Evaluation of computer-aided detection and diagnosis systems.

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Diagnostic Technology Evaluation

Background:

• Computer-aided detection and diagnosis (CAD) systems assist clinicians in disease detection and interpretation.
• Current lack of standardized performance evaluation hinders comparison and clinical translation of CAD devices.
• The American Association of Physicists in Medicine (AAPM) established the Computer Aided Detection in Diagnostic Imaging Subcommittee (CADSC) to address this gap.

Purpose of the Study:

• To convey the opinions of the AAPM CADSC members regarding CAD system performance assessment.
• To stimulate the development of consensus approaches and best practices for evaluating CAD systems.
• To encourage the creation of structured guidelines for CAD performance evaluation.

Main Methods:

• Discussion of assessment approaches for standalone CAD systems.
• Evaluation of the impact of CAD systems on end-users.
• Presentation of AAPM CADSC member opinions and recommendations.

Main Results:

• Identified deficiencies in current CAD performance evaluation methods.
• Proposed the need for standardized approaches to compare CAD devices.
• Highlighted the importance of understanding CAD's clinical utility and limitations.

Conclusions:

• Standardized CAD performance assessment is crucial for clinical adoption and efficacy.
• Awareness of CADSC recommendations can foster structured guidelines for evaluation.
• Proper assessment will enhance understanding, stimulate research, and improve CAD utility in practice.