Evaluation of freehand SPECT image quality depending on different reconstruction parameters

Thesis by: Alexander Zhdanov
Advisor: Philipp Matthies and Alvaro Sanchez (SurgicEye GmbH)
Supervision by: Nassir Navab
Due date: 12.02.2013

Abstract

This project is performed in an innovative area of an intra-operative 3D nuclear imaging which offers a novel approach for robust and precise localization of functional information to facilitate less invasive, image-guided surgery.

The base for the project is declipseSPECT which is a system used for generating 3D images of radioactivity distributions inside a human body. In order to accomplish the task the system acquires readings of a conventional gamma probe together with positions and orientations of the gamma probe at the time the readings were acquired. The acquisition takes place while a user of the system moves the gamma probe over a region of interest of the body.

In order to generate images from readings of a gamma probe, the declipseSPECT implements an image generation process, which consists of two steps. In the first step – the data acquisition – the readings of the hand held gamma probe are recorded synchronously with the position and orientation of it. In the second step - the image reconstruction – the 1D projections are used to generate a 3D volumetric image using conventional iterative reconstruction algorithms.

In practical terms the image generation is modeled as a linear system, where the readings of the gamma probe are the result of the values of the image to be generated and the influence of them over the readings. The process of image reconstruction is in this way the process of solving a linear system.

For acquisition guidance the declipseSPECT implements a visualization that enables the user to know which areas have been scanned and which still require information. The generated images are then made available for the surgeon using methods of augmented reality. In particular the video image of the optical camera is acquired and augmented with the generated images, which are overlaid in the right perspective using the position of the patient reference.

An alternative is to display the images from the perspective of the instrument. This makes it possible to generate any possible projection the surgeon requires. These images, as well as the ones combined with the video of the of the patient’s body, are displayed on a touch screen monitor.

The goal of the given project is a modification of a software system used for a generation and analysis of images in the declipseSPECT. The additional modification is necessary for the more convenient comparison of images which are visualized for surgeons. This is useful in order to see differences between images reconstructed with various algorithm parameters. The scanning is also performed with different speeds and the system needs to give a feedback about errors occurred for different speeds.

The project will contain two parts: * Development of a user interface to load multiple 3D declipseSPECT images. Implementation of mechanisms for their selection, scrolling through 3D slices, comparison and visualization of results. * Research and implementation of computer vision algorithms to analyze differences between the images.

The proposed modification in the user interface could be implemented as QT widgets containing images. The user could scroll through the images and select them for the comparison. The system should give the result in another widget containing differences.

The programming language selected for the project is C++ since the system’s software is written in this language. The analysis and the visualization of the images will be performed using VTK library. The user interface will be implemented using QT library.

The project’s development will be performed with Visual Studio since it offers embedded support for the QT projects. The VTK is also configured for building with Visual Studio and GIT. The operation system accordingly is Windows 7.