WO1994004977A1

WO1994004977A1 - Reference image sequence recording for interference signal elimination and filtering in medical image sequences

Info

Publication number: WO1994004977A1
Application number: PCT/DE1993/000702
Authority: WO
Inventors: Peter Strobach
Original assignee: Siemens Aktiengesellschaft
Priority date: 1992-08-11
Filing date: 1993-08-06
Publication date: 1994-03-03

Abstract

When taking X-ray image sequences in the heart area, the relevant objects in the image sequence are characterized by rhythmic movements determined by the heart activity. An effective interference filtering process for such image sequences uses this preliminary knowledge about these movements whose repetition is synchronous with the pulse of the patient. This process may generally be used to filter out interferences from other medical image sequences as well, in which movements synchronous with the heartbeat of the patient are detected.

Description

Image sequence reference storage method for interference signal suppression and filtering out of medical image sequences

The noise in X-ray image sequences increases with a decreasing X-ray dose. Therefore, efficient methods for noise filtering are required. They enable a reduction in the dose rate during the examination and thus a reduction in the radiation exposure to the patient and operating personnel.

Noise filtering of X-ray image sequences has already been intensively investigated in the literature. The previously known methods for noise filtering are based on the one hand on a motion-compensated averaging of several images over time (TS Huang and YP Hsu, Image Sequence Enhancement, In TS Huang, ed., Image Sequence Analysis, pages 290-310, Springer Verlag, 1981), on non-linear filtering in the single image, or on the method of background extraction with a Kalmann filter (A. v. Brandt and KP Karmann, "A method for the detection and tracking of moving objects in image sequences with a stationary background", US patent 5034986). In all of these processes, no specific inistic knowledge of the type of object movements is included in the process approach. In examinations in the area of the heart, however, the movement of objects of interest (eg catheters) in the image is essentially a result of a synchronous movement of the pulse of the

Patient-dependent course marked. The resulting deterministic in the course of the object movements has so far not been taken into account in any of the known methods for noise filtering of X-ray image sequences. The invention is therefore based on the object of specifying a new method for noise filtering in medical image sequences in which synchronous movement sequences are visible to the heartbeat, which takes into account the sequence of object movements given by the deterministics of the heartbeat.

This object is achieved by a method for filtering noise in medical image sequences with movement sequences synchronous with the heartbeat with features according to claim 1.

With this method, an image sequence is recorded in which a movement sequence synchronous with the heartbeat is visible. An electrocardiogram with R waves is recorded synchronously. The image sequence is segmented according to the R waves of the electrocardiogram using a trigger method. The images of a pulse period of the electrocardiogram are corrected for movement and accumulated in an R-wave synchronized manner in an image sequence reference memory.

The method is particularly suitable for use in the noise filtering of X-ray image sequences, in particular when recording the heart.

Advantageous further developments of the method result from the subclaims.

Fig. 1 shows the basic procedure for the execution of the method in a schematic representation.

Fig. 2 shows a signal flow diagram of the operations necessary to carry out the method for each picture in the picture sequence. The method is described in more detail below with the aid of a preferred exemplary embodiment and with the aid of the figures.

When recording X-ray image sequences in the area of the heart, the objects of interest in the image sequence are subject to a rhythmic movement determined by the cardiac activity. Using this prior knowledge of the movement sequence, which is repeated essentially synchronously with the patient's pulse, an efficient method for noise filtering can be developed. The patient's EKG is recorded in sync with the actual image sequence. It provides information about the phase of the heartbeat associated with each image. The image sequence is segmented according to the R waves of the EKG using a trigger procedure. Finally, a motion-corrected and R-wave-synchronous accumulation of the images of a pulse period takes place in an image sequence reference memory. Over time, a noise-compensated image sequence with the object movement typical for a heartbeat builds up in this reference memory. This noise-compensated image sequence is again periodically read out with the heartbeat and adapted to the current image sequence.

In principle, this method can be applied to all medical image sequences in which an object movement synchronous to the heartbeat takes place in the image area. The method is particularly suitable for noise filtering of X-ray image sequences during examinations in the area of the

Heart with special consideration of object movements caused by cardiac activity.

In the field of interference compensation biomedical zinischer Signals has developed a powerful procedure using the prior information on the deterministics of the heartbeat (German patent application P AI 26 46 3.0). In the process of this patent application, compensation becomes quasi-periodic

Disturbances in measurement signals separate the useful signal from the interference signal by estimating a sequence of trigger times from an auxiliary signal (blowout and an interference signal template from this measurement signal, in which a trigger-synchronous reference signal is also generated from this interference signal template, and this trigger-synchronous Re Some principles of this method can be applied to the problem of noise filtering of x-ray image sequences and other medical image sequences. This leads to the method according to the invention which, through additional evaluations of the electrocardiogram, provides patients with heart rate-synchronous noise filtering and thus one Inclusion of the deterministics of the object movements in the noise filtering process enables.

The patient's EKG is recorded in sync with the actual image sequence. It provides information about the phase of the heartbeat associated with each image. The image sequence is segmented according to the R waves of the EKG using a trigger procedure. Finally, a motion-corrected and R-wave-synchronous accumulation of the images of a pulse period takes place in an image sequence reference memory. Over time, a noise-compensated image sequence with the object movement typical for a heartbeat builds up in this reference memory. This noise-compensated image sequence is then periodically read out with the blow and adapted to the current image sequence, which ultimately results in the actual noise-filtered output image sequence. Fig. 1 shows the basic procedure for building the

Image sequence reference memory: An ECG recorded synchronously to the image sequence is segmented according to the R waves. The images of a segment are entered in an image sequence reference memory. As soon as the next segment limit is reached, the frame counter of the frame sequence reference memory is reset. The images in the next segment are now added motion-compensated to the image content already in the image sequence reference memory.

2 shows a signal flow diagram of the operations required for each image in the image sequence. A cumulative image in the image sequence reference memory is adapted by motion compensation to an image of an input image sequence that is segmented in phase with the ECG. Including an exponential forgetting factor, the input image and the movement-compensated cumulative image are now added. The result of this operation is an object-oriented updated cumulative image which is again entered in the image sequence reference memory (which has been updated with it). At the same time, the updated cumulative images arrive at the exit of the method and form the result of the noise-compensated images there.

Claims

Claims:

1. Method for filtering noise in medical image sequences with movement sequences synchronized with the heartbeat, with the following steps:

a) an image sequence is recorded in which a movement sequence synchronous with the heartbeat is visible;

b) an electrocardiogram with R waves is recorded synchronously;

c) the image sequence is segmented according to the R waves of the electrocardiogram using a trigger method;

d) the images of a pulse period of the electrocardiogram are motion-corrected and R-wave-synchronized accumulated in an image sequence reference memory.

2. The method according to claim 1 for noise filtering in medical X-ray image sequences.

3. The method according to claim 2 for noise filtering in medical X-ray image sequences of the heart.

4. The method according to any one of the preceding claims, in which the accumulated images are read out periodically with the heartbeat from the image sequence reference memory.

5. The method according to claim 4, wherein the read out images are adapted to a current image sequence.

6. The method according to any one of the preceding claims, in which a movement correction is carried out by object matching.