Increasing frame rate in ultrasound imaging by temporal morphing using tissue Doppler

The diagnostic value of ultrasound B-mode sequences of the human heart seems to diminish when the frame rate is low. This often enforces the operator to use a limited size of the scan in order to retain frame rate. The problem becomes more evident when doing duplex color Doppler and B-mode scanning, and increases even more in real-time 3D applications. An obvious way to improve the visual quality of a low frame rate B-mode display is to interpolate between successive frames according to the actual screen update frequency. In this work we used morphing based on velocity information to improve the quality of interpolated B-mode frames in sequences with low frame rate. Two different sources of information about tissue velocity were investigated: speckle tracking and Doppler signal processing. Based on a priori knowledge of the image, we calculated the velocity field in certain image locations throughout a cardiac cycle. Secondly, we extended the velocity information to the rest of each image frame to produce a velocity estimate for the entire data set. The velocity information was then used to create intermediate, morphed frames, were the frame rate could be chosen arbitrarily. We applied this technique to data previously used in diagnosis. The quality of the morphed data was evaluated by removing frames from duplex B-mode and tissue Doppler recordings, then replacing the removed frames with morphed ones. The decimation factor was chosen so that the frame rate in the remaining sequence was between 12 and 18 frames per second. The decimated and morphed sequences were compared to the original ones, both frame by frame and as synchronized cine-loops running in true time or in slow motion. Wall motion scoring is an evaluation of left ventricular wall motion based on the subjective impression of the endocardial echo inward motion towards the centre of the left ventricle and the degree of thickening of the myocardium. The wall motion score was performed on 16 segments of the myocardium for each patient, based on B-mode recordings from three standard cardiac views: apical 2-chamber, 4-chamber and long-axis. We investigated data from 20 different patients, 10 healthy with normal ventricle and 10 with pathology (myocardial infarction). The 60 data sets were object to wall motion scoring two different times: First as part of the diagnosis, then scored blinded again two years later by the same operator, but the last time with morphed data. The feasibility of the two scorings were compared: With original data, 3.4% of the segments were not possible to score due to poor gray scale visibility, while for the morphed data, the same applied for 5.3%. We found that 94% of the segments had identical scorings in the two cases. We conclude that much of the diagnostic value is retained in recordings with frame rates in the range 12 - 18 frames per second.