Survey
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Real-Time Kymogram Detection from Cardiac Spiral CT Scans Dirk Ertel Dipl.-Ing., Marc Kachelrieß Ph.D., Dirk-Alexander Sennst Ph.D., Stephan Achenbach M.D., Willi A. Kalender Ph.D. Institute of Medical Physics, University of Erlangen-Nuremberg, Erlangen, Germany, www.imp.uni-erlangen.de Purpose Results To generate a rawdata-based synchronization signal for cardiac spiral CT scans in real time. The signal may be used as an alternative to the ECG acquired at scan time and as a basis for cardiac phase-dependent on-line tube-current modulation. The already existing offline computed kymogram [1] is taken as the base for these enhancements. 9 of 12 patients showed a high correlation with the ECG to the kymogram where the phase lag value stayed below 25% for all three methods. The least reliable results were generated by the derivation of the projected mass with respect to the view angle. The prerequisite of being able to perform in real time was satisfied by all three methods. Methods and Materials Fig. 6 shows examples of coronal CT images using phase-correlated reconstruction. The quality of the different images point out the usability of the different synchronization signals. In contrast Fig. 7 shows coronal CT images with the usage of the predicted sync peaks derived from the different kymograms. It has to be mentioned that the predicted signal will be used for a phase-dependent tube current modulation and not directly for an image reconstruction. The synchronization signal –the motion or kymogram function– is generated by analyzing the periodic temporal variation of the mass distribution m(t) of the current section and correlates to the heart rate. Three different methods are used to generate the kymogram: A two dimensional center-of-mass tracking of the scanned object [1] (Fig. 1), the difference of the projected mass after a half rotation [2] (Fig. 2) and a derivation of the projected mass with respect to the view angle [3] (Fig. 3). Fig. 4: ECG signal and kymogram signal (COM) from patient 310179 For all three methods the data were integrated over several detector rows. The data of 12 patients scanned with a collimation of 12×0.75 mm, a table increment of 2.8 mm/rotation, a rotation time of 0.42 s and a concurrent ECG recording (Sensation 16, Siemens, Forchheim, Germany) was used for validation. The signal extrapolation was performed by predicting the heart rate according to the values observed so far with the normalized least-mean-square algorithm. The sync peak prediction shall allow for real-time performance. The degree of correlation is expressed by the value of the standard deviation σ of the lag function L(t), which is the difference of the absolute phase of the ECG and the kymogram (Fig. 5). COM(t) ξ c,2 ECG COM ∆M π0,p ∆ϑ ξ c,1 Fig. 6: Coronal CT images of patient 310179, phase-correlated reconstruction, based on kymogram (C=0 HU/W=1000 HU) Fig. 1: Two dimensional center-of-mass tracking of the scanned object: COM(t) [1] 0 π ,p m(t1) Heart area predicted signal: COM m(t 1+∆ t) predicted signal: ∆M π0,p Fig. 5: Lag functions and heart area from patient 310179 Fig. 2: Difference of the projected mass after half a rotation: ∆Mπ 0,p (t) [2] m(t3) m(t2 ) The phase lag value σ is computed for the ROI covering the heart area. A high correlation with the ECG leads to small values of σ. The time needed for signal processing was determined with a standard PC to verify the real-time capability of the kymogram calculation. Table 1 shows averaged values of σ for 12 patients. Here the correlation of the different kymogram methods and of the derived predicted sync peaks to the ECG can be seen. m(t1 ) predicted signal: ∆ϑ Fig. 7: Coronal CT images of patient 310179, phase-correlated reconstruction, based on predicted sync peaks (C=0 HU/W=1000 HU) Conclusion σ (kymogram) σ (prediction) COM 24.76 31.52 0 ∆M π ,p 32.66 40.57 ∆ϑ 54.55 48.57 Method Fig. 3: Derivation of the projected mass with respect to the view angle: ∆ϑ (t) [3] Three different methods for generating a kymogram Table 1: Averaged values of σ [in %] for L(t) of 12 patients For most patients predicted sync peaks on the basis of a kymogram correlated well with the ECG. We conclude that a kymogram-based tube current modulation for dose reduction in cardiac CT appears possible. References [1] Marc Kachelrieß, Dirk-Alexander Sennst, Wolfgang Maxlmoser, and Willi A. Kalender (2002). “Kymogram detection and kymogram-correlated image reconstruction from subsecond spiral computed tomography scans of the heart”. Medical Physics, 29(7):1489-1503 [2] H. Bruder, E. Maguet, K. Stierstorfer, T. Flohr “Cardiac spiral imaging in Computed Tomography without ECG using complementary projections for motion detection” Proc. SPIE, 5032:1798-1809, May 2003. Medical Imaging 2003: Image Processing [3] Dirk Ertel (2004). “Real-Time Kymogram Detection from Spiral CT Scans of the Heart” Diplomarbeit, Friedrich-Alexander Universität, Institut für Medizinische Physik and Lehrstuhl für Multimediakommunikation und Signalverarbeitung