2.4 X-ray Computed Tomography (CT)
2.4.2 Operating Principles
2.4.2.3 Parameters
CT image quality depends on different parameters that can be set prior to a scan. Among them, those of most importance are slice thickness, radiation dose, and reconstruction kernel (see Table 2.2).
Slice thickness represents the width of each slice of the image and can be changed after the image acquisition according to specific user needs. Narrow slice thickness leads to better edge definition, better high resolution contrast (meaning that small objects and details can be differentiated from surrounding structures with a high accuracy), and less partial volume effects (a phenomenon that causes
Table 2.2: This table shows the typical parameters of a CT scan.
Parameter Units Effect
Slice Thickness mm Edge definition and resolution contrast Radiation Dose mA Image quality and noise presence Convolution Kernel \ Edge definition and noise
a blurring effect across boundaries due to a loss of apparent intensity in small objects or regions because of the limited resolution), at the cost of a higher signal noise and poorer low contrast resolution (the ability to define objects and details which can be differentiated from surrounding structures with very little density difference). On the other hand, large thickness leads to less noise on the image and better low dose contrast, but it also results in poorer edge definition, worse high resolution contrast, and a higher presence of partial volume effects.
Radiation dose refers to the amount of x-rays emitted during the acquisition and has to be set before the scan starts. It is measured in mA as it represents the level of current within the x-ray source. The higher the current, the higher the amount of x-rays emitted and, thus, the dose. High radiation doses allow for better quality images, with more detail and less noise, but expose the patient to higher risks. Therefore, a good trade-off between radiation exposure to the patient and image quality which is useful for diagnostic purposes has to be considered.
The reconstruction kernel, known also as convolution kernel, is the filter or algorithm applied on the acquired data to reconstruct the final scan image. It has a significant impact on spatial frequency and noise characteristics of an image, and different scanner brands may have different kernels. In general, a reconstruction kernel varies in a range soft to sharp, with a soft kernel leading to smooth edges and reduced image noise, whereas a sharp kernel will enhance the edges at the cost of a higher image noise [53, 54]. There is no a perfect choice for CT parameters, and physicians tend to change them dependent on different patients.
2.5 3D Slicer
2.5
3D Slicer
The main goal of the research was to develop a free open-source tool for virtual bronchoscopy. For this reason, the freely-available and easily extendible open- source package, 3D Slicer, has been used. 3D Slicer is a software package for visualization (including volume rendering) and image analysis (including image registration and segmentation), natively designed to be available on multiple platforms, including Windows, Linux and Mac OS X. It also supports multi- modality imaging including, MRI, CT, US, nuclear medicine, and microscopy, and standard image file formats, such as DICOM. The application integrates interface capabilities to biomedical research software and image informatics frameworks.
3D Slicer was initiated as a master’s thesis project between the Surgical Plan- ning Laboratory (SPL) at the Brigham and Women’s Hospital and the MIT Artificial Intelligence Laboratory in 1998. Since then it has been continuously improved and in 2007 a completely re-architected version was released. The cur- rent version is Slicer 4.5 which was released in December 2015. To date, Slicer has been downloaded by thousands of users and developers worldwide and has enabled hundreds of academic publications. Slicer’s users and developers are also part of a community (through two different mailing lists) that provides help and suggestions when problems occur.
3D Slicer consists of more than one million lines of code, mostly C++ and Python, and is distributed under a Berkeley Software Distribution (BSD) license. However, while available for clinical research, it is not FDA approved for di- agnosis [55–57]. Permissions and compliance with applicable rules are the re- sponsibility of the user. 3D Slicer is built on a set of powerful and widely used software components, such as Tool Command Language (Tcl/Tk) [58], Visual- izaion Toolkit (VTK) [59], and Insight Segmentation and Registration Toolkit (ITK) [60–62] to which is added an application layer, incorporating a graphical user interface (GUI) which makes the system easily usable for non-programmer end-users. The main Slicer’s feature is its powerful plug-in capabilities for adding and modifying new algorithms and applications. In fact, along with the standard and basic algorithms available, Slicer provides a type of free App Store, called Ex- tensions Manager, from which it is possible to download, use or modify the Slicer
extensions, new algorithms added to Slicer from developers around the world in the same way as applications for a smartphone. More than 50 extensions and packages of extensions are currently available and these are in continuous im- provement. The Extensions Manager also allows for upload of new extensions to Slicer, so that these can be used and tested by the community and feedback can be obtained. Thanks to the modularity of Slicer, developers can focus on their area of expertise without extensive knowledge of the larger platform. In advance of this thesis, Slicer did not provide any system for virtual bronchoscopy, as it is used mainly for brain imaging. Beside 3D Slicer, there exist other freely avail- able imaging processing platforms. Among them, Osirix [63] and MITK [64] are probably the most popular.
Osirix and MITK
Osirix is a software dedicated to DICOM images which has been specifically designed for navigation and visualization of multimodality and multidimensional images. It supports a complete plug-ins architecture that allows for the expansion of the capabilities of Osirix according to personal needs, but it currently does not provide many tools for image processing operations. Moreover, Osirix is currently available only for Apple Mac OS computers.
On the other hand, the Medical Imaging Interaction Toolkit (MITK) is an open-source platform for medical image analysis which aims at providing support for an efficient software-development of methods and applications dealing with medical images. As well as 3D Slicer, MITK is available on Windows, Mac OS X, and Linux. It is based on ITK and VTK classes, and it is not FDA approved. However, being more recent than 3D Slicer, MITK contains less basic tools for medical image processing. 3D Slicer has been chosen for the present research as a new project is currently being developed at SPL in order to create a branch of Slicer exclusively dedicated to chest imaging. This is referred to as Slicer Chest Imaging Platform, or SlicerCIP, and it is still under development, with a future release in mind.