Learning-assisted multi-modality dielectric imaging

US11250601B2 · kind B2 · utility

Assignee

• University of Southern California · US

Inventors

• Guanbo Chen · Allen, US
• Mahta Moghaddam · Los Angeles, US
• Pratik Shah · Bengaluru, IN
• John Stang · South Pasadena, US

Key dates

• Filing date: Apr 3, 2020
• Grant date: Feb 15, 2022
• Expiry date: Jul 7, 2040

Classification

• Technology area (CPC G): Physics
• CPC primary: G06T2207/30004
• WIPO field: Medical technology
• WIPO sector: Instruments

Abstract

A Convolutional Neural Network (CNN) assisted dielectric imaging method is provided. The method used CNN to incorporate the abundant image information from Magnetic Resonance (MR) images into the inverse scattering model-based microwave imaging process and generate high-fidelity dielectric images. A CNN is designed and trained to learn the complex mapping function from MR T1 images to dielectric images. Once trained, the new patients' MR T1 images are fed into the CNN to generate predicted dielectric images, which are used as the starting image for the microwave inverse scattering imaging. The CNN-predicted dielectric image significantly reduces the non-linearity and ill-posedness of the inverse scattering problem. The application of the proposed method to recover human brain dielectric images at 4 mm and 2 mm resolution with single-frequency and multi-frequency microwave measurements is provided.