Automatic blood vessel segmentation for quantifying blood oxygenation
At present, delineation of blood vessels within the feto-placental unit can only be conducted manually by highly trained personnel. This is due to the suboptimal signal-to-noise ratio and image contrast, which are restricted due to the limitations on the allowable energy delivered to the human body during MR imaging. Furthermore, the lack of image contrast, coupled with small fetal blood vessels and variation in fetal position can complicate the interpretation of the anatomy. Manual identification of these vessels is very time consuming and highly observer-dependent, making large clinical trials currently unfeasible. Creating an automatic segmentation tool for the blood vessels of interest will increase the accuracy and reproducibility of measurements, thereby also enabling clinical investigations to detect minute changes in blood oxygen saturation and permitting earlier diagnosis of deteriorating cardiovascular physiology. We aim to create an automatic segmentation algorithm for identifying blood-vessel ROI. The algorithm will use anatomical landmark detection to predict the location and size of the blood vessel of interest. This process will be repeated for each echo time, creating a time series of ROIs for which the signal intensity will be calculated. A plot of these intensities over time and their exponential fit will yield the corresponding T1 and T2 of the blood, to be used in calculating the oxygen saturation.
Student is responsible to develop a segmentation algorithm for identifying blood-vessel ROI. For the segmentation, anatomical landmark detection should be utilized to predict the location and size of the blood vessel of interest.
Excellent knowledge and experience of computer programming in MATLAB, C++, Java or any other programming languages.
Javad Alirezaie : Automatic blood vessel segmentation for quantifying blood oxygenation | Thursday March 30th 2017 01:06 PM