SBIR Phase I: AI-Powered Low-dose, Low-cost, High-Quality CT imaging

The broader impact of this Small Business Innovation Research (SBIR) Phase I project is to enable local (and global) access to high-quality, low-cost, and low-radiation exposure three-dimensional (3D) computed tomography (CT) imaging using existing 2D equipment. Examples include walk-in clinics, lung cancer screening centers, rapid stroke assessment centers, mobile platforms (e.g., ambulances), battlefield hospitals, and clinics in rural and underserved areas.

This advance will result in greater public access to advanced healthcare and should result in substantially lower healthcare costs. For example, moving complex spine surgeries from hospitals to local ambulatory surgical centers (ASCs) can save payors $10B in costs annually. The ASCs will also benefit; a relatively low percentage of complex monthly procedures can double their profits.

Patient satisfaction should improve by moving more complicated spine surgical procedures to smaller ASCs closer to home with fewer infection risks. The useful life of legacy X-ray systems will be extended following conversion to 3D, thereby reducing waste and landfill space. Beyond medicine, the project technology has widespread applications in nondestructive testing, from manufacturing to failure analysis/prevention to archaeology and art! All these advantages will enhance US competitiveness.

This Small Business Innovation Research (SBIR) Phase I project will enable simple, small-footprint, mobile, two-dimensional (2D) X-ray imaging systems to generate three-dimensional (3D) computed tomography (CT) images at low cost, with one-third of the X-ray dose of a conventional CT scan. The project combines recent advances in imaging physics with artificial intelligence (AI) to overcome the limitations of current CT image acquisition.

This contrasts with conventional AI-based CT de-noising (image cleaning) algorithms that function only in the image domain with no physics input. The project has three primary research objectives. First, enhance deep learning-based image reconstruction's ability to produce high-quality images from limited data.

Second, devise real-time geometric calibration methods to overcome mechanical instabilities inherent to simple X-ray systems. Third, develop high speed and high image fidelity data transfer methods to interface existing hospital imaging systems to the project computing platform while maintaining FDA and HIPPA compliance and avoiding disruption of hospital workflow.

Successful development of the three core technologies described will be used to create a minimum viable product (MVP). Medical practitioners will use the MVP to evaluate the technology and refine the features needed for a clinical product.

This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.