Radiation-free synthetic CT for transcranial ultrasound
This technology is a multi-task 3D Mamba model that generates non-ionizing synthetic CTs from Ultrashort Echo Time/Zero Echo Time MRI scans for transcranial focused ultrasound planning.
Unmet Need: Non-invasive brain imaging modalities
Traditional computed topography (CT) imaging produces high amounts of radiation and associated cancer risks, limiting its use for preclinical planning and brain imaging diagnostics. The complexity and variation in skull anatomy requires non-ionizing, non-invasive brain imaging techniques to accurately assess patient anatomy. Machine learning and deep learning algorithms have demonstrated promise generating synthetic CTs (SynCT) as an alternative, non-invasive approach.
The Technology: Deep learning model for generation of non-ionizing synthetic CT for transcranial ultrasound
This technology describes a deep learning model that generates SynCTs from MRI scans. Using a multi-task model, Ultrashort Echo Time/Zero Echo Time MRI scans are processed with a Mamba state-space architecture. This algorithm integrates 3D spatial context and outputs CT intensity maps and skull segmentation in a single step, enabling a radiation-free modality for brain imaging and preclinical planning.
Training and validation on human and non-human primates demonstrated efficient processing, strong image fidelity, and cross-species generalizability.
Applications:
- Radiation-free preclinical planning for diagnostics
- Radiation-free transcranial ultrasound planning
- Radiation-free imaging for targeted neuromodulation therapies
- Preclinical research in non-human primate models
- Patient-specific acoustic simulation
Advantages:
- Radiation-free
- Cost-efficient
- Generates accurate synthetic CTs with limited datasets
- Leverages multi-task deep learning approaches
- Utilizes Ultrashort-echo time (UTE) and Zero-echo time (ZTE) MRI imaging
Lead Inventor:
Related Publications:
Tech Ventures Reference:
IR CU26245
Licensing Contact: Jerry Kokoshka