US 11,810,654 B2
Method and system for automatically generating a section in a radiology report
Jeffrey Chang, Berkeley, CA (US); Doktor Gurson, Berkeley, CA (US); Eric Purdy, Berkeley, CA (US); Brandon Duderstadt, Berkeley, CA (US); Jeffrey Snell, Berkeley, CA (US); Andriy Mulyar, Berkeley, CA (US); and Deeptanshu Jha, Berkeley, CA (US)
Assigned to RAD AI, Inc., Berkeley, CA (US)
Filed by RAD AI, Inc., Berkeley, CA (US)
Filed on Apr. 20, 2022, as Appl. No. 17/725,031.
Application 17/725,031 is a continuation of application No. 17/020,593, filed on Sep. 14, 2020, granted, now 11,342,055.
Claims priority of provisional application 62/900,148, filed on Sep. 13, 2019.
Prior Publication US 2022/0246258 A1, Aug. 4, 2022
This patent is subject to a terminal disclaimer.
Int. Cl. G16H 15/00 (2018.01); G16H 20/40 (2018.01); G16H 40/63 (2018.01); G16H 50/20 (2018.01); G16H 10/60 (2018.01); G16H 40/67 (2018.01)
CPC G16H 15/00 (2018.01) [G16H 10/60 (2018.01); G16H 20/40 (2018.01); G16H 40/63 (2018.01); G16H 40/67 (2018.01); G16H 50/20 (2018.01)] 18 Claims
OG exemplary drawing
 
1. A method for automatically generating an impression section of a radiology report, the method comprising:
receiving:
a radiologist identifier associated with a radiologist; and
a set of finding inputs;
with a trained machine learning model comprising a set of encoders and a set of decoders:
determining a radiologist style matrix based on the radiologist identifier, wherein the radiologist style matrix is determined based on a length metric associated with a set of impression sections in a set of manually generated radiology reports, the set of manually generated radiology reports generated by the radiologist;
referencing an ontology database to determine a set of concepts associated with the set of finding inputs;
determining, with the set of encoders, a context matrix based on the set of finding inputs and the set of concepts;
generating, with the set of decoders, the impression section of the radiology report based on the context matrix and the radiologist style matrix, wherein the impression section is configured to mimic a writing style of the radiologist;
retraining the trained machine learning model based on the generated impression section of the radiology report to determine a retrained machine learning model; and
with the retrained machine learning model, automatically generating a second impression section of a second radiology report associated with the radiologist.