In the AI Startup of the Week, the editorial staff of ai.nl is featuring promising AI startups, their innovations, solutions and challenges. In this nineteenth episode, we are taking a look at an Eindhoven-based AI startup called Autoscriber. The Dutch startup is building a voice recognition software to capture and summarise consultation between a health professional and patient.
Autoscriber was founded in June, 2021, by Jacqueline Kazmaier and Koen Bonenkamp to solve a known problem in the healthcare industry. As the healthcare industry becomes more and more AI-fuelled and is using data to make critical decisions, doctors also struggle to capture conversations with their patients in a structured way.
The voice recognition software built by Autoscriber is meant to solve this very pain point. The Dutch AI startup has built a speech recognition system and natural language processing technologies to extract structured data from conversations between physicians and patients. With automation at its centre and collaboration with doctors at the core, here is how Autoscriber is trying to change the healthcare industry.
Automating administrative actions in the electronic patient record
Autoscriber was formed as a research collaboration between CAIRElab and Cape AI. CAIRElab is an innovation unit within the LUMC focusing on the application of AI in healthcare. Cape AI is a consultancy focussing on the development of neural language processing applications.
The startup is working with the mission to better care and provide personal attention to each and every patient. It is doing so with the help of its innovative speech recognition technology designed to automate “administrative actions in the electronic patient record.”
With its technology, Autoscriber is looking to make healthcare affordable and accessible for all. It is also helping drive data-driven healthcare initiatives, and the service is being rolled out at a time when there is desire among patients to understand and self-determine their situation.
The speech to text solution developed by Autoscriber relies on natural language processing, a field of artificial intelligence. The solution allows healthcare professionals to convert audio recordings into text and even interpret them automatically with full context. While speech to text as a domain may not be new, Autoscriber has built its tool to ensure that the model is able to recognise medical terms in the doctor’s office.
It is one thing to find a problem and another thing to build a solution. For this need of a solution that can automatically record a consultation in the electronic patient file (EPD), Autoscriber directly worked with doctors. It has also partnered with Google Cloud and CTS for the development of its NLP techniques and deployment of Google Cloud technology.
Speech to EHR as a complete solution
Autoscriber is primarily using neural language processing, a field within AI, to accomplish its goal of helping medical professionals automate part of their administrative actions. To accomplish this goal, the startup has three products: Speech to EHR, Speech to text, and text to entries. Speech to EHR is the marquee product offered by Autoscriber and it also acts as the technological showcase for the startup.
As a flagship product, Autoscriber does everything from recording, transcribing, extracting to summarising and assisting with the administrative actions. It starts with Autoscriber installing microphones in the consulting room and a software is used to convert the conversation into text. It takes into account the different speakers and dialects.
The software is also able to extract clinical terms and automatically recognise and link it to standardised codes and ontologies. These clinical terms are then combined into a readable summary that can be placed in the file of the patient. The clever implementation is in how it works. The medical practitioner can begin the recording directly from the patient’s file and the file is automatically enriched afterwards.
On its website, Autoscriber reveals how it turns the data into structured information. There is also an option to edit the report. “The technology offers many advantages, including better contact with the patient and new analysis possibilities with the structured data obtained from the anamnesis,” says Martijn Bauer, Internist at LUMC.
Autoscriber and its other AI products
In addition to Speech to EHR, Autoscriber also offers other tools like speech to text and text to entities. These are conventional products and can even be dubbed as derivatives of Autoscriber’s marquee product. Here is how these two products work in a healthcare ecosystem.
Speech to Text: As the name itself gives away, this is a tool that allows users to turn speech into text. The only difference being that this tool is designed with medical practitioners in mind and is being offered as a service to be used within the software framework used by hospitals and healthcare facilities. Autoscrbers offers an API that users can integrate with their own workflow.
Text to Entities: This is the third product from Autoscriber and is also available as a service to be integrated with existing software or user’s own software. With this tool, users can extract all the medical terms from their conversation with the patient or conversation on the patient file. It is limited to Dutch language but can be used to extract symptoms, medication, procedures, diagnoses, et cetera.
AI in healthcare
Artificial intelligence (AI) was supposed to disrupt every industry but its impact has been felt the most by the healthcare industry. There is now AI in every level of the healthcare industry and AI is often seen in the form of natural language processing applications designed to understand and classify clinical documentation.
Autoscriber is also bringing NLP to the healthcare industry and focussing particularly on patient electronic records. It is also trying to not only analyse unstructured data but also helping medical practitioners get better insights that would help with treatment and deliver better results for patients.