Intel Is a Powerhouse in Healthcare AI: Here’s How

The Role of Hardware in AI Workflows

To support AI in healthcare, Intel’s hardware solutions form the essential infrastructure layer of the technology stack, including CPUs, GPUs, NPUs, field-programmable gate arrays, AI accelerators and edge AI systems.

These components enable efficient data processing, model training and inference for machine learning and deep learning applications. For instance, AI servers, which can be used to integrate these technologies, drive AI operations in enterprise data centers and edge environments, enabling advanced healthcare applications with speed and scalability.

Abhishek Khowala, Intel’s principal engineer for AI on the healthcare team, says the company provides the “incumbent platform architecture” in healthcare. This includes most of the electronic health records software deployed on Intel architecture in hospitals and clinics. Intel also collaborates with major players in medical imaging and life sciences.

For example, recent collaboration with ultrasound manufacturers has led to using CPUs and integrated graphics to employ AI to analyze imaging data, making it easier to conduct real-time assessments such as fetal head circumference measurements or cardiac evaluations.

Click the banner below to find out how infrastructure modernization increases healthcare agility.

Intel has equipped its workflow breadth architectures, such as Intel® Core™ Processors and Intel® Xeon® Processors, with algorithms to optimize AI. This includes Deep Learning Boost, which is optimized for deep learning tasks and enhances the performance of convolutional neural networks.

“DL Boost allows operations that used to take three instructions to be performed in just one,” Khowala says.

This streamlined processing enables faster and more accurate data handling on the computer chips, crucial for applications that require immediate insights, such as X-ray image analysis.

With enhanced edge processing capabilities, Intel’s hardware is vital for data-heavy tasks in patient-proximate environments. Noise and thermal management are key considerations in patient settings, adds Karen Perry, Intel’s chief solution architect for healthcare and life sciences sales. Intel’s optimized AI workloads can allow for edge processing right in the imaging room, removing the need to transmit data to the cloud for analysis, which in turn reduces latency.

Security and Data Management at the Edge

Healthcare organizations need secure, resilient infrastructure due to the sensitive nature of patient data and increasing cyberthreats.

“Healthcare is a prime target for hacking and ransomware,” Khowala explains.

By processing data locally rather than sending it to the cloud, hospitals can maintain compliance with data privacy regulations, navigate internet bandwidth issues and reduce vulnerabilities associated with external networks.

Resilience is another priority in healthcare environments, with redundant systems, backups and failover capabilities all considered essential. Perry says the resilience of Intel’s edge systems helps hospitals mitigate operational disruptions from security threats.

“In a healthcare setting, downtime isn’t an option,” she says, pointing to Intel’s embedded threat detection and out-of-band manageability features.

These help organizations detect, isolate and recover from potential cyberattacks without requiring a full shutdown.

PREPARE: Expert guidance helps healthcare organizations achieve meaningful transformation with AI.

Open-Source Software That Increases AI Application Adaptability

Open-source software tools, including Intel’s distribution of the OpenVINO™ toolkit, enhance the adaptability of AI applications across diverse healthcare environments.

“Healthcare facilities often use hundreds of applications, making flexibility essential,” Khowala says.

OpenVINO provides runtime environments and APIs, allowing healthcare organizations to integrate AI workflows and deploy models across multiple Intel hardware components without custom coding. This level of flexibility allows hospitals to adapt their AI implementations to their specific needs and existing infrastructure.

Federated Learning Overcomes Data Limitations

Khowala notes that Intel’s commitment to privacy extends into data sharing through federated learning, an approach that allows healthcare organizations to train AI models collaboratively without sharing actual patient data.

“We’re enabling hospitals to benefit from collective data wisdom while keeping their data secure,” he says.

Federated learning allows institutions to train models locally and aggregate insights across multiple organizations, resulting in robust AI models without violating data privacy rules.

It also supports diversity in AI training data, helping to reduce model bias and improve accuracy to create analytics that clinicians can trust.

Real-World Use Cases and Best Practices for Healthcare AI Adoption

The rapid advancement of AI technologies in healthcare is leading to practical applications such as ambient clinical documentation, which Khowala highlights as a top use case.

“Clinicians are experiencing burnout from the focus on electronic health record documentation,” he says.

Generative AI tools can automate tasks such as message generation, visit summaries and clinical concept extraction, reducing administrative burdens and allowing clinicians to focus on direct patient care.

Intel’s collaboration with the AI platform Hugging Face is another example of how it is enhancing accessibility to AI models.

“Through our integration with Hugging Face’s optimum layer, healthcare providers can access a library of optimized AI models and adapt them quickly to their specific needs,” Khowala says.

EXPLORE: How is quantum computing being used in healthcare?

These models, while based on large foundation models, are scaled down for practical, cost-effective applications in healthcare environments.

Khowala explains that by breaking down large AI models into smaller, manageable units, Intel enables healthcare organizations to deploy models tailored to their particular requirements.

These advancements help healthcare organizations avoid cloud dependencies and optimize performance directly on Intel’s processors, keeping both costs and response times low.

As healthcare organizations advance in their AI programs, it’s important that IT and AI leaders ensure that the right hardware is in place to support these initiatives.

Brought to you by: