Research On Extraction Of Hemodynamic Information From Liver DSA Sequences

Digital subtraction angiography(DSA), which reveals the vessels from surrounding tissue by injection of contrast agent, then subtract injected images from the un-injected image. Doctors can observe the morphology of vascular from subtraction images to diagnosis tumor or other pathological, but the flow density change with time cannot be observed intuitively and exactly from the images. In order to offer more accurate information for doctor diagnosis, this thesis studies the extraction method of flow dynamic information from the DSA sequences. The main work is as following:The DSA sequences which is DICOM format were analyzed, in order to research under laboratory conditions, useful data elements were read and their value were adjusted according to window width and window center value, then saved as BMP format. For the utilization of spine position information, window width and center values were adjusted to optimally show the spine region.A segmentation method which combined spinal symmetry and rough subtraction projection is proposed here to obtain angiography regions. The symmetry line was found by the spine’s gray symmetry; Binary image was segmented by threshold from rough subtraction images, processed using several morphological algorithms, and projected on vertical axis. Angiography regions were devised according to the line of symmetry and projection position region.Contrasting the commonly subtraction and registration method, this thesis proposes a local dynamic subtraction method. After selection of the region of interest, it was matched with corresponding regions in other frames automatically. The match method used the normalized histogram energy of subtraction images as the measurement standard of similarity. Vascular areas were segmented from every corresponding region of interest using OTSU threshold, then the flow density was obtained. This method can ensure accuracy and be more efficiency because it does not require registration the whole image.Lastly, T-D curve was drawn by the time, flow density values of every frame as the horizontal and vertical axis respectively. Time density curve can response flow density changing process of the sequence.Peak-valve, peak-time, K-max-decent, etc. seven parameters of the T-D curve were computed, and compared with the values calculated by artificial observation simulate a real diagnosis. The experimental result shows that this method can provide more accurate and credible flow dynamic information for diagnosis.