Introduce Amazon Transcribe Medical

People always inform their doctors of their symptoms when they see a doctor, and the doctor will keep recording the corresponding medical records during the process. Although the current electronic health record system (EHR) has greatly simplified these recording processes, due to long-term work fatigue, physicians are unable to concentrate on listening to patients, resulting in problems such as reduced medical quality and inefficiency.

At present, there are many solutions to related issues on the market, such as: hiring an additional scribe to record the condition, so that the doctor can focus on communicating with the patient, but this approach in disguise increases the personnel costs of the hospital, so It is difficult to popularize; more studies have pointed out that when a scribe is nearby, it will increase the patient’s pressure and make it difficult to tell the truth about himself, thus reducing the diagnosis.

In addition to scribes, there are also many solutions for transcribing voice notes into text, which are usually outsourced to manufacturers, but most of them require one to three working days to get the post-transcription results, which will cause problems with the timelines of document archiving. There is no solution that can effectively improve this problem for now.

Amazon launched Transcribe Medical, a fully managed automated speech recognition (ASR) service that is compliant with the HIPAA Act, in December 2019.

Source：Amazon Transcribe Medical

Scenario

Amazon Transcribe Medical is very similar to the previously launched Amazon Transcribe. The difference is that this service focuses on speech recognition in the medical industry. It can highly recognize medical-related professional terms. Developers can integrate with their own systems through this service, or input the transcribed results. An electronic medical record (EHR) system allows physicians to focus more on patients without interrupting conversations and taking medical notes. In addition to being used as medical text records, it can also be used in telemedicine subtitles, drug safety monitoring, and even subsequent medical analysis and other use scenarios, reducing the burden on physicians and improving the patient’s medical experience.

At present, Transcribe Medical supports medical fields including family medicine, internal medicine, pediatrics, obstetrics and gynecology (OB-GYN). Many well-known foreign EHR system providers have integrated Transcribe Medical into their own EHR system, which will be explained in detail used in [Use Cases](#Use Case).

How it Works

Because this service is fully managed, developers only need to prepare a device with a microphone and stream real-time voice to Transcribe Medical through the API to get the transcribed result. To establish a connection using the WebSocket protocol, please pay attention to the following related matters:

• Transcribe 16 kHz audio in real time
• Transcribe up to 4 hours of audio streams
• Transcribe audio in US English
• Word-level timestamp in transcripts
• Word-level confidence in transcripts
• Number normalization
• Supports both dictation and conversation speech types

How to Use

Developers can use Transcribe Medical through the console or SDK. If you want to test this service, you may do simple tests through the console:

• After logging in to the AWS console, verify that the user has IAM permissions to enter the Transcribe console.

• Select Real-time transcription on Transcribe Medical on the left.

• Select Dictation.

• Click on Start streaming and try to read the following sentence, you can see the transcribed result in Transcription output.

  The liver attachments to the adrenal kidney were divided and the liver was reflected superiorly.

The transcription result may cause misjudgment due to factors such as pronunciation and intonation.

• As for the application side, there are also some company which have implemented a simple web application and provided related demonstration videos by calling APIs through SDK.

Use Case

Cerner is a well-known EHR system provider. Its system is deployed in thousands of medical centers around the world. At present, Amazon Transcribe Medical acts as an automatic scribe in its system, and performs medical written records through the API. Even though it is still in the testing stage, which a major breakthrough.

Source：Amazon Transcribe Medical

With Transcribe Medical, the text can be transcribed and further analyzed with Comprehend Medical.

Comprehend Medical is also a natural language analysis (NLP) service for the medical field, so these two services are very suitable for use in combination. The text analyzed by Transcribe Medical is sent to Comprehend Medical for further semantic analysis.

Conclusion

Although Amazon Transcribe Medical can effectively solve many of the above problems, after all, this service has just been released and there are still many feature for improvement, such as supported languages, streaming time, etc. In addition, Amazon still calls for this service to not replace professional medical advice, diagnosis or treatment, which can only be used as a tool to assist the physician.