T-Pro Unveils MirAI

T-Pro Unveils MirAI: The Future of Medical Automatic Speech Recognition, exclusively for T-Pro Customers

Introducing ‘MirAI’

T-Pro is proud to announce the launch of MirAI, our latest and most advanced Automatic Speech Recognition (ASR) model, available exclusively to T-Pro customers. MirAI represents a significant leap forward in speech recognition technology, offering unparalleled accuracy, enhanced accent recognition and superior medical terminology precision, making it the ideal solution for healthcare professionals.

The name MirAI draws inspiration from Japanese and Irish languages, 'Mirai' (未来) is the Japanese word for future, and 'Mír' is the Irish (Gaeilge) word for segment or musical phrase, with MirAI evoking our use of AI to capture the rhythm and structure of spoken language.

Key features: 

• Improved word error rate: MirAI delivers a dramatic reduction in word error rate compared to its peers, saving time for clinicians and transcribers by recognising and delivering transcriptions that require less post-editing effort.
• Enhanced Accent Recognition: Recognising the global nature of T-Pro’s client base, MirAI excels in understanding a wide range of accents, making it a versatile tool for hospitals and clinics worldwide.
• Superior Terminology Accuracy: MirAI sets a new standard in the accurate transcription of specialised terminology, particularly across medical specialities. This precision ensures that critical information is captured precisely and efficiently.
  
  
  

How have T-Pro delivered such a significant improvement in the ASR’s word error rate?

Word error rate (WER) is the industry standard metric for tracking ASR model improvements. The lower the WER the greater the time and efficiency benefits that can be delivered directly to clinicians. 

In order to calculate the WER we require a test set of medical dictations, with corresponding transcriptions written by human experts. Unlike many benchmarks that rely on general domain or open-source test sets, we focus specifically on our customers' use case, evaluating on a test set of audios typical of T-Pro users. 

To ensure a comprehensive and fair evaluation, we extract a balanced test set from our vast database of anonymised medical documents produced annually across North America, Europe, and Australia. By using in-domain data, representative of our customers, rather than generic open-source datasets, we offer a true reflection of MirAI's performance in the specific contexts where it will be deployed, providing a robust and reliable measure across various real-world scenarios. 

Using this specialised test set to measure our performance we can see that MirAI achieved a WER of 9.3%. Putting this into the context of other speech recognition services available on the market reveals that MirAI delivers the fewest errors, further emphasising MirAI's exceptional accuracy in this field.

Enhanced Accent Recognition

We are committed to making impactful improvements for our global client base. A standout feature of MirAI is its enhanced recognition of diverse accents, particularly Irish and UK accents.

The chart below demonstrates substantial accuracy gains of 48% for Irish accents, 14% for UK accents, and 37% for other accents compared to the closest competitor.

These improvements allow us to recognise and deliver transcriptions that require less post-editing effort, saving time for clinicians and transcribers.

Superior Terminology Accuracy

Accuracy is crucial when transcribing specialised terminology for T-Pro users. We use F1 scores to gauge the accuracy of important terms in our dictations, particularly place names and medical terms. F1 scores range from 0 to 1, with 1 being perfect accuracy. They balance precision (correctness of predictions) and recall (completeness of correct answers found).

Using F1 scoring we can see that T-Pro MirAI achieves higher accuracy on place names and medical terms when compared to our closest competitor, another domain-specific ASR service. 

What this means for you?

T-Pro’s MirAI speech recognition model represents a significant leap forward in accuracy, accent awareness, and medical terminology handling. For our customers this delivers more accurate recognition across diverse speech patterns, reduces time spent on corrections and seamlessly handles complex medical terms. Clinicians will experience smoother workflows, improved documentation, and the increased ability to focus more on patient care. 

Hear directly from our customers about the accuracy we deliver...

• Neil Perry, CIO at Dartford & Gravesham NHS Trust
  
  "We’ve invested in the latest proven technology. T-Pro’s AI-powered speech recognition engine is super‑fast and accurate. The workflow makes life easier for our clinical teams, making cross-organisational communications and transfers of care efficient. Healthcare staff are more mobile than ever, both inside and outside the hospital. There is also an increased need for patient stories to become immediately available. Integrating T-Pro with our Alcidion Electronic Patient Record system will enable us to meet both of these needs - and accelerate our digital transformation programme."

• Mike Bailey, Contract Project Manager at Hywel Dda University Health Board

"We found T-Pro’s software to be most helpful and cost-effective compared to other providers. Their AI-powered voice solutions including Digital Dictation & Speech Recognition are the best in the market with close to 100% accuracy and the support from the team has been great, they are always available to assist and train our teams here when needed."

• Brian Anderson, Project Manager at Ramsay Healthcare UK

"Having worked with many digital organisations over the past 15 years, T-Pro has been one of the most consistent and available groups to be working with. They continue to shape the speech recognition platform to meet our immediate and future needs. When we are met with any challenges, T-Pro continues to work with us to provide solutions.”

• Michell Atkinson, ICT-Coordinator at Naas General Hospital

"T-Pro has freed our clinicians from their desks and made their mobile devices a working extension of our PAS, IMS MAXIMS. Using T-Pro’s user-friendly cloud-based digital dictation platform, we have been able to streamline clinical documentation and boost staff productivity and turnaround.”

Interested to learn more about T-Pro? 

Why T-Pro

T-Pro is an award-winning global healthcare technology and services provider. We provide solutions that use machine learning and AI technologies, such as speech recognition and clinical language understanding, to enhance the way clinicians produce clinical documentation. T-Pro makes it easy for clinicians to document patient encounters whilst also providing valuable insights which can help to reduce risk, drive efficiencies across healthcare organisations and provide more time for patient care.

For more information, please visit: https://info.tpro.io/ or contact sales@tpro.io.