Monday August 5, 2019 0 comments
DENVER -- Eon announced the addition of two new applications to their product portfolio: EonEndu, a deep learning model powering the identification of patients at risk for future disease, and the expansion of EonDirect into Abdominal Aortic Aneurysm (AAA) patient tracking.
The company said EonEndu utilizes deep learning models that support Natural Language Processing (NLP) to interpret radiologic reports and identify patients with incidental findings for various diagnoses, including pulmonary nodules and aortic aneurysms.
Deep learning is a subfield of machine learning that utilizes algorithms in a manner inspired by the neural networks in brains and have the ability to learn and modify themselves when exposed to more data.
NLP is a technique widely used in healthcare for keyword search. However, keyword search alone, when applied to identification of incidental findings, results in significantly higher false positive rates.
Eon said it takes standard NLP further by using natural language processing outputs to build the EonEndu deep learning models. Eon said it feeds these NLP datasets into convolutional neural networks (CNNs) that continually refine the models.
This unique two-part process has resulted in the most accurate model for identifying abnormal incidental findings, with a positive predictive value over 90%, the company said.
Eon said its EonDirect is the market leader for lung cancer screening management and is expanding its proven platform to additional disease states. Eon will utilize EonEndu's identification and extraction technology to read radiologic reports and identify patients with an incidental AAA finding.
These incidentals will be segmented by risk, size and previous treatment, then registered to the EonDirect dashboard for surveillance and intervention monitoring. EonDirect "listens" for pertinent events and alerts end users when a patient misses a scheduled exam, the company said.
The EonDirect AAA alpha pilot launched in March at Sky Ridge Medical Center in Lone Tree, and following a three-month review, had identified 13 AAAs greater than 3.5cm and 10 AAAs greater than 5cm.
Serial surveillance of small aneurysms is recommended, while large aneurysms are often recommended for surgical intervention.
As with most incidental findings, positive AAAs are not well followed. Patients with a AAA will experience little to no warning before rupture, and of those patients with a rupture, 50% reach the hospital alive and 30-50% of those won't make it through surgery.
As a result, approximately 11,000 deaths occur each year in the United States due to AAA rupture. Incidentally identified AAAs may not be surveilled as effectively as those detected in a structured screening program. EonEndu + EonDirect can locate, register, and track AAA patients and monitor appropriate next steps for each individual patient, the company said.
Eon said EonEndu + EonDirect now identify and track both lung cancer and abdominal aortic aneurysm populations, greatly impacting patient outcomes and benefitting hospital systems.
Additionally, Eon has developed deep learning models for pancreatic cysts, adrenal and thyroid nodules that will be released later this year. Eon said it will develop an additional 10 disease modules by the end of 2019.