RPI ID: 2020-047-201
Innovation Summary: A deep learning-based method is introduced to correct distortions in photon-counting X-ray imaging data. The system compensates for spectral distortions and pile-up effects that degrade image quality in photon-counting detectors. Neural networks are trained to reconstruct corrected images from raw detector outputs, improving diagnostic accuracy. The approach is compatible with existing photon-counting CT systems and supports real-time processing.
Challenges / Opportunities: Photon-counting detectors offer spectral imaging advantages but suffer from data corruption due to pile-up and charge sharing. This invention addresses those limitations using AI-driven correction, enabling broader clinical adoption. It opens opportunities for enhanced tissue characterization and reduced radiation dose. The method supports integration into commercial imaging platforms.
Key Benefits / Advantages: ✔ Corrects spectral distortions ✔ Enhances image quality ✔ Enables real-time processing ✔ Compatible with existing systems ✔ Supports low-dose imaging
Applications: • Medical imaging • Photon-counting CT • Diagnostic radiology
Keywords: #photoncounting #xrayimaging #deeplearning #medicaldiagnostics #spectralcorrection #radiology
Intellectual Property: US Issued Patent US12099152B2, 17/704520, filed 25-Mar-2022