Ultrasound And Pet-Ct Image Fusion For Prostate Brachytherapy Image Guidance

ABSTRACT

Fusion of medical images between different cross-sectional modalities is widely used, mostly where functional images are fused with anatomical data. Ultrasound has for some time now been the standard imaging technique used for treatment planning of prostate cancer cases. While this approach is laudable and has yielded some positive results, latest developments have been the integration of images from ultrasound and other modalities such as PET-CT to compliment missing properties of ultrasound images.

This study has sought to enhance diagnosis and treatment of prostate cancers by developing MATLAB algorithms to fuse ultrasound and PET-CT images. The fused ultrasound-PET-CT image has shown to contain improved quality of information than the individual input images. The fused image has the property of reduced uncertainty, increased reliability, robust system performance, and compact representation of information. The objective of co-registering the ultrasound and PET-CT images was achieved by conducting performance evaluation of the ultrasound and PET-CT imaging systems, developing image contrast enhancement algorithm, developing MATLAB image fusion algorithm, and assessing accuracy of the fusion algorithm.

Performance evaluation of the ultrasound brachytherapy system produced satisfactory results in accordance with set tolerances as recommended by AAPM TG 128. Using an ultrasound brachytherapy quality assurance phantom, average axial distance measurement of 10.11 ± 0.11 mm was estimated. Average lateral distance measurements of 10.08 ± 0.07 mm, 20.01 ± 0.06 mm, 29.89 ± 0.03 mm and 39.84 ± 0.37 mm were estimated for the inter-target distances corresponding to 10 mm, 20 mm, 30 mm and 40 mm respectively. Volume accuracy assessment produced measurements of 3.97 cm3, 8.86 cm3 and 20.11 cm3 for known standard volumes of 4

iv

cm3, 9 cm3 and 20 cm3 respectively. Depth of penetration assessment of the ultrasound system produced an estimate of 5.37 ± 0.02 cm, indicating the system’s ability to visualize low contrast objects 5.4 cm into a patient.

PET-CT system’s performance evaluation also produced satisfactory results in accordance with set tolerances as recommended by IAEA Human Health Series 1. Computed tomography laser alignment test ensured that all CT gantry lasers were properly aligned with the patient bed. Image display width test ensured that volume of patient or organ being measured and displayed was equivalent to that selected on the CT scanner console, to a deviation of ± 1 mm. Results from CT image uniformity test showed that mean CT numbers in peripheral regions of interest deviated from the central mean to within recommended tolerance level of ± 5 HU, indicating a good level of uniformity. Computed tomographic dose indices for head and body phantoms were estimated as 44.30 mGy and 20.08 mGy, comparative to console displayed doses of 42.40 mGy and 19.49 mGy respectively. Registration accuracy for PET-CT images was to have displacements of less than 1 mm in x, y and z directions. Image quality of PET-CT images was performed to produce images simulating those obtained in a total body imaging study involving both hot and cold lesions. Percentage contrast estimates of 49.3% and 52.6% were obtained for hot spheres of diameters 1.3 cm and 2.2 cm respectively, while contrast estimates of 74.8% and 75.6% were obtained for cold spheres of diameters 2.8 cm and 3.7 cm respectively. The PET-CT system resolution was estimated as 0.5 ± 0.01 cm, indicating the system’s ability to image tumours of the size of about 5 mm.

Satisfactory results from the performance evaluation of ultrasound and PET-CT systems, paved way for them to be used in acquiring prostatic images for the study. Developed MATLAB image enhancement algorithm enhanced the quality of prostatic images before fusion. The algorithm

v

was developed by mapping the intensity values in raw images to new values in a modified image using imadjust function. Contrast enhanced prostatic images of ultrasound and PET-CT were then co-registered with developed MATLAB fusion algorithm. The fusion algorithm was developed on the theory of mutual information and rigid body transformation. Fused image of ultrasound and PET-CT in this study has been assessed to have well defined and good visualized prostate capsule, urethra and implanted seeds, which would otherwise not be the case in either of the two images separately. The resultant image could therefore produce much more accurate results in treatment planning of prostate cancer cases. Assessment of image registration error for the ultrasound-PET-CT fused image produced a root mean square error estimate of 1.3 mm.