AI takes over medical transcription with Nova-3 Medical

As artificial intelligence (AI) continues to shape our everyday lives, medical transcription has been one of its popular use cases, and Deepgram’s new speech-to-text Nova-3 Medical designed for clinical environments aims to bring more precision into the field.

Specifically, Deepgram’s medical speech-to-text model claims to have unmatched accuracy in healthcare settings, filtering out irrelevant noise and grabbing critical details like diagnostic terms, procedure details, and medication names, the company announced on March 3.

Nova-3 Medical specifications

Furthermore, Deepgram said its HIPAA-compliant and secure model provides flexible self-service customization, with Keyterm Prompting that facilitates fine-tuning of the model by adding up to 100 custom terms. It builds upon its predecessor, the Nova-2 Medical, which already features ultra speed, enterprise-grade security, and compliance.

Meanwhile, it inherits advanced in-context learning from Nova-3 that allows it to adapt in real-time to specialized terminology, as well as using Keyterm Prompting, a proprietary feature that increases accuracy and recognition of up to 100 important terminology, product and company names, industry jargon, and more.

The AI model covers fields such as cardiology, neurology, and oncology, and supports dynamic updates to keep up with the latest medical advancements in terms of drugs, procedures, and diagnostic terms, while deployment options include managed cloud, on-premises, and Virtual Private Cloud (VPC).

It’s also important to note that Deepgram, in collaboration with AWS, carried out a demo powered by Nova-3 Medical, showcasing how it can seamlessly integrate with AWS services and existing Electronic Health Record (EHR) systems, capturing clinical notes, processing drug dispatching commands, and scheduling tasks – all in real-time.

AI in healthcare

Elsewhere, AI may soon help with early detection and treatment of glaucoma, one of the most dangerous eye diseases in the world and the leading cause of irreversible blindness, for which scientists have developed an AI-based Glaucoma Screening (AI-GS) tool.

On top of that, a new AI program called EchoNet-Liver and trained to study patterns across more than 1.5 million echocardiogram videos can identify chronic liver disease from pictures picked up during the heart test, delivering similar results as using patients’ abdominal ultrasounds or MRI images.