I have a frontal and lateral chest x ray projection for a given patient as shown in the attached figures. I am thinking about performing pixel-level mappings across these views. What are the steps involved in this process? Do we have a standard method/code in Matlab/Python to perform image registration and pixel-level mapping? A detailed response is appreciated to this end. Thanks.
Hi,
You may follow one of these methods:
1. https://www.mathworks.com/help/images/registering-multimodal-3-d-medical-images.html
2. https://www.mathworks.com/help/images/registering-an-aerial-photo-to-an-orthophoto.html
In order to align, chest images where there is elastic/local deformation (i.e certain areas such as organs tend to deform greatly during inhalation and exhalation, while the bones no not), rigid (translation and rotation) and affine registration (rigid+scaling and shearing) may be helpful only as giving a rough connection between the corresponding features of the images. After doing rigid and affine registration, deformable registration must be applied in order to align these xray images. The best library for medical image processing is ITK.
https://itk.org/
Many medical applications are based on it, such as 3D sliced
https://www.slicer.org/
which is capable of doing both rigid and non-rigid image registration
Unfortunately this can not be done.
A pixel in a frontal X-ray image is a sum of all densities along the straight line of the pixel location in the detector, to the X-ray source. All these points on that straight line would be located on a more or less horizontal line in a lateral X-ray image, assuming the two X-rays were made simultaneously (probably not the case). Furthermore, the pixel values of that more or less horizontal line are determined by the density of all tissue in the plane defined by the line and the X-ray source.
So, a one to one correspondence of pixels in the two images does not exist!
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping