Transcript IMAGE REGISTRATION AND DYNAMIC IMAGING IN PACS

IMAGE REGISTRATION AND
DYNAMIC IMAGING IN PACS
Igor Vujović, Ivica Kuzmanić, Maja Krčum
INTRODUCTION
PACS systems are increasingly developing every day. They are not
invented to be useful aid to medical experts, but to serve in reducing
space for archive purposes, and in transmitting medical images to the
distance, they are, in combination with HIS, becoming a basic part of
expert systems in development.
The problem in processing the images and the possible cause of
misdiagnosis in the future expert systems could be SNR
Is the SNR the right criteria for the quality of the image?
IMAGE RECONSTRUCTION
Nevertheless every day’s improvements, some questions still remain:
1) can we use a prior knowledge about image structure or function to design better
systems of computer-aided diagnostics and PACS (if we are looking for a
fibrosis in X-ray, or anomaly in bones, there is a prior knowledge of what are we
looking for, where it could be found),
2) how can be clear image be reconstructed from nonuniformly sampled data,
3) can the resulting image be qualitatively and quantitatively characterized (can we
find a measure of the quality of the reconstruction).
DYNAMIC IMAGING AND
IMAGE REGISTRATION
In the dynamic imaging, a lot of images in series are acquisited from the
same anatomical site. Dynamic imaging is useful for functional MRI,
mammography and interventional MRI. Conventional Fourier imaging
methods require a tradeoff between spatial and temporal resolutions
(the Heisenberg's equation of uncertainty). Spatial and temporal
resolutions are (if N encodings are collected for each image):  x  1 / N  x
and T a cq = NT R . For the dynamic imaging (we need high speed), if a large
N is used to obtain high spatial resolution, temporal resolution will be
compromised. The key issue of the image reconstruction is how can we
effectively use the reference data for reconstruction of the dynamic
images so that data truncation due to encoding can be minimized.
EXAMPLE FROM PRACTICE
Figure 1 - Example of sharpening pulmonary X-ray: a) analyzed image, b) highly sharpen
image, c) normally sharpen image, d) enlarged detail of the histogram differences.
CONCLUSIONS
Whatever we do with matrix, the goal issue should not be SNR  3,
4 or saving memory and time, but to find ways of not changing
the medical information contained in the medical image 6, 8. It
is the simple question that should be answered: whether the
result of processing is the truth – or lie. Even a simple matrix
operation (A = A + B) leads to the possible loss of the real
medical information, because the original is changed. And it is
possible, in that simple operation, to lose the truth of the
patient's health.