Anatomically-Informed Deep Learning on Contrast-Enhanced Cardiac MRI for Scar Segmentation and Clinical Feature Extraction

Case ID:
C16606
Disclosure Date:
10/15/2020

Unmet Need

Heart disease is the leading cause of death worldwide with 30.3 million Americans diagnosed with heart disease in 2018, according to the CDC. The leading diagnostic method is to identify myocardial fibrosis (scarring), which is a leading indicator of sudden cardiac death (SCD). Current image segmentation involves cardiac magnetic resonance (CMR) imaging with contrast enhancement (i.e. late gadolinium LGE). Identifying scarring through image segmentation is challenging, as the process can be manual and tedious with low generalizability. There is a strong need for a more automated method of identifying scarring to limit the burden on physicians and technicians and more accurately aid preventative identification markers of SCD.

Technology Overview

Researchers at Johns Hopkins have developed a novel Anatomical Convolutional Segmentation Network or “ACSNet” technology, which is a fully automated approach that can identify myocardium and fibrosis in LGE-CMR images. This technology is capable of improving the image data leading to enhanced anatomical accuracy and proper handling of ambiguous regions such as the apex and base. The invention additionally enables clinical use by ensuring robust applicability across different CMR modalities, scar distributions arising from different heart pathologies, and different imaging centers. Furthermore, ACSNet automatically extracts clinical features that aid a physician in diagnosing diseases and decision-making factors.

Stage of Development

Experimental data is available with graphic user interface in development.

Publication

N/A

Patent Information:
Title App Type Country Serial No. Patent No. File Date Issued Date Expire Date Patent Status
ANATOMICALLY-INFORMED DEEP LEARNING ON CONTRAST-ENHANCED CARDIAC MRI FOR SCAR SEGMENTATION AND CLINICAL FEATURE EXTRACTION PCT: Patent Cooperation Treaty United States 18/249,583   4/19/2023     Pending
Inventors:
Category(s):
Get custom alerts for techs in these categories/from these inventors:
For Information, Contact:
Lisa Schwier
lschwie2@jhu.edu
410-614-0300
Save This Technology:
2017 - 2022 © Johns Hopkins Technology Ventures. All Rights Reserved. Powered by Inteum