Masters Degrees (Medical Physics)

Permanent URI for this collection

https://scholar.sun.ac.za/handle/10019.1/70981

Browse

Recent Submissions

Now showing 1 - 5 of 7
  • Thumbnail Image
    Item
    The application of a deep convolution neural network for the automated delineation of the target and organs at risk in High Dose Rate Cervical Brachytherapy
    (Stellenbosch : Stellenbosch University, 2022-12) Duprez, Didier Raphael Roger; Trauernicht, Christoph Jan; Stellenbosch University. Faculty of Medicine and Health Sciences. Dept. of Medical Imaging and Clinical Oncology. Medical Physics.
    ENGLISH SUMMARY: Low/middle income countries suffer from large deficits in experienced Radiation Oncologists, Medical Physicists and Radiation Therapists. Due to these deficits, the bottlenecks experienced in the High-dose rate (HDR) cervical brachytherapy treatment planning workflow are amplified. Image-guided HDR cervical brachytherapy is a complex, labour intensive, manual, time-consuming and expertise driven process. Automation in radiotherapy treatment planning, especially in brachytherapy, has the potential to substantially reduce the overall planning time however most of these algorithms require high level of expertise to develop. The aim of this study is to implement the out of the box self-configuring deep neural network package, known as No New U-Net (nnU-Net), for the task of automatically delineating the organs at risk (OARs) and high-risk clinical target volume (HR CTV) for HDR cervical brachytherapy. The computed tomography (CT) scans of 100 previously treated patients were used to train and test three different nnU-Net configurations (2D, 3DFR and 3DCasc). The performance of the models was evaluated by calculating the Sørensen-Dice similarity coefficient, Hausdorff distance (HD), 95th percentile Hausdorff distance, mean surface distance (MSD) and precision score for 20 test patients. The dosimetric accuracy between the manual and predicted contours was assessed by looking at the various dose volume histogram (DVH) parameters and volume differences. Three different radiation oncologists (ROs) scored the predicted bladder, rectum and HR CTV contours generated by the best performing model. The manual contouring, prediction and editing times were recorded. The mean DSC, HD, HD95, MSD and precision scores for our best performing model (3DFR) were 0.92/7.5 mm/3.0 mm/ 0.8 mm/0.91 for the bladder, 0.84/13.8 mm/5.2 mm/1.4 mm/0.84 for the rectum and 0.81/8.5 mm/6.0 mm/2.2 mm/0.80 for the HR CTV. Mean dose differences (D2cc/90%) and volume differences were 0.08 Gy/1.3 cm3 for the bladder, 0.02 Gy/0.7 cm3 for the rectum and 0.33 Gy/1.5 cm3 for the HR CTV. On average, 65 % of the generated contours were clinically acceptable, 33 % requiring minor edits, 2 % required major edits and no contours were rejected. Average manual contouring time was 14.0 minutes, while the average prediction and editing times were 1.6 and 2.1 minutes respectively. Our best performing model (3DFR) provided fast accurate auto generated OARs and HR CTV contours with a large clinical acceptance rate. Future work should focus on including larger datasets to eliminate inconsistencies, as well as focus on automating the generation of treatment plans.
  • Thumbnail Image
    Item
    Application of gradient dose segmented analysis as a treatment quality indicator for patients undergoing volumetric modulated arc radiotherapy
    (Stellenbosch : Stellenbosch University, 2022-12) van Reenen, Christoffel Jacobus; Trauernicht, Christoph Jan; Stellenbosch University. Faculty of Medicine and Health Sciences. Dept. of Medical Imaging and Clinical Oncology. Medical Physics.
    ENGLISH SUMMARY: The gamma analysis metric is a commonly used metric for volumetric modulated arc radiotherapy (VMAT) plan evaluation. The major drawback of this metric is the lack of correlation between gamma passing rates and dose-volume histogram (DVH) values for planning target volumes (PTV). The novel gradient dose segmented analysis (GDSA) metric was developed by Steers et al. to quantify changes in the PTV mean dose (Dmean) for patients undergoing VMAT. In this study, the GDSA metric was applied to 115 head-and-neck cancer patients treated on the Varian Halcyon v2.0 linear accelerator between August 2019 and July 2020 in the Division of Radiation Oncology. The GDSA indicated that a total of 13 patients had received at least one treatment fraction where the PTV Dmean exceeded 3% compared to the first treatment fraction. The kilovoltage cone-beam computed tomography (kV CBCT) images of these patients were analysed to determine the cause. The maximum predicted change in the PTV Dmean was 4.83%. Measurable changes in anterior-posterior and lateral separations were observed for 8 out the 13 patients (62%) where the change in PTV Dmean exceeded 3%. The maximum calculated effective separation change diameter was calculated as 3.86 cm. In cases where the change in PTV Dmean was less than 3%, no measurable separation changes were observed. The pitch-, roll- and yaw-rotational errors were quantified as the Halcyon treatment couch does not allow for online rotational corrections. The maximum pitch, roll and yaw rotational errors were 3.91º ± 0.89º, 3.07º ± 0.51º and 2.62º ± 0.40º, respectively. The mean errors were 0.9º, 0.45º, and 0.43º, for pitch, roll and yaw, respectively. The obtained results demonstrated that large deviations in PTV Dmean (>3%) were more likely due to change in effective diameter, whereas small deviations in PTV Dmean combined with separation changes less than 1 cm, were more likely caused by errors in pitch for long treatment fields. Weight loss during radiotherapy is well documented and proven to be the highest among head-and-neck cancer patients. The GDSA easily be implemented to identify setup/immobilization errors, as well as aid the department in scheduling new CT scans for patients experiencing continuous weight loss before significant differences in dose delivery occur.
  • Thumbnail Image
    Item
    A multi-institutional quantitative survey of multi-leaf collimator accuracy using a digital picket fence test with sub-millimetre detection capabilities.
    (Stellenbosch : Stellenbosch University, 2022-04) Boonzaier, Willem Petrus Engelbrecht; Trauernicht, C. J.; Stellenbosch University. Faculty of Medicine and Health Sciences. Dept. of Medical Imaging and Clinical Oncology. Medical Physics.
    ENGLISH SUMMARY: When patients are treated with modern radiotherapy, multi-leaf collimators (MLC) are used to shape treatment fields. With recent advances in technology, MLCs have also been used to modulate dose in certain areas inside patients to obtain better treatment outcomes. For the desired treatment outcomes to be valid, the equipment used to deliver the radiotherapy must function optimally. This study was aimed at creating software that can measure the accuracy of the MLC independently with routine picket fence test measurements. With this in place, the software was used to conduct a national survey of the accuracy of MLC systems on various medical linear accelerators in South Africa. Using on-board imaging and the in-house created software, MLC position and width errors could be found efficiently and with a very low degree of user input. When the results from the software were validated with measurements, other commercial software, and log files from linear accelerators, the average errors found agreed within 0.27mm, 0.49mm and 0.11mm for the measurements, commercial software and log file comparisons respectively. A national survey could be done with data from participating volunteer centres. A distinction could be made between units with more advanced capabilities such as VMAT and SRS capabilities by including picket fence tests acquired during arc delivery and picket fence tests done as a measure of the performance of the MLC for VMAT and SRS applications. The results showed that the software was easy and non-costly to compile. Results indicate that the software is robust enough to be used across a multitude of vendor types but sensitive enough to detect MLC errors in the order of 1 pixel of the onboard detector used. The software indicated that all units that participated in this study met the national standards of 2mm MLC accuracy. SRS capable units all had MLC accuracies under 1mm.
  • Thumbnail Image
    Item
    Mandibular morphological variation : implications for fracture repair
    (Stellenbosch : Stellenbosch University, 2019-12) McKay, Cyrilleen; Baatjes, Karin J.; Kotze, Sanet; Stellenbosch University. Faculty of Medicine and Health Sciences. Dept. of Medical Imaging and Clinical Oncology. Medical Physics.
    ENGLISH SUMMARY : Achieving a predictable clinical outcome during mandibular fracture repair necessitates thorough knowledge of variations of the neurovascular bundle, the location of tooth roots and bone quantity in the region of interest. In South Africa, the prevalence of mandibular body fractures due to alcohol-related interpersonal violence is increasing and is largely stratified socio-economically. Approximately 80% of people presenting with mandibular fractures rely on the resource-limited public health care system. In addition, South African population groups have a high prevalence of structures resembling accessory mental foramina (AMF) which may impede fracture fixation and outcomes. These considerations form the basis of this study which aimed to define population specific information on interforaminal variations and assess their applicability in clinical decision-making prior to fracture repair, using dry mandibles. Hemi-mandibles (N = 213) with known age and varying tooth loss patterns were obtained from four ancestry and sex subgroups, namely South African Coloured (SAC) females, SAC males, Black (SAB) males and White (SAW) males. The location of the mental foramen (MF) and AMF was determined in relation to mandibular topographical landmarks. Buccal cortical plate (BCP) and buccal bone thickness was assessed at 12 points – four points on transverse planes