Improved False Positive Reduction by Novel Morphological Features for Computer-Aided Polyp Detection in CT...

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Radiology

Background:

• Computer-aided detection (CAD) systems enhance computed tomographic colonography (CTC) interpretation.
• Existing false positives (FPs) in CAD systems hinder their effectiveness.
• Reducing FPs is crucial for improving CAD performance in CTC.

Purpose of the Study:

• Develop novel morphological features to decrease FPs in CTC.
• Maintain high polyp detection sensitivity alongside FP reduction.
• Introduce new 3-D circular transformation (CcT) and quantized convergence index (QCI) filters for colonic polyps.

Main Methods:

• Computed volumetric feature maps using 3-D geodesic distance transformation, CcT, and QCI filters.
• Developed new morphological features based on curvature, fractal dimension, and volumetric maps.
• Evaluated features on 456 oral contrast-enhanced CT scans (228 patients, 130 polyps ≥5 mm).

Main Results:

• Reduced FP rate from 4.2 to 2.0 FPs/scan (52.4% reduction) at 96.2% by-polyp sensitivity.
• Reduced FP rate from 4.5 to 2.1 FPs/scan (53.3% reduction) at 93.9% per-scan sensitivity for polyps ≥5 mm.
• Demonstrated effective FP reduction without sacrificing detection sensitivity.

Conclusions:

• Newly developed morphological features effectively reduce FPs in CTC CAD systems.
• These features maintain high detection sensitivity, crucial for clinical utility.
• Advancements in CAD systems will better assist radiologists in CTC image interpretation.