Biomedical Ultrasound and Electromagnetic lab

FCM based extraction of heart with motion compensation for

accurate estimation of tracer perfusion

Introduction

To measure the blood flow in different regions of the heart, tracers are used. The tracer is injected in the blood stream that reaches the heart. When the PET scanning is conducted the tracer appears to be of high intensity and other regions appear to be of low intensity. Using this phenomenon the flow of the blood can be measured based on the wash in and wash out curves obtained from tracer intensities. If a region stays dark throughout, it shows that the blood has not reached that part of the heart. During the process of image acquisition different kind of movements induce calculation errors along with the decrease in image resolution. Once you get the complete data, it has to be made sure that you are corresponding same portions of heart in different images taken at different times.

Motion Compensation

During a beat cycle, heart not only expands and contracts but it also rotates along the axes. To incorporate this motion in the parameter estimation, a compensation technique has to be applied to nullify the effect of these motions. For the rotation along the two axes, instead of calculating a global rotation factor for the alignment along y-axis, different rotation angles are calculated for different time bins. In this way at all times the heart is exactly aligned no matter in which state of a beat cycle it was in. For expansion and contraction, the biggest size is selected at a certain time and all others are interpolated to make the size equal to the biggest one. If this is not done, you can end up comparing a sector of heart with a black portion in any other time bin.

Processing of Data

FCM

Once the raw data is acquired, it needs to be processed to extract the heart from all sorts of noise. For that image is passed through the Fuzzy clustering routine that sifts out the major clusters from the image, the heart being one of those.

Masking

To remove the unwanted clusters, masking is applied. In that All the slices in one time instance are added together but with giving higher weight to the higher intensities to broaden the gap between intensities level and enhance the contrast. This process brighten up the heart region far more that the other smaller and random clusters. This brightest cluster is then selected by adjusting a proper threshold. After this process we get the heart extracted out (almost) completely.

Rotation

Once the heart is extracted it is aligned to one of the axes. This is done using the projections on the plane. Due to the elliptical shape of the heart, it can only be aligned to y-axis if its projection on x-axis is smallest. In this way it can be made parallel to y-axis.

Polar Map

Polar map is a standard way of arranging the slices of heart in one image. With polar maps made for each time slice each corresponding sector can be compared to estimate the tracer curve.

Parameter Estimation

The model we are working on is proposed in a paper by Gary Hutchins. It is governed by the following three rate parameters

Using the above model the perfusion parameters are calculated and the maps for all these parameters is shown below.