Real Time Adaptive Filtering for Digital X-Ray Applications

Abstract

Over the last decade, many methods for adaptively filtering a data stream have been proposed. Those methods have applications in two dimensional imaging as well as in three dimensional image reconstruction. Although the primary objective of this filtering technique is to reduce the noise while avoiding to blur the edges, diagnostic, automated segmentation and surgery show a growing interest in enhancing the features contained in the image flow. Most of the methods proposed so far emerged from thorough studies of the physics of the considered modality and therefore show only a marginal capability to be extended across modalities. Moreover, adaptive filtering belongs to the family of processing intensive algorithms. Existing technology has often driven to simplifications and modality specific optimization to sustain the expected performances. In the specific case of real time digital X-ray as used surgery, the system has to sustain a throughput of 30 frames per second. In this study, we take a generalized approach for adaptive filtering based on multiple oriented filters. Mapping the filtering part to the embedded real time image processing while a user/application defined adaptive recombination of the filter outputs allow to change the smoothing and edge enhancement properties of the filter without changing the oriented filter parameters. We have implemented the filtering on a Cell Broadband Engine processor and the adaptive recombination on an off-the-shelf PC, connected via Gigabit Ethernet. This implementation is capable of filtering images of 5122 pixels at a throughput in excess of 40 frames per second while allowing to change the parameters in real time.

Real time adaptive filtering for computed tomography applications

“…Adaptive filtering has demonstrated in the past that it could provide better images with a given X-Ray doses [7,9]. Reducing the dose significantly impacts the signal-to-noise ratio.…”

“…Therefore, although those methods show extremely good results, they suffer from various shortcomings that prevent a wide adoption in other modalities than those for which they have been designed for [6,7]. In this study, we start from a generic algorithm that can theoretically address all types of de-noising and edge enhancements endeavors and that has already been successfully used for filtering digital X-Ray frames in real time 9 . However, instead of displaying the filtered 2D projections, we use them as raw data for a cone beam reconstruction algorithm in a way similar to [7], except for CArm based scanners.…”

Implementing real-time adaptive filtering for medical applications on the cell processor using a generic multicore framework

“…For the Basic Filter Set, we have opted for a simple filter LP given by equation [1]. The complementary Band Pass (BP) filters are based on equation [2] and oriented along three directions, regularly distributed on the interval [0, π]. The first path gives the estimation of the local anisotropy.…”

On the Implementation of Boundary Element Engineering Codes on the Cell Broadband Engine

“…The Cell Broadband Engine is a new architecture that is already playing a significant role in the computing industry in some specific areas [13][14][15] and the knowledge of its strengths and also its current limitations is a decisive factor for engineers and scientists willing to find high-performance solutions to the increasing complexity of their problems and applications. To achieve this goal, this paper introduces the Cell Broadband Engine Architecture and describes in some detail the porting of a well-known serial engineering code to an efficient multi-level parallel implementation.…”