Medical Transcription Analysis with Machine Learning - Doctor/Patient Conversation Demo

Hello and welcome to the video recording of the Medical Transcription Analysis, also known as MTA demo. MTA is a simple medical practitioner patient ambient recording demo solution that integrates Amazon Transcribe Medical and Amazon Comprehend Medical to automate medical conversation transcription processes involving recording, data extraction, comprehension in clinical settings.

On the landing page, we have the option to automatically take notes or play a sample recording with a single click.

For demonstration purposes, we are going to use a recording of a telehealth encounter that was created recently, so we can see the solution in action.

Hello Amy, I'm Dr. Jones. How are you doing today?

I'm okay, but it hurts when I go to the bathroom when I pee.

That's called dysuria and it's pretty common. When did this start?

Two days ago.

Have you taken anything for it?

I tried Tylenol and drank cranberry juice, but it didn't help.

Have you had this before?

Yes, I had this several years ago before you were my doctor.

Do you remember about what year that was?

I think it was 2018.

Okay. How was it treated back then?

The clinic gave me an antibiotic, Bactrim, and that made it better.

Great.

Is there any chance that you're pregnant?

I use protection, but there's always a chance, I guess.

So when was your last menstrual period?

Three weeks ago.

Okay, I think we should check a pregnancy test just to be sure.

Okay.

Have you had any vomiting, nausea, abdominal pain, back pain, shortness of breath, blood in your urine, constipation, diarrhea, or skin rashes?

Yes, I've had an I've had abdominal pain down low.

Any vaginal discharge or drainage or painful intercourse?

No.

Do you have any other problems that you'd like me to address?

No.

Okay. I would like to do a simple physical exam and have you give us a urine sample. I will order both pregnancy test and urine test and that we will do now and determine treatment when we have the results. Does that sound okay to you?

Okay, thanks.

Okay, thanks.

As we could see during the demo, the audio was being recorded and transcribed in real time on the left side with Amazon Transcribe Medical, and the corresponding medical entities were also being identified in real time on the right panel via Amazon Comprehend Medical.

MTA converts unstructured data like audio clips and written notes into structured data that can then be used by EHR systems to generate notes, assessments, and automate codes for billing.

Something was transcribed incorrectly, we can simply double click the specific line we would like to fix and update it to reflect the correct information.

Now diving into the right panel, we can see any and all important information identified by Comprehend Medical.

On the right hand side, we see the legend which color codes information by category such as PHI, medical condition, medications, test treatments, and procedures.

Additionally, Comprehend Medical can provide additional information such as relationships and traits of entities that are defined that can be used by automating generations of notes such as procedure notes or recording patient history.

If we change the transcription manually, the analysis panel will also change to reflect the true value of the entity.

For example, let's fix this P to reflect the correct term.

Another example is type patient also presents with rashes on the right arm.

You will see Comprehend Medical results have been updated on the right hand side.

For medical conditions, we get a description, ICD 10 CM codes, along with corresponding confidence scores such that which indicate how confident Amazon Comprehend is that the correct medical entity has been selected.

If the correct medical condition has not been selected, we can select an alternate medical condition with the options provided.

The results are sorted by their confidence scores.

Additionally, the user can manually input medical conditions as well as remove medical conditions that are no longer relevant.

Likewise, the same type of features and capabilities are available for medications, where we see medication name, RX norm, and confidence scores, as well as for anatomy, test treatment procedures, and PHI.

Anytime a field is marked red, that means it has a low confidence score and could and would require human review to verify the results.

Right, to recap, MTA uses Amazon Transcribe Medical and Amazon Comprehend Medical to be able to extract information from unstructured data into a more structured format that can be used to automate multiple clinic workflows, including note generation, augmenting nursing assessments, automating data entry into EHR systems, and assisting the coding and billing process.

Thank you for walking through the MTA demo with me.