SOFTWARE FOR REGIONS OF INTEREST RETRIEVAL ON MEDICAL 3D IMAGES

Abstract

Background. Implementation of software for areas of interest retrieval in 3D medical images is described in this article. It has been tested against large volume of model MRIs. Material and methods. We tested software against normal and pathological (severe multiple sclerosis) model MRIs from tge BrainWeb resource. Technological stack is based on open-source cross-platform solutions. We implemented storage system on Maria DB (an open-sourced fork of MySQL) with P/SQL extensions. Python 2.7 scripting was used for automatization of extract-transform-load operations. The computational core is written on Java 7 with Spring framework 3. MongoDB was used as a cache in the cluster of workstations. Maven 3 was chosen as a dependency manager and build system, the project is hosted at Github. Results. As testing on SSMU's LAN has showed, software has been developed is quite efficiently retrieves ROIs are matching for the morphological substratum on pathological MRIs. Conclusion. Automation of a diagnostic process using medical imaging allows to level down the subjective component in decision making and increase the availability of hi-tech medicine. Software has shown in the article is a complex solution for ROI retrieving and segmentation process on model medical images in full-automated mode. We would like to thank Robert Vincent for great help with consulting of usage the BrainWeb resource.