JoVE - Journal of Visualized Experiments
JoVE Visualize
Contact Us

Computer-aided diagnosis tool utilizing a deep learning model for preoperative T-staging of rectal cancer based on three-dimensional endorectal ultrasound

  • 0Department of Medical Ultrasonics, Biomedical Innovation Center, The Sixth Affiliated Hospital, Sun Yat-sen University, Guangzhou, China.
Abdominal Radiology (new York) +

|

April 30, 2025

|

April 30, 2025

View abstract on PubMed

Summary

This summary is machine-generated.

A new deep learning computer-aided diagnosis (CAD) tool for rectal cancer staging using three-dimensional endorectal ultrasound (3D-ERUS) significantly improved radiologist accuracy and consistency in preoperative assessments.

Area Of Science

  • Oncology
  • Medical Imaging
  • Artificial Intelligence

Background

  • Accurate preoperative evaluation is crucial for rectal cancer patient prognosis and treatment.
  • Three-dimensional endorectal ultrasound (3D-ERUS) shows high accuracy in T staging.
  • Deep learning offers potential for enhancing diagnostic tools.

Purpose Of The Study

  • To develop a computer-aided diagnosis (CAD) tool using deep learning for preoperative T-staging of rectal cancer.
  • To evaluate the diagnostic performance of the CAD tool with 3D-ERUS images.
  • To assess the impact of the CAD tool on radiologists' interpretations.

Main Methods

  • Retrospective analysis of 216 rectal cancer patients who underwent 3D-ERUS.
  • Random assignment to training (n=156) and testing (n=60) cohorts.
  • Evaluation of radiologist performance with and without the CAD tool on test cohort images.

Main Results

  • The CAD tool demonstrated high diagnostic efficacy across all T stages, particularly for T1 tumors (AUC, 0.85).
  • Radiologist performance improved with CAD assistance: AUC for T1 tumors increased from 0.76 to 0.80 (junior radiologist 2).
  • Diagnostic consistency (κ value) improved significantly for both junior (0.31 to 0.64) and senior (0.52 to 0.66) radiologists with CAD use.

Conclusions

  • A deep learning-based CAD tool utilizing 3D-ERUS shows strong performance for rectal cancer T-staging.
  • The CAD tool has the potential to enhance radiologist performance and inter-reader consistency.
  • This AI tool can improve preoperative assessment accuracy for rectal cancer patients.

Keywords: