Abstract
Purpose: :
To develop a web-based system to quantitatively assess vascular caliber and tortuosity from Nidek NM200D non-contact camera retinal images in ROP patients. It will allow clinicians to recover a quantitative estimation of these parameters in an objective, fast and user-friendly way.
Methods: :
The analysis is performed with a client-server setup by using a regular web browser. Images are uploaded to the web server and a graphical user interface allows the client (i.e., user) to select the region of a vessel to be analyzed by clicking on its end-points. A variable number of additional points are then added to highlight the vessel course and guide vessel tracking. An interpolating spline, connecting the inserted points, is automatically drawn and displayed over the vessel, updated every time a new point is added (top panel). When the whole spline lies inside the vessel, processing can be started. Vessel axis and edges are detected using the constraint that vessels present a tubular structure delimited by two parallel borders. After image enhancement, all vessel edge points are extracted by means of a Canny filter. Refined axis and calibers along the vessel are then identified by linking corresponding pairs of points on opposite edges with a customized algorithm. From tracking results, retinal caliber and the retinal tortuosity score recently proposed by our group are computed (Grisan et al, IEEE TMI 2008). All processing is performed server-side and user receives the computed parameters in less than one minute (bottom panel).
Results: :
The system is currently under beta testing by clinicians at Children’s Hospital of Philadelphia and Scheie Eye Institute (Philadelphia, USA) and a protocol for clinical investigations is being developed. In a preliminary inter- and intra-observer study, ROPnet provided vessel calibers and tortuosities in excellent agreement with clinicians’ assessment.
Conclusions: :
A system for the quantification of vessel caliber and tortuosity in ROP retinal images will be soon available for general use.
Keywords: imaging/image analysis: clinical • retinopathy of prematurity • image processing