Effictive 3-D rendering for abdomen Boosting performance of existing transfer function specification methods Data adaptive medical volume enhancement via space frequency analysis TF specification controls the visual illustration of medical volumetric data by mapping data values to color and opacity and it is an integrated part of interactive Direct Volume Rendering (DVR). In recent years, the importance of generating multi-dimensional domains representing the texture properties has been emphasized in several studies. Accordingly, the superior performance of the brushlet based TF design method and its effective use in 3D visualization is reported in comparison with other statistical or spacefrequency based methods. This previously developed method uses only radiologist selected Space-Frequency Blocks (SFBs), which are produced by the brushlet transform of 3D medical image series, for reconstruction. The optimal SFB weights are calculated through SVM in order to minimize the error obtained by the comparison of the weighted SFB reconstruction and the desired 3D visualization. Figure A. Training strategy of the proposed systemPurpose: The purpose is to improve visualization quality of medical volumes by enhancing features of interest in the data using reconstruction of weighted over-complete SFBs of 3D Brushlet transform.
Theory and Methods:Existing approaches use machine learning to find weighted combination of filters inside a predefined set, such that the difference between desired texture and obtained signature is minimized. In this paper, instead of using a limited filter bank, the optimal weights of SFBs in an expansion are determined to extract a desired texture. Accordingly, a novel method is proposed for reconstruction with optimally weighted SFBs.
Results:Results show that weighted SFB reconstruction provides slightly higher performance compared to SFB selection. The performance increase in FP rates is higher than FN due to the removal of selected quadrants. The proposed automatic quadrant selection method has slightly better performance than manual selection. The results of its application showed enhanced visualization capabilities especially for the abdominal organs.
Conclusion:This paper proposes a new strategy for spatio-temporal identification and extraction of textures. It expands the entire image to Brushlet bases through SFBs, each of which include textures at varying scales and orientations. Since original image can be reconstructed exactly using the inverse transform, it is safe to claim that SFBs include all texture information in the image. The novel idea is shown to be able to find the optimal weights of SFBs such that only the texture of interest is reconstructed and others are suppressed.