Computed tomography colonography: feasibility of computer-aided polyp detection in a "first reader" paradigm

Area of Science:

• Medical Imaging
• Gastroenterology
• Artificial Intelligence in Medicine

Background:

• Computed tomography colonography (CTC) is a key tool for colorectal cancer screening.
• Computer-aided detection (CAD) systems aim to improve the accuracy and efficiency of medical image interpretation.
• Evaluating CAD as a primary reader is crucial for optimizing its clinical workflow integration.

Purpose of the Study:

• To assess the feasibility of using a computer-aided detection (CAD) algorithm as the initial reader in computed tomography colonography (CTC).
• To compare the diagnostic performance and efficiency of radiologists reading CTC with and without CAD assistance.

Main Methods:

• A two-part, blinded trial involving 41 CTC studies.
• Phase 1: Radiologists interpreted CTC studies without CAD.
• Phase 2: Radiologists interpreted the same studies with CAD providing a list of potential polyps.

Main Results:

• Unassisted readers detected an average of 63% of polyps >= 10 mm.
• CAD use resulted in 74% sensitivity for polyps >= 10 mm (not statistically different from unassisted reading).
• CAD significantly decreased interobserver variability (P=0.017) and reduced the time to detect the first polyp, without affecting specificity.

Conclusions:

• Computer-aided detection (CAD) as a first reader in CTC demonstrates comparable per-polyp and per-patient sensitivity to unassisted interpretation.
• CAD integration into CTC workflows can enhance consistency among readers and expedite the identification of significant findings.