Radiologic image assessment using information loss theory by specially designed low contrast detail phantoms and extending it to CT

Alghamdi, S 2016, Radiologic image assessment using information loss theory by specially designed low contrast detail phantoms and extending it to CT, Doctor of Philosophy (PhD), School of Health and Biomedical Sciences, RMIT University.

Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Alghamdi.pdf Thesis application/pdf 5.80MB
Title Radiologic image assessment using information loss theory by specially designed low contrast detail phantoms and extending it to CT
Author(s) Alghamdi, S
Year 2016
Abstract All radiographic imaging modalities are imperfect; they produce images that are affected to some degree by loss of object information or details. This can be addressed by exposing the subject being imaged to beams of different exposure values or extending the exposure time, but this result in an increase in the radiation dose delivered to the subject, a factor which must be minimised in clinical applications. In order for radiographic imaging modalities to be calibrated to minimise the dose delivered to a patient while still capturing images of sufficient detail to facilitate diagnosis, various methods of image quality (IQ) assessment have been developed which determine the efficiency and low contrast detectability of these modalities. A further approach to the reduction of the dose delivered is applied in the latest types of Computed Tomography (CT) modalities, in which incomplete slices are obtained and utilised in the reconstruction of images. These slices are made through very short periods of time which also reduces the dose to the patient but leads to some artefacts in the image in addition to some extra information loss. IQ assessment is therefore critical to these new types of CT modalities as well [1].

Assessing IQ is normally either done by using equipment (objective method) or by visualisation of images by professionals (subjective method) [more information about different methods of image quality evaluation is discussed in (Chapter 2, Section 2.8)]. Objective assessments of image quality associated with diagnostic imaging systems are most often equipment-based, such as noise analysis or modulation transfer functions [2]. These methods do not consider the effects of the image assessor, who is typically a radiologist, nor the effects of the viewing system and conditions.

Image quality can also be assessed subjectively. The subjective method is based on the observer’s perception of the assessor. This method requires multiple observers who individually identify a visible object (threshold details) for every detailed parameter available in the image. The human decision criteria are considered a fundamental element, as they can be included in the imaging chain when evaluating image quality, due to their crucial role in the medical diagnosis process. In this method, the radiologists usually assess the diagnostic images with the receiver operator characteristics (ROCs) to compare the performance of different imaging systems [more details about the ROC in (Chapter 2, Section 2.8.3)]. Despite the fact that ROC analysis considers the whole imaging chain, such as human observers and equipment, this analysis is a time-consuming process and cannot be readably adopted as a quality assurance (QA) method in a busy clinical practice [3].

The most common alternative approach to assess IQ is the use of a contrast detail phantom (CDP). A CDP can provide useful information on contrast detail detectability and is considered the most reliable form of IQ assessment, particularly in low-contrast conditions [3]. In fact, the CDP is referred to as a low-contrast detail (LCD) phantom, and the commercially available phantom is called CDRAD 2.0. The CDRAD phantom is made of acrylic (Perspex; polymethyl methacrylate) which is 10 mm thick and in which 225 cylindrical holes of various sizes and depths are drilled. The diameter of the holes varies in size from 0.3 mm to 8 mm. This range is equally distributed across 15 depths. These depths range from 8 mm (providing high contrast) to 0.3 mm (providing low contrast). Hence, the CDRAD phantom uses the air–acrylic interface to create image contrast [3]. This method involves both equipment and observers.

This thesis is mainly focused on the use of CDPs in the evaluation of IQ in two X-ray-based modalities: Conventional radiography and CT. The evaluation of the former modality includes computed radiography (CR) and digital radiography (DR) systems. A modified CDRAD phantom which is based on the CDP approach is proposed and evaluated. This modified phantom utilises a smaller attenuation differential than unmodified CDRAD and it is more closely representative of the tissues found in the human body.

The three main aims of this thesis are then as follows:

The first aim:
To modify the current CDRAD phantom by replacing the air-filled holes fully with water or contrast media. Replacing the air-filled holes at the same phantom with the contrast media can create a gradation of contrast measurements that can be varied and extended by adding different amounts (concentrations) of contrast media into the holes (Chapter 5) because air attenuates much less radiation compared with other media, including Perspex. After creating a low contrast phantom by replacing the air-filled holes with water at the CDRAD, the LCD of these two interfaces (air-Perspex and water-Perspex) is investigated by utilising of information loss (IL) theory in the DR system (Chapter 6). The investigation of applying the IL theory with the CDRAD is extended to include the evaluation of the image quality of the CR system using two different techniques anti-scatter grid and non-grid on the unmodified CDRAD containing air filled holes (Chapter 7). Finally, the study of the grid effect includes the flat panel direct (DR) system by using the IQFinv factor to assess the image quality of different DR systems with and without the grid by using the CDRAD phantom containing air-filled holes (Chapter 8).

The second aim:
To create a special contrast-detail phantom to evaluate the IQ and assess the LCD of CT scanners more efficiently than the commercially available Catphan phantoms by including a wide dynamic range that can be modified to assess any required level of LCD, which will be called CTCDP. Like the CDRAD phantom used in conventional radiography, the CTCDP incorporates a central slide that contains holes of different diameters. The diameter of these holes incrementally increases in size from the middle of the phantom to the phantom edge, as follows: 1.0 mm, 2.5 mm, 5 mm, 6 mm, 7.5 mm, 9.0 mm, 10.0 mm, 11.0 mm and 12.5 mm (more details about the CTCDP in Chapter 9). The detection of the LCD at this phantom is measured by two factors: IQF and IL (Chapters 9 and 10). This phantom complements the existing Catphan and extends its applicability to much lower contrast values and also it extends its contrast dynamic scale.

The third aim:
To investigate the effects of the contrast media on the CTDI value and relate it to the CT numbers by using theoretical and experimental methods. The theoretical method employed a new derivation of the known CTDI formula . This new derivation accounts for the presence of the contrast media as a factor when determining the dose enhancement. The experimental part includes the determination of the dose enhancement by using the contrast media and Gafchromic films (Chapter 11).
These studies will be of great value to the radiology community and to all the CT users because it develops a method of estimating the level of information loss during imaging procedures that significantly enhances the X-rays based modalities currently employed.

Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre School of Health and Biomedical Sciences
Subjects Radiology and Organ Imaging
Keyword(s) Image quality
Information loss (IL)
Computed tomography (CT)
Contrast detail phantom CDP
Image quality assessment
Version Filter Type
Access Statistics: 173 Abstract Views, 420 File Downloads  -  Detailed Statistics
Created: Fri, 13 Jan 2017, 09:30:55 EST by Adam Rivett
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us