Signs Based Lung CT Image Retrieval System

Medical Image is an important basis for Disease diagnosis. With the large increase number of medical images, there requires effective way for medical image retrieval. Compared to text-based retrieval methods, the advantage of content-based medical image retrieval is the ability to find visual similar images for doctor to reference and comparison. This can play an important role in the domain of Medical teaching, research and diagnosis.Signs mean image information shown by disease in different pathological stage and level. Doctor can diagnose disease or narrow diagnosis range by typical signs. There are many kinds of Signs in lung CT image and the performances of these signs are complex, which make false or missed detection easily occurred in doctor’s image-reading process. So building a signs-based lung CT image retrieval system which can detect signs type contained in query image and retrieval images contain the same type of signs can assist doctor identify signs correctly and assist disease diagnosis.In this paper, author first built a lung CT signs database. Nine kinds of hand-marked lung CT signs are storied in this database. Annotation program use VTK for CT image display and implement magnification, window adjusting, moving and delineation operation. User delineate signs contour and annotate signs sub category or significance in every signs category。System design relational database based storage and XML file based storage to store Signs information, in order to supply data to lung CT image retrieval system and for data publish.As a part of the lung CT image retrieval system, database providing data support for CT image retrieval system. The lung CT image retrieval system use C/S architecture contains clients, server and database. The server automatically recognize the signs contained in suspicious area in clients’ query images, returns the images in database which contains similar signs. System use WCF for communication between client and server, use TCP protocol, and apply singleton pattern and concurrency mode to improve telecomunication efficiency and throughout capacity. System use DCMTK to implement JPEG2000 based image lossless compression. When download result images, clients download compressed image and decompress at local. System uses this way to reduce network’s transmission data and improve telecommunication efficiency.