High-Performance CAD-CTC Scheme Using Shape Index, Multiscale Enhancement Filters, and Radiomic Features

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Gastroenterology

Background:

• Computer-aided detection (CAD) systems for computed tomography colonography (CTC) aim to detect colorectal polyps.
• Current CAD-CTC systems struggle with numerous false positives (FPs) due to complex colon structures.

Purpose of the Study:

• To propose a novel CAD-CTC scheme to address the FP issue in polyp detection.
• To improve the accuracy of automated polyp detection in CTC examinations.

Main Methods:

• The proposed scheme combines shape index and multiscale enhancement filters to generate polyp candidates.
• A supervised classification model utilizing 640 radiomic features (440 established, 200 new) was developed to reduce FPs.

Main Results:

• The CAD-CTC scheme achieved 98.1% by-polyp sensitivity and 95.3% per-scan sensitivity for polyps ≥5 mm.
• The system maintained a low FP rate of 1.3 FPs per dataset for polyps ≥5 mm.

Conclusions:

• The proposed CAD-CTC scheme demonstrates high sensitivity and a low FP rate.
• This advanced CAD-CTC system shows promise as a valuable tool for clinical colon examinations.