Biography
Alpaslan KOÇ is a PhD student in the Institute of Biomedical Engineering in Boğaziçi University. He is also a lecturer in the Vocational College of Health Services in the Kırklareli University. His research areas are related with the Nuclear Medicine Imaging Instrumentation, Optimization, and Radiotherapy Planning.
Abstract
AbstractrnAccurate estimation of tumor volume on Positron Emission Tomography (PET) images is an important requirement in radiotherapy planning and other similar applications in Oncology. Tumors with less than 2 cm diameters are particularly difficult to quantify mainly due to the limited spatial resolution of PET cameras and the large pixel sizes. The purpose of the present work is to present a new method for the accurate volume estimation of small tumors and demonstrate its efficacy on real oncological PET images that contain simulated tumors. First, images in three dimensional form were reduced to small volumes incorporating the individual tumors in order to increase the efficiency of the image restoration process and reduce the effects of the varying point spread function (PSF) across the FOV. A blind deconvolution algorithm based on the Lucy-Richardson method was used to restore the cropped volumes by reducing the effect of the resolution loss. Parameters were optimized by the use of a relative difference index. The resultant image was then resampled using interpolation methods in order to reduce the pixel size. The tumor in the final image was delineated with the 50% thresholding method. The entire procedure was applied to the [18F] FDG PET images in the ONCOPET database. The ONCOPET image database consists of image sets of 128x128x375 voxels. Real patient scans were used and tumors of different sizes were simulated with spheres of varying activities. Finally, projection data were reconstructed using the fully AW-OSEM 3-D algorithm using six iterations and 16 subsets with a Gaussian isotropic postfiltering of 8 mm. The MIPAV image processing platform was used for visualization, cropping and segmentation. The volume estimation error was reduced from 583 % to less than 9 % for small tumour of 14 mm diameter and a signal to background ratio of four. First results indicate that a substantial improvement in accuracy of small tumor volume estimation can be achieved by the use of this method. This may contribute to higher precision in radiotherapy planning and other oncological applications using PET.rn
Biography
Adem Cihan Arslan has completed his MS at the age of 25 years from Bogazici University and he started Phd studies at Bogazici University Biomedical Engineering. He is also teaching assistant at Bogazici University Mathematics Department.
Abstract
Accurate detection and characterization of lung nodules is an unresolved issue. Although chest CT has been demonstrated to increase diagnostic accuracy, it suffers from several disadvantages such as higher costs, increased radiation exposure and a large number of false positives which may result in longer reading periods and unnecessary invasive interventions. Improving CAD techniques on chest radiography has been proposed as a potential method for improving lung nodule detection and characterization [1-5]. Once a lung nodule is detected, then the next step is to assess the nature of the lesion.The purpose of this work was to develop a lung nodule characterization scheme on bone subtracted images and evaluate its performance on a public database.