We are developing cutting-edge AI software technology for data analytics and decision making for human health and well-being.
Deep Reasoning AI Inc. was founded by two recognized experts in the AI research community. Our Deep Reasoning Engine combines AI reasoning, deep learning, knowledge representation and optimization to solve complex tasks. We have initially applied our technology to the generation of personalized 3D digital anatomical models from medical images. We have fully automated the accurate production of 3D digital models in a few minutes compared to a manual process, which can take up to weeks of a skilled technician’s time; thereby, significantly reducing the time and the cost of production. Our work with the Mayo Clinic has confirmed that our generated models are highly accurate.
Our
Technology
The Deep Reasoning Engine is based on over ten years of research by our co-founders. By combining deep learning's ability to leverage unlabeled data with higher-level reasoning modules, the Deep Reasoning Engine can reason about prior knowledge and physical constraints and compensate for the scarcity of labeled data. The underlying AI technology was ranked fourth on the “Top 10 List of 2019 Coolest Science and Technology Advances” by the US Army Basic Research program.
Check out our 3D segmentation platform!Work by team members on the underlying deep reasoning networks was featured in the September 2021 issue of Nature Machine Intelligence magazine.
-
The Deep Reasoning 3D Engine rapidly and accurately generates patient-specific 3D digital anatomical models from CT and MRI scans. Studies show that 3D anatomical models can improve surgical outcomes by providing the surgeon with a better understanding of a patient’s anatomy and the ability to plan the surgical procedure in advance of the operation. Widespread adoption of 3D models, however, has been limited by current methods which require time consuming and expensive input from a technician.
-
One of the most time-consuming tasks in the process is image segmentation, which is done automatically by the Deep Reasoning 3D Engine and allows models to be generated in a few minutes. We believe the reduction in time and costs resulting from use of our technology would allow 3D models to be more readily available and not limited to only the most complicated medical cases.
-
The Deep Reasoning 3D Engine is also able to classify the components of the segmentation, i.e. classify and label each vertebra or different parts of a hip. Our work with the Mayo Clinic (a shareholder in the Company) has confirmed that our generated models are highly accurate, save weeks of expert time and would be helpful in surgical planning and medical training.
Deep Reasoning 3D
Engine uses
a multi-step
process consisting of 3 modules:
STEP 1:
Pre-
Processing
Input data is pre-processed to ensure the Deep Reasoning 3D Engine is robust to variations in:
Type of Scanner (works with scans from all the major manufacturers)
Patient Orientations
Scanner Resolutions (can process scans of any input resolution and provides 1 mm voxel segmentation resolution)
Multi-Organ
Segmentation
STEP 2:
Image
Segmentation
The creation of 3D anatomical models requires the extremely tedious task of segmenting the anatomy of interest on a slice-by-slice basis for every CT/MR volume image.
This task requires expert analysis and time and can take days to weeks for a single patient anatomy.
The Deep Reasoning 3D Engine uses artificial intelligence to automate this process and completes it in a few minutes.
Multi-Organ
Segmentation
Refinement of individual segmentations via
anatomical
reasoning and 3D rendering.
Refinement of joint segmentations via
anatomical
reasoning and 3D rendering.
STEP 3:
3D Rendering
- The Deep Reasoning 3D Engine uses artificial intelligence to automate the rendering of the segmented image into a 3D model.
- The task involves the refinement of individual and joint segmentations using anatomical reasoning and 3D rendering techniques.
THE TEAM
Human
Intelligence
We collaborate with Mayo Clinic, a shareholder in our company and a globally renowned medical institution. Specifically, we work closely with Mayo’s 3D Anatomic Modeling Lab, which is recognized worldwide as a leading institution in 3D medical anatomical modeling.