Overview
Artificial intelligence (AI) in healthcare is a catch-all term to describe the use of algorithms trained by machine learning (ML) in three major categories of the healthcare market: the application of healthcare (life sciences), the provision of healthcare (providers), and the consumption of healthcare (payers).
Advances in this type of algorithmic machine learning now allow use cases for artificial intelligence in healthcare to expand beyond the reactive AIs of the past. These advances can play a significant role in driving healthcare transformation and modernization.
In order to analyze and act upon medical data, data must be made accessible and actionable. Once the model is trained with sufficient data, the machine can inference in new settings. AI inference is the operational phase of AI, where the model is able to apply learning from training and apply it to real-world situations. Only then can we have broader, more effective and efficient applications of health services.
AI can broadly be categorized into four types:
- Reactive AI: Artificial intelligence that does not use machine learning to improve and reacts in the exact same way every time it encounters an identical situation.
- Limited memory AI: This artificial intelligence uses machine learning to craft algorithms based on past performance. This is where many of the current advancements in AI are happening right now, and why you might see artificial intelligence referred to as AI/ML. Furthermore, a subset of machine learning, called deep learning, involves several layers of analysis to extract more meaning out of raw data.
- Theory of mind AI: This is an AI that can understand and remember emotions and interact with people based on that model. Theory of mind AI is largely nascent or theoretical at this point.
- Self-aware AI or "true" AI: Self-aware AI is aware of its own emotions and has a level of consciousness similar to that of a human. Self-aware AI is currently in the theoretical stage.
When we talk about AI in healthcare, we are largely referring to the rapid advancements both in the algorithms and the applications of limited memory AI.
What are the benefits of AI in healthcare?
New advancements in AI can fundamentally shift patient outcomes by helping doctors and other medical practitioners deliver more accurate diagnoses and plans of treatment. These advancements can also help administrators precisely and accurately allocate medical resources.
This can benefit the three pillars of the healthcare market (life sciences, providers, payers), in many ways. Efficiently acquiring, distributing, and leveraging the most up-to-date information can help clinicians better treat patients, quickly cull data from multiple sources to better manage existing conditions, and aid in the prediction or identification of new conditions or disease onset.
Better distributed data processes allow administrators to more efficiently prioritize and verify claims and streamline the overall claims process, improving the accuracy and speed of information communicated to patients, customers, and providers. Overall, the collation of data into healthcare algorithms can help predict future risk, and give healthcare administrators more power to manage and improve the care available to society.
A few of the ways that AI in healthcare can be a benefit to patients, healthcare providers, and payers:
Faster diagnosis
Data insights processed by AI algorithms and real-time predictive analytics can be used to speed up diagnosis, meaning that patients get treatment more quickly, leading to better outcomes and fewer overall resources used to solve the problem. An example of this is HCA Healthcare, one of the largest healthcare service providers in the United States, that used Red Hat solutions to create a real-time predictive analytics product system to more accurately and rapidly detect sepsis, a potentially life-threatening condition.
Claims management
The bureaucracy of medical claims and payment can take thousands of work hours. Doing each claim manually also raises the risk for errors to creep into the process, which is neither good for the patients making the claims, nor the providers trying to balance the books. AI can help automate the filing and provide insightful recommendations based on claims management data analysis. This could accelerate claims processing, improving employee and customer experiences.
Fraud, waste, and abuse
Robotic process automation (RPA) is able to rapidly go through documents with a pace and accuracy that manual intervention can not match. These algorithms can then flag fraudulent activity or waste, and over time as the algorithms improve, they get more effective at finding issues.
Expand access to healthcare offerings
AI assisted diagnosis can widen patient groups receiving services. For example, AI-assisted radiology and medical imaging could allow a larger number of professionals to interpret ultrasounds, which could reduce the bottleneck on a handful of specialists, and expand the number of patients who have access to the technology.
