Governed, secure, and private computational access to federated patient data is revolutionizing healthcare AI

In recent years, the medical technology (MedTech) sector has witnessed a rapid increase in AI adoption. This surge has largely been driven by the increasing availability of pre-trained, commoditized machine learning (ML) models. These models have simplified the development process, making it easier for MedTech companies to build AI applications on top of their medical hardware providing massive customer value, such as improved assisted diagnosis, assisted surgery with AI robotics, or streamlined clinical operations. However, amidst this progress, a significant challenge remains: How to leverage the valuable machine-generated data from the machine operators to customize pre-trained ML models to make better predictions.

The data dilemma: ownership, privacy and compliance

MedTech companies face a unique dilemma when it comes to leveraging the sensitive customer data generated by their machines. MedTech devices usually find their way into clinics, laboratories, and hospitals worldwide. These machine operators generate sensitive and personally identifiable information (PII) along with IP. The sensitive data generated is owned by the machine operators, not by the MedTech companies. Understandably, machine operators are reluctant to share that data often due to privacy and compliance concerns. Evolving privacy and AI regulations, as well as compliance with multiple privacy laws depending on the geographical location of machine operators, add to the challenge of MedTech companies leveraging machine generated data.

To gain access to valuable data to improve their software, MedTech organizations have explored compliant methods, such as the Digital Imaging and Communications in Medicine (DICOM) standard or the Structured Reporting System (SRS) logs. When these standards are applied to medical imaging data, the data can be shared directly. However, these methods often come at a cost: they inadvertently dilute the inherent value of the data and hinder its potential to fuel the advancement of artificial intelligence (AI). As a result, many MedTech companies have turned to other data anonymization methods that fall short of compliance requirements, especially in Europe. This leaves MedTech at a crossroads - the need to balance innovation with the demand for compliance.

"In the next 1-2 years, I expect severe fines for the non-compliant structures that many organizations have built up to get access to patient data for ML training”Partner at Jones Day – Life Science & Medical Device industry

Most MedTech companies know that to remain at the forefront of their field, they are compelled to adopt new, evolving algorithms, use new, cutting-edge AI technologies, and make use of the valuable data being captured by their customers. So how is it possible to get computational access to relevant data for AI applications in line with different regulatory requirements, without it becoming a time-consuming and costly ordeal?

Apheris' innovative solution: enabling secure computational access to data

The Apheris product addresses the core problem of getting computational access to sensitive, distributed customer data in a compliant and privacy-preserving manner.

Here’s how:

Federated infrastructure: bringing computations to the data

• Allows MedTech companies to perform computations on the machine operators’ data without the data ever leaving the machine operators’ IT environment.

• Only privacy-preserving results are shared with the MedTech organization.

• The machine operator remains the data controller and processor.

The Apheris Governance Portal: empowering machine operators

• Set the required level of privacy

• Define access control at the computational level, dictating who can run which computations on which of their datasets, and for what purpose.

The Apheris Compute Gateway: ML-powered insights without data sharing

• The Compute Gateway is installed within the machine operator’s IT environment.

• Ensures only approved computations, that are in line with the machine operator’s requirements, are launched on the data.

• Compute Gateways can communicate with each other, allowing MedTech organizations to scale computations across multiple machine operators.

The Apheris Trust Center: staying compliant across borders

• Provides guidance to the MedTech organization on model properties for computational data access

• Provides guidance to the machine operator on how to prepare data and on governance settings to comply with privacy and AI regulations.

• Hands-on practical examples from the model registry demonstrates compliant practices to help get started.

Apheris Model Registry: running computational jobs out-of-the-box

Contains a collection of Apheris approved models for MedTech organizations categorized on properties including:

• Privacy of models – applicability of privacy reconstruction attacks

• Safety of models – capabilities, limitations, robustness, applicability of security attacks (e.g., adversarial attacks)

• Fairness of models – explainability, trustworthiness, reliability

• Other properties required for transparency such as risks, how the model was tested, and what data the model was trained on

Apheris’ product offers a comprehensive and streamlined solution that integrates data governance and compliance management, ensuring that organizations can leverage AI technology without sacrificing security or broader compliance obligations.

Revolutionizing MedTech: Advantages of the Apheris product

Implementing Apheris' solution brings about transformative benefits to MedTech companies looking to leverage ML effectively:

• Accelerated model development: by eliminating the need to move data, MedTech organizations can get computational access to more data faster.

• Rapid updates: the Apheris solution enables rapid model updates to keep up with the evolution of AI technologies.

• Leverage diverse data: MedTech companies can train models on larger and more diverse datasets from multiple customers, resulting in more robust and accurate ML applications.

• Regulatory compliance: Apheris ensures compliance with various regulatory requirements, giving MedTech companies the confidence to navigate complex and evolving regulatory landscapes.

Seizing the paradigm shift: Capitalize on ML with Apheris

The AI revolution in MedTech is underway, and Apheris empowers you to realize its full potential. By productizing your customers' data in a compliant manner, you can be at the forefront of innovation while maintaining privacy and security. We invite you to reach out to us and explore how Apheris can be your partner in revolutionizing your AI journey.

Example use cases where Apheris solution for private, secure, and governed computational access to distributed sensitive machine data will revolutionize the MedTech sector include:

Clinical workflow support

AI-powered solutions that automate routine clinical workflows, freeing healthcare professionals from repetitive tasks and allowing them to focus on more critical aspects of their practice. Training ML models on distributed clinical workflow data from multiple hospitals enables the system to make the best predictions in different settings. Easily bring your data products through FDA approval to help your customers interpret datasets and receive analysis results for review, confirmation, and potential inclusion in final reports.

Read more on AI-assisted clinical workflow support: Developing reliable AI tools for healthcare (Google Deep Mind), AI-Rad Companion (Siemens Healthineers)

Multimodal 3D brain tumor segmentation

Quickly bring improved data products for your multimodal resonance imaging (MRI) machines to market by training cutting-edge ML models (e.g., U-Net Medical or K-means clustering) on multiple distributed MRI datasets from different clinics, labs, and hospitals. Help your customers accurately detect and segment brain metastases and predict tumor boundaries and sub-regions. Using the Apheris product, your data products will be able to accurately predict potential complications, recurrences and response to treatment and suggest the most appropriate course of action on a case-by-case basis.

Read more on multimodal 3D Brain Tumor Segmentation: Multimodal 3D Brain Tumor Segmentation with Azure ML and MONAI (Towards Data Science), Artificial Intelligence in Brain Tumor Imaging: A Step toward Personalized Medicine (National Library of Medicine)

AI-supported mammography screening

By bringing new data products for your mammography machines to market you can assist radiologists in their mammography interpretation. Training your ML algorithms on distributed mammography data via Apheris helps you to massively increase early breast cancer detection. Make your customers happy by reducing radiologists’ workload and help them to massively increase breast cancer detection accuracy.

Read more on AI-supported mammography screening: Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial (The Lancet Oncology), AI Outperformed Standard Risk Model for Predicting Breast Cancer (RSNA)

Empowering the future of MedTech AI

In a world where AI is reshaping industries, MedTech stands at the precipice of transformation. Apheris' solution enables you to overcome the data access dilemma, supercharging your AI initiatives. By placing privacy, security, and compliance at the forefront, we're enabling MedTech companies to develop, update, and deploy AI models with unprecedented speed and confidence. Join us in shaping the future of AI in MedTech. Contact us today to embark on this transformative journey.