Image registration

[nagdawi]

Hi everyone,

I am Ph.D. student try to implement image registration for a medical multimodal image pair. I am looking to solve a new problem may lead to enhance diagnoses of any disease.

Many algorithms are developed to register medical images, such as brain MRI with PET.

As a radiologist, what images pair was taken by different sensors are better to register and it may help to reach a new level of disease diagnosis accuracy?

Thanks,

---

Members don't see this ad.

 

[JoshSt]

What you describe already exists for applications between different modalities where it is marginally useful.

It’s most helpful with nuclear medicine studies, such as PET, and anatomic Imaging not acquired as part of the PET CT, such as how you mentioned fusing PET or another NM study with brain MR.

Some places find fusing PET with pelvic MR for rectal or cervical cancer patients.

I don’t remember the name of the program we use offhand, but robust systems already exist. I think if you did this, you’d be re-inventing the wheel for something that mny radiologists find not helpful to only marginally helpful

More and more academic centers are getting PET MR systems and they’re still working out some bugs with coregistering the data. This would likely be the most clinical important research you could do, if your center has a PET MR system.

 

[maxxor]

OP should check out DynaCad (multi parametric prostate MR analysis tool). You can fuse diffusion / DCE etc all to whatever random MR sequence. Maybe something similarly powerful for liver MR would be useful. I particularly like the DCE viewer where you can simply page through the contrast phases while on the same image. Pretty nifty.