This technology is an AI-based software for denoising and reconstructing MR images.
Unmet Need: Fast, anatomically accurate and de-noised reconstruction of medical MR images
Currently the acquisition of magnetic resonance images (MRI) requires a long period of time for scanning and several steps for post-processing and reconstruction. The long scan time is a logistical burden for patients and medical professionals. Additionally, slow reconstruction time and poor accuracy quality impedes the speed and quality of diagnoses from the MR images. These factors necessitate the development of a unified software to rapidly reconstruct MR images with high anatomical accuracy.
The Technology: A deep learning-based MR image reconstruction and denoising software
This MRI denoising and reconstruction software is built using an Artificial Fourier Transform (AFT)-Net to implement deep learning-based reconstruction. This network has been trained on publicly available datasets in order to perform robust reconstruction on MR images from various sources with variable quality of acquisition.
This network has demonstrated equivalent performance compared to conventional reconstruction software for preclinical scans in mice.
Applications:
- Preclinical MR image reconstruction
- Preclinical MR image denoising
- Robust MR image reconstruction with variable contrast agent dosing and acquisition time & resolution
- Blood-brain barrier opening detection using a low dose of contrast
- Disease detection and anatomical evaluation of the brains of preclinical research subjects
Advantages:
- Unified and robust reconstruction
- Can be incorporated into existing deep learning networks
- Trained on various datasets to enable robust to variable acquisition and experimental parameters
- Enables lower doses of contrast agent and shorter scan times by denoising and reconstructing with high anatomical accuracy
- Enables repeated or longitudinal scanning due to reduction in scan time and necessary contrast agent dosing
Lead Inventor:
Jia Guo, Ph.D.
Patent Information:
Patent Pending
Related Publications:
- [Yang Y, Laine AF, Guo J. “Deep Learning-based MRI Reconstruction with Artificial Fourier Transform (AFT)-Net” Internatinoal Society for Magnetic Resonance in Medicine. 2023 Jun. 3-8](https://submissions.mirasmart.com/ISMRM2023/Itinerary/ConferenceMatrixEventDetail.aspx?ses=D-15
Tech Ventures Reference:
