Mumbai: Now, AI-based speech-to-text tool for doctors to manage medical records

Dr Aditya Shetty, consultant radiologist at Breach Candy Hospital, says the mobile mic app creates a more convenient hands-free capability to dictate reports on his airpods. Pic/Ashish Raje

This year, four senior radiologists from NM Medical, a diagnostics chain in Mumbai, started using Augnito, an AI-based speech-to-text tool designed to help doctors maintain medical records in a faster and accurate manner. Until now, medical professionals would largely rely on dictation and getting their reports typed by transcriptionists. According to the centre, within three weeks, the turnaround time improved by nearly 60 per cent as the doctors were able to edit and finalise their reports within minutes of going through the study. "Earlier, they would have to proofread the typed reports hours later, when they had already moved on to other cases," says Rahil Shah, director of NM Medical. He says that it also addressed the sloppy handwriting that doctors are infamous for, and the trouble that others have deciphering it - a 2017 study by South Africa's National District Hospital, Bloemfontein had found that just 75 per cent of pharmacists could read doctors' instructions from a sample group.

The tool is the brainchild of entrepreneur Rustom Lawyer. For the past seven years, Lawyer has been researching how he could help healthcare providers with efficient systems for medical reporting. The pandemic only necessitated it. "Doctors end up spending as much time inputting patient data as they do in patient consultation," he says. "We wanted to address this pain point and ease the effort through an intelligent voice-driven user interface."

With Augnito, he adds, doctors can produce twice the number of reports in the same time with close to 100 per cent accuracy, which allows them to focus more on patient care and has a direct impact on the bottom line. Moreover, it minimises human contact and allows doctors to have dynamic working hours. In Mumbai, Breach Candy Hospital and Jaslok Hospital have already begun using the app to streamline their operations. Dr Aditya Shetty, consultant radiologist at Breach Candy Hospital, has been using the app since its inception. "What I like is the introduction of the mobile mic app, which creates a more convenient hands-free capability to dictate reports on my AirPods, while I remain focused on viewing and reporting the scans."

Rustom Lawyer

Augnito's parent company Scribetech is a 20-year-old organisation that pioneered clinical documentation in the NHS, UK. The experience helped Lawyer package the technology into a first-of-its-kind AI. Presently, it offers two products: Augnito Everywhere, a voice-to-text medical software that converts a physician's voice to text in real time. It's a multi-platform solution available on desktop, mobile, web, browser plugin and is also offered as a Speech API & SDK that integrates directly with any clinical system. The other is Augnito VoiceRx, which enables a physician to complete a prescription in less than 30 seconds. "VoiceRx accurately transcribes and structures medical data through its latest Clinical Natural Language Understanding technology," he explains. Lawyer normally works with the finance departments at the hospitals and takes them through the "Value Calculator". "A large healthcare group of 2,000 beds could save R20-30 crore a year with the adoption of Augnito."

Augnito does not only type what the doctor says, but also categorises it into medically relevant discrete data. It also contains the entire language of medicine, covering 50-plus specialities and sub-specialties

In the span of a year, Augnito has reached over 5,000 doctors and currently, has a presence in 24 states. Within healthcare, converting speech into text is a challenging task. "A doctor's conversation in a surgery theatre is very clinical and precise, when compared to his interaction with patients outside of it. The latter is free flowing and informal. So, a lot of research goes into making sure the AI understands these contextual differences," says Lawyer. Accents, background noise, multi-party conversation, how fast someone talks, whether they speak in paragraph or sentences, annunciation, context, acronyms were factored while designing the app. This app, he says, offers 99 per cent accuracy for diverse accents.