Deep learning-based multi-parametric magnetic resonance imaging (mp-MRI) nomogram for predicting Ki-67 expression in rectal cancer

Sikai Wu ¹, Neng Wang ¹, Weiqun Ao ²

⁰Zhejiang Chinese Medical University, Hangzhou, Zhejiang, China.

Abdominal Radiology (new York) +

|

March 15, 2024

View abstract on PubMed

Summary

A deep learning nomogram using multi-parametric MRI accurately predicts Ki-67 expression in rectal cancer, outperforming traditional clinical models. This tool aids clinicians in precise rectal cancer prognosis.

Area Of Science

Oncology
Radiology
Artificial Intelligence

Background

Ki-67 expression is a crucial prognostic marker in rectal cancer.
Accurate prediction of Ki-67 levels is essential for treatment stratification.
Current prediction methods may lack precision and convenience.

Purpose Of The Study

To evaluate a deep learning-based multi-parametric MRI (mp-MRI) nomogram for predicting Ki-67 expression in rectal cancer.
To compare the predictive performance of the deep learning nomogram against a clinical model.

Main Methods

Retrospective analysis of 491 rectal cancer patients from two centers.
Extraction and selection of deep learning features from mp-MRI data.
Construction of a deep learning model, a clinical model, and a combined nomogram.
Performance evaluation using ROC curves, calibration, and decision curve analysis.

Main Results

Deep learning identified key mp-MRI features for prediction.
The deep learning model and nomogram achieved high AUCs (0.88-0.98).
The deep learning nomogram significantly outperformed the clinical model (P < 0.001).

Conclusions

A deep learning-based nomogram effectively predicts Ki-67 expression in rectal cancer.
This tool offers accurate and convenient prediction for clinical decision-making.
Deep learning enhances the predictive value of mp-MRI for rectal cancer biomarkers.

Keywords:

Deep learning Ki-67 Multi-parametric MRI Nomogram Rectal cancer