Project toolkit: Streaming, serverless and ML
ClearScale’s telemedicine application, which runs on macOS, iOS and iPadOS, among other operating systems, interfaces with patients through their personal smart devices. The system lets physicians conduct virtual appointments using a medical coding application, BugMD, that ClearScale built for CPS in a previous project.
To create the telemedicine system, ClearScale wove together several Amazon services. Amazon AppStream 2.0, a desktop streaming service, lets CPS deliver its application’s security to any number of physician users, ClearScale said. Amazon Connect, an omnichannel cloud contact center service, enabled ClearScale to bundle physician services such as medical billing, coding and telemedicine. ClearScale deployed AWS Lambda and AWS Fargate to provide the processing power for automated billing and video/audio streaming. Lambda is a serverless compute service, while Fargate lets organizations run containers on AWS’ cloud platform.
Vyacheslav Gorlov
In addition, CPS wanted to save audio recorded during remote patient visits for billing purposes. BugMD already provided the ability to accurately translate the medical appointment notes a physician dictates and records after a patient visit. ClearScale built upon that feature for CPS so BugMD could record telemedicine appointments. The telemedicine application supports standalone telephony — a virtual phone number provisioned for a physician via Amazon Connect — as well as the telephony features of services such as Zoom, Skype or WhatsApp, said Vyacheslav Gorlov, senior solutions architect at ClearScale.
Audio quality posed a challenge for BugMD’s use in a telemedicine setting. With phone conversations, “you have 8 kHz of data only, the line is shared with other data, and dropped frames are the common issue,” Gorlov said. “Processing of such audio requires tremendous compute resources and power.”
ClearScale used the AudioKit.io framework, which lets developers add audio functionality to applications, to tackle the problem. AudioKit provides background noise removal and normalization and backfills dropped frames, Gorlov said. Those audio improvements take place on the physicians’ devices. AudioKit’s linear and AI-backed algorithms boost the audio quality enough for the Amazon services to properly handle the recorded appointments, Gorlov explained.
The telemedicine app sends BugMD’s recordings of medical appointments to the cloud — specifically, to an AWS Simple Storage Service bucket — and encrypts the data at rest and in transit to comply with HIPPA.
The application then uses the Amazon Transcribe Medical service to convert English language audio into text. For appointments conducted in languages other than English, the application can use Amazon Translation in tandem with Amazon Transcribe to render English transcriptions, Gorlov said.
Matt Dallmann
Once the transcription is created, AWS’ machine learning (ML) algorithms process the text, extracting information on diagnoses, procedures and prescriptions for coding.
“A variety of medical coding rules and if-then scenarios were implemented to give the AI a head start on accuracy and ensure compliance for certain coding regulations,” said CPS President Matt Dallmann.
Physicians can review the results of the ML text processing and send them on to insurance companies to receive reimbursement.