AWS HealthScribe and Bedrock: introducing generative AI into healthcare

How AWS is becoming a large-scale generative AI provider in healthcare.

From the early days of its evolution, Artificial Intelligence (AI) has been the focal point of attention in healthcare. According to recent Morgan Stanley research, 94% of healthcare enterprises have integrated AI and Machine Learning (ML) into their operational frameworks to various degrees. So, when the generative AI gold rush hit the scene, it became abundantly clear that we could expect even more impactful strides to be made in healthcare and life sciences.

Tech industry giants are closely monitoring the integration of generative AI, as they seek to drive innovation forward. Amazon Web Services (AWS), a trailblazer in cloud computing, has been deepening its strategic foothold in the domain. In the last couple of months, the technology heavy-lifter made several vital announcements: it released agents for Amazon Bedrock, which enable companies to build customized AI apps, and launched AWS HealthScribe, an AI-driven service designed to alleviate the workload of healthcare professionals. Let’s take a closer look at these tools and what they bring to the industry.

What is AWS HealthScribe and how does it work?

In July 2023, AWS announced the release of HealthScribe, a solution that helps build applications for automated and efficient clinical documentation. This service leverages speech recognition and generative AI, and can analyze patient-clinician interactions. Its core functionality leans on the ability to convert spoken medical interactions into structured and actionable data. Here is a closer look at HealthScribe as presented by AWS (see Fig. 1):Figure 1. How interaction transcript transfers into clinical documentation with HealthScribe

In operation, AWS HealthScribe ingests audio recordings and converts the content into text. Subsequently, the platform analyzes this transcribed health data to identify key details such as patient information, diagnoses, medications, and treatment plans. The AI engine also assists in formatting the data into structured templates, which makes it more accessible for further processing and integration into electronic health records (EHR) systems. Beyond that, AWS HealthScribe allows:AWS HealthScribe makes it possible for companies to avoid one fundamental challenge: the numerous complexities of integrating generative AI and speech recognition into healthcare solutions. Creating something like HealthScribe would require businesses to work with their own Large Language Models (LLMs), use gargantuan amounts of health data, and have the necessary complete capacity at hand.

Another layer of complexity stems from the sheer diversity of medical terminology. The healthcare domain employs an extensive lexicon of terms that encompass diagnoses, procedures, and medications, to say the least. Integrating accurate medical terminology recognition requires developing specialized models that can accurately identify and contextualize these terms within transcripts. This complexity is further compounded by variations in accents, dialects, and speech patterns among both clinicians and patients.

Last but not least, the intricacies of the healthcare sector, combined with stringent regulations and patient privacy concerns, present hurdles for building AI-driven solutions. In the ever-changing healthcare industry, privacy protection is the cornerstone upon which trustworthy and effective AI-driven solutions should be constructed. Yet, balancing technological advancement and ethical responsibility requires capacity, expertise, and careful navigation. AWS reiterates that HealthScribe, as a HIPAA-eligible solution, has it all.

Amazon Bedrock

AWS has been eager to offer their clients innovative services within the AI segment. Unlike organizations collaborating with OpenAI, AWS, as a cloud provider, continues to work on its own models. In fact, it created Bedrock, a library designed to empower companies to deploy various AI models. Bedrock encapsulates a wide array of tools and resources that cater to the entire AI development lifecycle. It serves as a platform where businesses can access Anthropic, Stable Diffusion, Claude, Jurassic, and Amazon’s own Titan LLMs.

Bedrock integrates with AWS’s cloud infrastructure to facilitate the training process and leverages its scalable compute resources to accelerate model training and optimization. This offers significant advantages for multiple industries, including healthcare and life sciences. As a result, even resource-intensive tasks can be completed efficiently, regardless of the project’s scale.

Bedrock’s support for distributed training further empowers biotech researchers to use the potential of multiple GPUs or even entire clusters, and allows for quicker convergence in the development of AI models for disease diagnosis or drug development. This combination of AI innovation and cloud scalability promises to propel new ways of thinking regarding medical solutions.

