When it comes to launching AI/ML-Powered applications, AWS is undoubtedly the goto platform. AWS offers agility, global scalability, best security, and all tools needed to sustain and grow such an application.
This document elaborates one such use case where a HIPAA compliant MVP telemedicine solution was built and launched in just weeks. It uses pre-tested and pre-trained AI/ML models as AWS managed services, which tremendously simplify the development and deployment of the solution. Some of the technologies used in this solution are:
A Healthcare startup has developed a new-gen solution for telemedicine to better deal with CDC guidelines for the safety of healthcare professionals and patients during the Covid-19 pandemic. This solution allows patients to request a telemedicine visit from a healthcare provider.
Healthcare staff can then schedule an audio/video-based conference call with a physician or physician-assistant based on staff’s availability. Conference call link and scheduled appointment information are then emailed to the patient. Once call is started by the Patient, AWS ML services are employed to:
Convert call audio into turn-by-turn call script with accuracy. The solution allows the use of custom vocabulary for better accuracy of the transcription process. Once transcribed contents are auto redacted to protect PII data and saved in appropriate HIPAA compliant storage
Medical comprehend uses call text to extract medically significant information such as prescription and dosage information, symptoms, etc and create structured records for integration with any medical information system /electronic medical record systems
Saved information can be used to process doctor notes and any other medically significant audio recording
The whole solution is built on a secure computing platform and meets all HIPAA requirements for access and authentication, security of sensitive data at rest and during transit, proper identification, classification, handling of protected health information, and reliable archival and back up of PII data.
Following is the reference architecture for a secure and reliable telemedicine service which fully leverage AI and Machine Learning for safety and efficiency.
AWS AI/ML services for healthcare can be used without developing deeper AI/ML knowledge or experience. These managed services provide pre-tested and pre-trained ML models ready for production use. Additional Custom ML models can be developed and deployed alongside using AWS Sagemaker. AWS AI services tier fast-track the overall development process as these models have been trained using multiple complex algorithms to attain better accuracy. Following AI services were used:
AWS Transcribe to convert call audio into turn by turn call log and redact any PII data
AWS Comprehend Medical to convert texting into medical data for better integration with healthcare applications.Comprehend Medical uses pre-trained NLP models and algorithms to do accurate Named Entity Resolution (NER) for healthcare related terms.
AWS Chime SDK to stream call audio data for scalable transcription process.
AWS provides a rich set of structured and unstructured data management services, e.g. AWS RDS and Elasticache. We used AWS RDS Aurora engine that provides required Database scalability as well as compatibility with existing MySQL scripts and dev work at a fraction of cost as compared to on-prem databases. We used Elasticache for storing/caching various computation results to achieve better response time and compute efficiency.
A Highly scalable ECS cluster which runs telemedicine UI and scheduling apps. ECS cluster provides highly scalable serverless compute resources for containerized applications and provides excellent performance, agility, security, and isolation. ECS is feature-rich as it supports multiple compute capacity providers, supports Blue/Green deployments, and can run a variety of workloads. ECS service mesh provides a solid foundation for microservice architecture. For this solution, the ECS cluster runs a service that uses Chime SDK and Node.JS to run Telemedicine applications. AWS CloudFront distribution and WAF V2 protect applications from a variety of threats and attacks.
AWS provides a fully serverless application build system including CodeCommit repos, serverless compute resources for the containerized app, secure private container repository, and CICD pipelines to tie them all together and create super-efficient build, test and release cycle, Blue/Green deployment of ECS application remove all risks of new and uncertain code push to production.
AWS Organizations and Single sign-on services provide required Isolation and policy controls to meet or exceed HIPAA access and authentic requirements. AWS managed security services used include AWS Macie for identifying and location PII data, AWS Config/Audit trail for capturing, configurations and access to services, AWS Guard Duty to detect threats. AWS Inspector for application security, AWS WAF with Shield for web application security and DDOS attack prevention, and AWS security hub to generate compliance reports.
High Plains Computing Advantage
At High Plains Computing (HPC), we specialize in launching all sorts of applications in the AWS. Our team of AWS DevOps is dedicated to AWS automation using Infrastructure as code, AWS security, and setting up dev environments on AWS. Our team has built/accumulated a vast repository of Terraform modules that are pre-tested and can provision infrastructure for any application or service within days/weeks rather than months.
Our AWS Certified Solution Architect professional and DevOps professional staff will ensure all AWS infrastructure code, as well as the application that is getting deployed, follow “AWS well Architected” program guidelines and best practices to achieve operational excellence, security, reliability of operation, cost efficiency, and sustainability.
High Plains Computing (HPC) professionals work with your team to deploy your amazing application on AWS infrastructure.