Recently, Amazon Cloud Technology launched Amazon HealthScribe at the New York Summit. The service complies with the relevant requirements of HIPAA ("Health Insurance Portability and Accountability Act") and provides medical software suppliers with a generative AI based on speech and text recognition. technology to help it create clinical applications that can automatically generate documents to provide more convenient diagnosis and treatment services for patients. With the help of Amazon HealthScribe, medical software suppliers can use API to extract key information in the diagnosis and treatment process, find the key content in the conversation between doctors and patients, realize the automatic generation of patient information, and enter it into the electronic health record system middle.
Amazon HealthScribe, backed by Amazon Bedrock, will make it easier and faster for medical software vendors to integrate generative AI capabilities into their applications. In common general medicine and orthopedics, medical software suppliers have already applied Amazon HealthScribe to daily diagnosis and treatment services. They no longer need to manage the underlying machine learning infrastructure, and they no longer need to train their own medical-specific large language models.
Amazon HealthScribe has taken a responsible approach to deploying AI systems that use textual information generated from raw conversations to make it easier for doctors to review clinical records before they are entered into the electronic health system. Amazon HealthScribe has attached great importance to security and privacy since its establishment. Users can control the storage location of data, encrypt data in different states, and will not use the input or output data generated by this service for model training. .
Generative AI technology is rapidly affecting many fields such as medicine and biomedicine. In this context, research on generative AI is gradually emerging, and medical software suppliers are using it in clinical practice, which is expected to fundamentally solve common clinical problems. Among them, the transcription of clinical documents after doctor-patient communication is one of the most common problems. On the one hand, it is of great significance for legal compliance, quality control, and reimbursement of medical expenses; but at the same time, it is a cumbersome task that consumes a lot of time in the communication between doctors and patients. Although many medical software suppliers have adopted text-to-speech and natural language processing methods to achieve rapid analysis of patients' conditions, due to the lack of generative AI technology, it is impossible to quickly convert patient's condition information into concise clinical documents, and entered into the electronic health record system. At present, the application of generative AI technology is very complicated, and it will consume huge engineering resources to realize the effective integration of multiple AI systems.
To build these generative AI capabilities, vendors need to improve understanding of clinical documentation by tuning their own large language models to accurately generate clinical documentation. Therefore, they need to professionally learn complex medical vocabulary including general internal medicine, pediatrics or orthopedics, and understand, analyze and summarize it, and finally realize the understanding and analysis of various dialogues between doctors and patients and identification.
For the system to work properly, software vendors also need to build generative AI products responsibly, one of which is to allow doctors to trace the source of generated text, thereby reducing the number of errors. At the same time, in order to ensure the security and privacy requirements of the medical industry, medical software suppliers also need to spend a lot of time and resources. The above bottlenecks restrict the development of artificial intelligence technology. Although artificial intelligence technology has great benefits for doctors and patients, its application still faces great challenges.
Amazon HealthScribe is an AI-driven, HIPAA-compliant medical service that helps medical software vendors build clinical applications that save a lot of time by automatically creating clinical records and analyzing communications between doctors and patients. With easy-to-use APIs, medical software suppliers can quickly provide patients with personalized diagnosis and treatment plans, and provide patients with different diagnosis and treatment experiences, reducing the difficulty of integrating and optimizing multiple artificial intelligence systems. By integrating Amazon HealthScribe into clinical applications, medical software vendors can use its text-to-speech capabilities to create a reliable transcript of a conversation and be able to identify the speaker and classify different types of speech by clinical relevance. session. Secondly, Amazon HealthScribe’s powerful natural language processing capabilities and intelligent analysis capabilities can be used to realize automatic identification of patient conditions, medications and other characteristics; finally, according to the doctor’s analysis of the patient’s condition, the patient’s diagnosis and treatment information is automatically generated for the patient to consult.
Amazon HealthScribe will use generative AI technology provided by Amazon Bedrock to create clinical records for general practice and orthopedics that facilitate doctor-patient communication. In the clinical records generated by AI, every sentence comes from the conversation between the doctor and the patient, so that the doctor can more easily check the historical information of the clinical records, making the clinical records more transparent and accurate. In addition, this service also guarantees the privacy and security of data, and will not save any data of customers after processing customer requests, and will also encrypt information in various states. Medical software suppliers have control and ownership of the storage location and data of their written records and preliminary clinical records. Additionally, the input and output data generated by this service will not be used for Amazon HealthScribe training.
"Health care customers and partners tell us they want to spend more time creating innovative clinical care and research solutions for patients and less time creating, maintaining and operating underlying medical data capabilities." Amazon Cloud Technology Bratin Saha, vice president of machine learning and artificial intelligence services, said, "This is why Amazon Cloud Technology has spared no effort to build a complete set of AI-driven, high-performance, health applications for the general population, so that clinicians can spend more time with patients themselves. Documentation is a particularly time-consuming task for healthcare professionals, so we are proud that Amazon HealthScribe’s powerful generative AI capabilities can help them ease this burden. The The official launch of the service further demonstrates Amazon Cloud Technology's determination to serve the healthcare and life science industries, as well as our responsible attitude towards technologies such as generative AI, which in turn helps doctors reduce the burden of clinical documentation and improves patient consultation experience. "
Amazon HealthScribe, as a professional technology for customers in the healthcare industry, has successfully provided many customers with a variety of different technical supports, and Amazon Cloud Technology will continue to promote technological innovation in healthcare to provide users with deeper Serve. In addition, Amazon Cloud Technology also announced a new service, Amazon HealthImaging, which can help customers more easily store, convert and analyze PB-level medical imaging data, thereby reducing the burden on the underlying infrastructure configuration and ensuring the operating efficiency of the system .
As an industry leader, 3M Health Information Systems (3M HIS) has applied its M*Modal text understanding, dialogue and surrounding AI solutions to more than 300,000 clinicians. "Amazon Cloud Technology's machine learning service helps 3M HIS change the tedious workflow of clinicians, and helps medical institutions simplify clinical documentation and billing," said Garri Garrison, President of 3M HIS, "3M HIS is working with Amazon Cloud Technology to bring conversational and generative AI directly into the clinical documentation workflow. Amazon HealthScribe will be at the heart of our clinical applications, helping to accelerate, optimize and scale the delivery of 3M’s clinical documentation work and virtual assistant solutions.”