Automated Generation of Sentence-Based Descriptors from Images

Case ID:
Disclosure Date:

Unmet Need
Radiological examinations of brain MR images can be solely based on subjective judgment utilizing radiologists' knowledge and expertise.  The reports generated by radiologists are free text, and are not structured, difficult to search, and hinder population-based analyses.

Technology Overview
Johns Hopkins faculty member, Dr. Susumu Mori, has developed a novel way to automatically generate human-understandable language, such as radiological reports, based on quantitative image analysis results.  This is a key technology that connects outputs from computer algorithms and inputs to the human brain.

Stage of Development
Dr. Mori maintains a large database of MR images created from past clinical cases, with an enabling cloud platform.  The proof of concept of this and related technologies is based upon a 419 patient study whereby patients were stratified in accordance with clinical diagnosis, and the data therefrom was analyzed. The software for this method is fully developed, and capable of providing machine-generated sentences that describe the change of structures in natural language.



Patent Information:
Title App Type Country Serial No. Patent No. File Date Issued Date Expire Date Patent Status
Automated Generation of Sentence-Based Descriptors from Images PCT: Patent Cooperation Treaty United States 15/771,030 4/25/2018     Pending
For Information, Contact:
Mark Maloney
Save This Technology:
2017 © Johns Hopkins Technology Ventures. All Rights Reserved. Powered by Inteum