AI agents for Bedrock

In July 2023, AWS also released agents for Amazon Bedrock, a platform with new assisting capacities that allow companies to automatically execute tasks and optimize operations in innovative ways. Consider the following scenarios in which healthcare organizations can leverage generative AI-powered apps with broad functionalities:

Scenarios

• Healthcare patient care coordination: schedule appointments, send reminders, and provide timely follow-ups.
• Healthcare medical records management: securely manage electronic health records in healthcare, support compliance, and ensure easy access to patient data.
• Healthcare diagnosis assistance: help healthcare professionals with clinical decision support through AI’s analysis of a patient’s symptoms, history, and lab results.
• Personalized treatment plans: generate recommendations based on a patient’s medical history, genetics, and current health status.
• Medical text summarization: summarize lengthy medical documents and research papers through generative AI in order to extract key insights and facilitate quicker literature reviews.

Agents for Amazon Bedrock is part of a more considerable sentiment towards AI agents. Companies like Meta and OpenAI now pledge to invest more into these software tools. This collective movement towards broader AI availability aligns with the evolving nature of the technology itself. As generative AI becomes increasingly intertwined with healthcare, the transition from exclusivity to inclusivity becomes imperative. The collaboration and open approach among industry leaders fosters an environment where groundbreaking ideas can emerge from unexpected sources.

Why it matters

With the AI arms race intensifying, healthcare software vendors are researching solutions that could align with the great demand for speed and accuracy. The goal is to create generative AI systems that can rapidly identify patterns, anomalies, and potential health risks as well as help healthcare practitioners with daily routines. This movement covers everything from medical imaging and genomics to electronic health records and real-time patient monitoring. Here is a closer look at the spheres where AI could generate impactful change in the years to come, according to a McKinsey study:Figure 2. Areas of healthcare where AI can accumulate substantial transformations, according to McKinsey.

Generative AI in healthcare has witnessed several pivotal advancements. In the quest for novel solutions, Google introduced Med-PaLM 2, an LLM explicitly designed for the medical domain. The company reportedly shared access to the model with a range of organizations, including the Mayo Clinic, in order to test and examine the model’s capacities, the Wall Street Journal highlights. The core characteristic of Med-PaLM 2 is that it was trained on health data, presumably making it more helpful and knowledgeable of the specific medical context.

Another development in that area comes with Google’s Care Studio and Cloud Healthcare Data Engine. While Care Studio embodies a centralized data repository for clinicians that helps organize information in an integrated manner, Cloud Healthcare Data Engine strives to eliminate silos and make data more structured, accessible, and accurate, all within the HIPAA compliance framework. These tools lend themselves to the challenge of enhancing healthcare data management and accessibility, both of which are critical for delivering more effective and efficient patient care.

In response to the generative AI expansion, Microsoft forged a series of partnerships to strengthen collaboration across the healthcare sector. The organization joined forces with Epic, Paige, and Nuance in order to support patient-centric care and enhanced diagnostics. In addition, the company recently introduced Microsoft Copilot, which can embed advanced AI possibilities into the healthcare workplace.

And, this is just a drop in the ocean. A 2022 McKinsey finding underscores the trillion-dollar potential of generative AI in the U.S. healthcare sector. It is a reminder of the immense value that technology can bring to the industry. This value isn’t solely measured in financial terms, but also in terms of improving healthcare outcomes, increasing access to quality care, and honing the healthcare experience. This potential can also be interpreted as a call to action for the entire healthcare ecosystem to come together and collaborate in ways that will re-imagine the domain.

Envisioning the future: AWS, generative AI, and healthcare

As HealthScribe and agents for Amazon Bedrock mature, and AWS continues to expand its offerings, the democratization of generative AI tools is expected to catalyze a wave of innovation in healthcare. During this phase of exploration and adaptation, we are witnessing the beginnings of a pivotal transformation that promises to substantially impact the prevention, treatment, and diagnosis of diseases, as well as healthcare delivery and patient outcomes.

This is the right time for generative AI adoption. Tap into the expertise of Avenga’s specialists and transform your business ecosystem with new healthcare solutions and AWS services in a cohesive way: contact us.