proliferation and prognosis for effective management.
• Multiparametric magnetic resonance imaging (MRI) offers rich data for developing predictive models.

Purpose of the Study:

• To develop and validate a predictive model using multiparametric MRI, integrating clinical, radiomics, deep learning (DL), and machine learning (ML) techniques.
• To predict tumor cell proliferation status (Ki-67 expression) and prognosis in LARC patients.

Main Methods:

• Retrospective enrollment of 384 LARC patients from three centers.
• Extraction of radiomics and DL features from multisequence MRI.
• Construction and validation of predictive models using 12 ML algorithms and integration into a nomogram.

Main Results:

• The optimal radiomics-DL (RDL) model achieved high predictive performance for tumor cell proliferation.
• The integrated nomogram demonstrated excellent discrimination (AUCs: 0.939-0.895) for predicting prognosis.
• The nomogram effectively stratified patients based on recurrence-free survival (RFS).

Conclusions:

• A multiparametric MRI-based nomogram effectively integrates clinical, radiomics, DL, and ML data.
• This model shows robust performance in predicting Ki-67 expression and stratifying prognosis in LARC patients.
• The developed nomogram supports personalized management strategies for LARC.