A new version of Hyper-Fractal Analysis: .Net WPF module for RGB 3D reconstruction of medical three-channel images

Abstract

In Grossu et al. (2022) [1] we presented a module for fractal analysis of three-channel (most commonly CT: native, arterial contrast, and venous contrast) medical images stored in csv specific files coming from various sources (CT, MRI etc). In current paper we extended these functionalities to file series. We also developed a .Net WPF user control (a “three-dimensional Picture Box”) for facilitating 3D image reconstruction. For a better representation, the application allows various ways of filtering and highlighting the volumes of interest (specify parallelepipedal selections, apply classical or fuzzy band-pass filters, set voxel transparency and/or slice spacing, and representing each slice matrix component on a separated RGB channel). New version program summaryProgram Title: Hyper-Fractal Analysis v07CPC Library link to program files:https://doi.org/10.17632/z9knmny56p.3Licensing provisions: GPLv2Programming language: C# 7.3 /.Net Framework 4.7.1External routines: Prism, https://github.com/PrismLibrary/Prism-DocumentationJournal reference of previous version: Comput. Phys. Commun. 273 (2022) 108255Does the new version supersede the previous version?: YesNature of problem: Estimating the fractal dimension of images, multi-dimensional objects, and three-channel images. 3D reconstruction of medical images.Solution method: .Net WPF user control for representing 3D three-channel medical images stored in specific csv file series.Reasons for the new version: Develop a RGB 3D image reconstruction module for CT, MRI etc. three-channel images.Summary of revisions:• Hyper-Fractal Analysis v06 [1] allows opening specific csv files (.ct.csv) containing up to three matrices of integers (three-channel images). Although this functionality was initially intended for CT images [2], one could notice the high generality of the proposed solution. Thus, the csv data could come from various other DICOM sources (e.g. MRI). In order to better illustrate this fact, we changed the expected file extension to slice.csv.• The current version was also extended for opening folders containing image series. Following this purpose, each data “slice” must be stored in a separated three-channel csv file conforming to the following file name convention: [Name].[SliceIndex].slice.csv. The folder must also include a metadata file (meta.csv) containing various information of interest (pixel size, slice thickness, spacing between slices, and so on). The application GUI allows navigating between slices.• The ThreeDimPictureBox component was developed in Windows Presentation Foundation (WPF) UI framework [3] on top of the Model View-ViewModel (MVVM) pattern [4]. It was designed as a “3D PictureBox” user control for graphically representing a set of voxels specified as an input parameter.• One important challenge in 3D medical image reconstruction is related to the massive amount of data (e.g. 94,371,840 triangles for only 30 slices at 512 X 512 pixels). Taking into account the medical specific requirements, we mainly focused on representation's accuracy. Thus, for avoiding artifacts generated by interpolation algorithms, we did not implement any method of adapting slices with different resolutions [5]. The impact on performance is mainly managed by all filtering the pixels using the existing band-pass tool, restricting the analyzed data to a desired subset (parallelepipedal selection), choosing a lower resolution, and reducing the number of RGB colors. The algorithm will also merge the row adjacent voxels with the same color. As a transparency alternative, the user could opt for increasing the slice spacing (Fig. 1). For each combination of parameters, the numbers of necessary triangles and colors are priorly calculated, thus allowing the user to decide rendering or not the output voxels, as a function of known available resources.• In the previous version, we used three image channels for storing the native, arterial contrast, and venous contrast CT information. Each of the corresponding matrices is represented on a different RGB channel [6,7], which results in highlighting the image differences. The RGB colors resulting from this technique are also assigned to the voxels in the 3D reconstruction algorithm.• For avoiding the OutOfMemory exception, the application platform target was changed from Any CPU to x64.• Considering the 3D visualization difficulties involved by voxels occultation effects (even when setting high transparency levels) all previously mentioned functionalities were designed for helping the physician isolating and highlighting specific 3D data sub-sets, which is expected to be of significant interest in various medical areas [8–10].