through, superior, and inferior to the MF midpoint. Transverse planes correspond with possible locations for mini-plate fracture fixation and the four points on each plane corresponds with locations for mono-cortical screw insertion. The MF was most commonly located between the first and second premolar teeth and the distance from the symphysis menti to the anterior border of the MF was smaller in SAC males when compared to SAC females. However, this parameter had a greater reading on right hemi-mandibles of SAB males when compared to SAC males. Tooth loss was associated with a decreased height of the mandible superior to the MF and age was associated with an increased MF diameter. Accessory mental foramina were observed in 6.54% of hemi-mandibles and was most commonly located mesial and superior to the MF. The BCP differed between subgroups and showed negative associations with tooth loss and age at selected assessment points. The buccal bone was thickest at the foraminal transverse plane when compared to superior and inferior transverse planes. It was thicker in SAC females when compared to SAC males on the inferior transverse plane of left hemi-mandibles. Overall, the influence of tooth loss and age on mandibular morphology did not vary between sex and ancestral subgroups. Results show that in comparison to superior and inferior transverse planes, the foraminal transverse plane had the lowest risks for inadvertent injuries to vital structures. Risks on this plane increased from 1.9 to below 8% for screw lengths 4 – 8 mm bilaterally. These findings expand population-specific knowledge of anatomical variations which could aid clinical and preoperative decision-making in the repair of mandibular fractures in South Africa.
  • Thumbnail Image
    Item
    The value of different reconstruction algorithms for quantification of FDG PET brain imaging
    (Stellenbosch : Stellenbosch University, 2016-12) Moalosi, Tumelo Carel Godwin; Ellmann, Annare; Mix, Michael; Warwick, James; Du Toit, Monique; Doruyter, Alexander; Stellenbosch University. Faculty of Medicine and Health Sciences. Dept. of Medical Imaging. Medical Physics.
    ENGLISH SUMMARY : Modern reconstruction techniques of positron emission tomography/computed tomography (PET/CT) data are optimized for whole body imaging. Such optimization is less developed for brain imaging. This study aimed at investigating the effect of different image reconstruction parameters (varying number of iterations, scan duration, relaxation parameter (smoothing levels) and the use of time of flight (TOF)) on PET/CT images with the objective of evaluating the algorithms for quantification of fluorodeoxyglucose (FDG) PET brain imaging. Materials and methods: A Philips Gemini TF Big Bore PET/CT scanner was used for acquiring the data. The study was based primarily on phantom and limited patient data for preliminary validation. Three dimensional (3D) Hoffman brain phantom (HBP) data and data of patients attending the Western Cape Academic PET/CT Centre for oncological purposes, with low probability of neurological pathology, were included in the study. The data was reconstructed using two different iterative reconstruction algorithms, row action maximum likelihood algorithm (RAMLA) and spherically symmetric basis function ordered subset algorithm (BLOB or BLOB OS), with variation in the number of iterations, scan acquisition duration, switching TOF on and off for BLOB OS and by varying the relaxation parameter. The set of output images were analyzed using MATLAB code. Results: From the HBP data, in all regions of the brain, the grey matter/white matter ratio, and the mean and the normalized mean counts increased as the number of iterations increased, reaching a plateau after 15 iterations for all algorithms. When comparing the algorithms with relaxation values λ=0.7 and λ=1.0, it was found that the latter converged faster. Overall, BLOB TOF (λ=1.0) proved to have faster convergence followed by BLOB TOF (λ=0.7). The coefficient of variation (COV) for all volumes of interest showed BLOB TOF to be superior compared to all the other algorithms. The COV results for different scan durations showed that there is minimal improvement after 5 min in high-activity regions (GM) and after 10 min in low-activity region (WM). The patient data was used as proof of principle but the numbers were too small to analyze further, as no pattern of behaviour could be identified for the different algorithms in the three patient images available. Conclusions: A higher number of iterations, such as 15, than currently used by the vendor of the PET scanner led to improved image quality for all algorithms. An acquisition time of 10 min provided an optimal trade-off between image quality and scan time irrespective of the reconstruction algorithm used. Including the TOF in the reconstruction algorithm improved the image quality, proving that TOF also improves image quality for small objects such as the brain similar to that seen for larger anatomical diameters as indicated in the literature.