Study Of Key Techniques On Electrocardiogram Printouts Digitizing

In the procedure of clinic diagnosis, electrocardiography (ECG) was relied heavily byhospital to detect the electrical activity of human’s heart. Unfortunately, the recording paper isheat-sensitive paper, it is easily broken and the line of ECG waveform is fading with time.Cause of clinical diagnosis value and fast development of HIS (Health Information System),ECG paper recordings are more precious. These ECG data were necessarily to be extractedand keep the ECG information as digital type for clinical information sharing, onlinediagnosing and ECG database establishing.Typical ECG printout is composed of ECG black waveforms, lead characters, andestablished voltage, these information are appeared on the background with horizontal andvertical red gridlines. ECG printouts digitization is not the procedure of converting ECGpaper to image with the help of scanner or digital camera, but it is the intelligent processingtechniques to extract the wave data from ECG printouts.Firstly, Hough transforming was used to detect the edge between ECG paper andbackground, and then the slant was adjusted by transforming the coordinate system, thismethod can effectively correct the ECG tilt.Moreover, an improved k-means method based on Sobel_MCO (Morphological ClosedOperation) was presented to extract the waveform, and to realize effective separation betweenwaveform and complex background. This method used Sobel_MCO as preprocessing, andretained pixels belonging to the waveform as much as possible, and then determines theclustering center according to the2-D histogram (gray-position). After the iteration ofimproved k-means, the wave curve was extracted. This method increases the extractionaccuracy rate and reduces time complexity of traditional k-means. Based on this, the ECGwave data extraction and background reconstruction method were given. At last, because ofhigh medical value of ECG, the wave data extracted demands high accuracy.Considering the ECG form, the paper put forward a new evaluation method. It piecewised the waveform according to the characteristic parameters first, and then evaluate twowaveforms by the similarity of inclined angle between segments. This method has obviousphysical meaning and does not sensitive to translation and rotation.The paper tries to solve some key problems in ECG printouts digitizing, and a scientificevaluation strategy was established. In this study direction, some necessary research workswere completed.