Drug development
Novel drugs require the discovery of suitable dosage amounts and delivery characteristics. Computational AI tools can enhance or even replace trial-and-error approaches, and allow for quicker and more efficient models to monitor the entire process. This can allow for the more rapid development of new and novel drugs, saving both pharmaceutical organizations and the end customer money.
Red Hat resources
Challenges in operationalizing AI/ML
While AI in healthcare can offer numerous advantages, implementation can pose several complex challenges. A few of the challenges AI is causing the healthcare industry to face include:
Data management and operalization collection
Many challenges occur in the process of collecting, analyzing, and applying healthcare data.
For AI to correctly feed relevant algorithms, a huge volume of data needs to be processed in real-time. The data collection challenge is therefore multifaceted.
Hardware, software, and procedures to collect the data need to be inserted into healthcare workflows. Healthcare workflows are built around specific structures, hierarchies, and certain levels of manual input. Health data is spread across different varying networks and not centralized into single databases, or, in some cases, never even copied from sheets of paper into digital form.
Getting alignment between all of the different stakeholders in the process—including data scientists, IT, operations, healthcare practitioners, providers, independent software vendors (ISVs), vendors, and others—is necessary to reduce friction in this process and make sure organizations can make the best use of AI and ML implementations. To face this challenge, stakeholders may need to use agile, vendor-agnostic software to best articulate issues and leverage clean and scalable data that is compatible with multiple ISVs.
The data needs to be collated and converted into interoperable and usable formats that work with information collected from various sources. A large amount of bandwidth is required to transmit data from the points in the network where it is collected, sometimes through edge devices. Storage space is expanding at an alarming rate due to the explosion of data being collected in healthcare systems, especially for things such as medical imaging, IOMT, and edge.
Cloud computing offers both high-performance and capacity to meet these challenges, however, this can be impractical in many cases, especially in rural settings and areas not served by robust IT and healthcare infrastructure. Solving this key challenge involves cost-effective solutions enhancing operations at the edge of the network and analyzing data at the point of care.
How to scale AI faster
Successfully deploying your AI workloads at scale depends on how efficiently and effectively your moving pieces are working together. Specifically, inference servers that can support larger AI models (like LLMs) and their more complex inference capabilities are essential to scaling AI workloads for the enterprise.
These AI tools use resources more efficiently to scale faster:
- llm-d: LLM prompts can be complex and nonuniform. They typically require extensive computational resources and storage to process large amounts of data. An open source AI framework like llm-d allows developers to use techniques like distributed inference to support the increasing demands of sophisticated and larger reasoning models like LLMs.
- Distributed inference: Distributed inference lets AI models process workloads more efficiently by dividing the labor of inference across a group of interconnected devices. Think of it as the software equivalent of the saying, “many hands make light work.”
- vLLM: vLLM, which stands for virtual large language model, is a library of open source code maintained by the vLLM community. It helps large language models (LLMs) perform calculations more efficiently and at scale.
Find out how Red Hat AI incorporates these tools and capabilities to help customers use AI at scale.
How Red Hat can help
Red Hat AI is a platform of products and services that can help your enterprise at any stage of the AI journey - whether you’re at the very beginning or ready to scale. It can support both generative and predictive AI efforts for your unique enterprise use cases.
With Red Hat AI, you have access to Red Hat® AI Inference Server to optimize model inference across the hybrid cloud for faster, cost-effective deployments. Powered by vLLM, the inference server maximizes GPU utilization and enables faster response times.
Red Hat AI Inference Server includes the Red Hat AI repository, a collection of third-party validated and optimized models that allows model flexibility and encourages cross-team consistency. With access to the third-party model repository, enterprises can accelerate time to market and decrease financial barriers to AI success.
Predictive AI use cases with Red Hat AI
With the right AI platform, you can use predictive AI to connect patterns, historical events, and real-time data to predict future outcomes with extremely high accuracy.
