Mediman: an object oriented programming approach for medical image analysis

Mediman is an image analysis package which has been developed to analyze quantitatively PET (positron emission tomography) data. It is object-oriented, written in C++, and based on InterViews on top of which new classes have been added. Mediman accesses data through NSF using external data representation or an import/export mechanism which avoids data duplication. Multimodality studies are organized in a database which includes images, headers, color tables, lists and objects of interest (OOIs), and history files. Stored color table parameters allow one to focus directly on the interesting portion of the dynamic range. Lists allow one to organize the image database according to modality, acquisition protocol, and time and spatial properties. OOIs are stored in absolute 3-D coordinates, allowing correlation with other coregistered imaging modalities such as magnetic resonance imaging or single photon emission computed tomography. OOIs have visualization properties and are organized into groups. Quantitative region-of-interest analysis of anatomic images consists of position, distance, and volume calculation on selected OOIs.