A public image database to support research in computer aided diagnosis

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Pulmonology

Background:

• Computer-aided diagnosis (CAD) systems require large, well-documented datasets for optimal performance.
• The creation and availability of such training data are significant challenges in CAD research.
• Existing databases often lack the scope or accessibility to support diverse CAD applications.

Purpose of the Study:

• To introduce the Public Lung Database (PLD) as a resource for computer-aided diagnosis (CAD) research.
• To provide accessible datasets for the characterization of pulmonary nodules and other lung diseases.
• To facilitate the development and evaluation of CAD systems by offering documented data.

Main Methods:

• Development of the Public Lung Database (PLD) for drug response research.
• Augmentation of the PLD with datasets for CAD research beyond pulmonary nodules.
• Provision of small, well-documented datasets, including manual and automated documentation.
• Inclusion of datasets with special properties for task complexity and system calibration.

Main Results:

• The PLD provides readily available, documented datasets for CAD research.
• The database supports research in pulmonary nodule characterization and other lung disease applications.
• Specialized datasets are available for evaluating CAD system performance and complexity.

Conclusions:

• The PLD serves as a valuable starting point for researchers exploring new CAD applications.
• This resource helps define issues related to specific CAD tasks and promotes field development.
• Making datasets available overcomes limitations in data accessibility and cost for CAD research.