US 12,133,762 B2
Three-dimensional modeling and assessment of cardiac tissue
Jon Klingensmith, Carbondale, IL (US); Colin Gibbons, Glen Carbon, IL (US); Michaela Kulasekara, Glen Carbon, IL (US); and Vu Quang Dinh, Edwardsville, IL (US)
Assigned to Board of Trustees of Southern Illinois University, Edwardsville, IL (US)
Filed by Board of Trustees of Southern Illinois University, Edwardsville, IL (US)
Filed on Apr. 27, 2022, as Appl. No. 17/730,380.
Claims priority of provisional application 63/250,568, filed on Sep. 30, 2021.
Claims priority of provisional application 63/184,492, filed on May 5, 2021.
Prior Publication US 2022/0370033 A1, Nov. 24, 2022
Int. Cl. A61B 8/08 (2006.01); A61B 5/00 (2006.01); A61B 34/10 (2016.01); G06T 7/00 (2017.01)
CPC A61B 8/085 (2013.01) [A61B 5/7275 (2013.01); A61B 8/0858 (2013.01); A61B 34/10 (2016.02); G06T 7/0012 (2013.01); A61B 2034/105 (2016.02); G06T 2207/30048 (2013.01)] 4 Claims
OG exemplary drawing
 
1. A system for patient cardiac imaging and tissue modeling, comprising:
an echocardiograph configured to acquire patient 3D cardiac imaging data with raw 2D spectral cardiac imaging data;
a processor to receive the patient 3D cardiac imaging data with raw 2D spectral cardiac imaging data; and
software run by the processor to provide the patient 3D cardiac imaging data and the raw 2D spectral cardiac imaging data to a trained learning network to identify a left ventricle, a right ventricle and fat and generate a 3D patient cardiac image labelled with fat; and
a user interface and display to interact with the patient 3D cardiac image;
wherein the trained learning network was trained by obtaining training magnetic resonance imaging 3D cardiac imaging data and training echocardiograph 3D cardiac imaging data with training raw 2D spectral cardiac imaging data by using deep learning to find left and right ventricles in the training magnetic resonance imaging 3D cardiac imaging data and training echocardiograph 3D cardiac imaging data, by fitting a training 3D model to a labeled image volume including fat labels, and by registering the training raw 2D spectral cardiac imaging data with the training 3D model.