Modeling and analysis of soft-test/repair for CCD-based digital X-ray systems

Authors:
Publisher website
Google Scholar
Abstract
Modern X-ray imaging systems evolve toward digitization for reduced cost, faster time-to-diagnosis, and improved diagnostic confidence. For the digital X-ray systems, charge coupled device (CCD) technology is commonly used to detect and digitize optical X-ray image. This paper presents a novel soft-test/repair approach to overcome the defective pixel problem in CCD-based digital X-ray systems through theoretical modeling and analysis of the test/repair process. There are two possible solutions to cope with the defective pixel problem in CCDs: one is the hard-repair approach and another is the proposed soft-test/repair approach. Hard-repair approach employs a high-yield, expensive reparable CCD to minimize the impact of hard defects on the CCD, which occur in the form of noise propagated through A/D converter to the frame memory. Therefore, less work is needed to filter and correct the image at the end-user level while it maybe exceedingly expensive to practice. On the other hand, the proposed soft-test/repair approach is to detect and tolerate defective pixels at the digitized image level; thereby, it is inexpensive to practice and on-line repair can be done for noninterrupted service. It tests the images to detect the detective pixels and filter noise at the frame memory level and caches them in a flash memory in the controller for future repair. The controller cache keeps accumulating all the noise coordinates and preprocesses the incoming image data from the A/D converter by repairing them. The proposed soft-test/repair approach is particularly devised to facilitate hardware level implementation ultimately for real-time telediagnosis. Parametric simulation results demonstrate the speed and virtual yield enhancement by using the proposed approach; thereby highly reliable, yet inexpensive, soft-test/repair of CCD-based digital X-ray systems can be ultimately realized.

This publication has 20 references indexed in Scilit: