How Leveraging Healthcare Real-World Data Fuels AI Innovation in Biopharma

GE Healthcare

Big data and artificial intelligence concept. Machine learning and cyber mind domination concept in form of women face outline outline with circuit board and binary data flow on blue background.

To unlock the promise of artificial intelligence (AI) for Biopharma, the right health data must be aggregated at the right time and provided to the right decision makers to integrate it with the overall clinical picture of health. This data is the source code of AI, underpinning the value and potential of the technology.

At present, the healthcare data ecosystem is fragmented. The future of Real-World Data aims to expedite the transmission of patient data to the managing physician and identify health problems earlier so that the physician can start treating these problems earlier. To get there, Biopharma needs greater access to Real-World Data from healthcare providers, enabling benefits including better disease identification and the prescription of relevant pharmaceuticals.

In recent years, AI across healthcare and biopharma has transformed from a “work-in-progress” technology into a viable solution. Similarly, companies are improving patient care through data with continuous glucose monitors (CGM) wearables. These CGMs demonstrate the immense value of contextual, Real-World Data, and coincide with an increasing uptake of wearable consumer medical devices. Further wearable capabilities include heart rate monitoring, SpO2 (pulse oximetry), and sleep tracking. This is what is currently available for consumers. Still, this data offers valuable context and insight into fundamental health metrics.

Concurrently, there is an increasing need for data sharing between healthcare providers and developers. By using healthcare provider data, developers can leverage the massive amounts of actionable Real-World Data being generated each day. The data supply chain must support data sharing in this context, improving accessibility to maximize the value of Real-World Data to deliver Precision Health.

Overcoming hurdles to data aggregation

There are numerous hurdles to collecting, aggregating, and translating this ready-for-use, Real-World Data in a clinical setting.

Regulatory frameworks

The first hurdle is changing regulation frameworks, from the challenges of maintaining patient privacy to the introduction of first-of-a-kind AI applications. When the FDA sees a new AI application integrated with an existing medical device, the result is an entirely new medical device augmented by software. Developers do not have the right to “piggyback” on the underlying device’s past approval by the FDA; the technology must be reassessed for its new intended use.

Interoperability

The second hurdle is the lack of interoperability in the healthcare data supply chain. Fragmented data collection in siloed IT environments reduces data agility, and prevents interoperability of IT systems, medical devices, and AI software. This interoperability is one of the most challenging but promising aspects of Real-World Data, creating a more complete picture for providers and patients to better understand their health. Furthermore, to drive interoperability, integrations or partnerships between pharma, healthcare providers, and developers are a necessity. This converts competition into cooperation, catalyzing innovation in the healthcare AI industry.

Adoption

The third hurdle is healthcare provider adoption. Providers manage a large range of service lines, which they need to monitor. When looking to adopt an AI solution, healthcare providers wonder how broadly they can apply these new technologies. Additionally, how much would it cost to integrate these AI technologies with every service line, thus maximizing the benefits? Large investments in technology must be justified through cost-benefit analysis and outcomes, however, data contained within Real-World Datasets can assist with this. Likewise, this justification through Real-World Data is what triggers reimbursement support from stakeholders, providing a pathway to deployment across all service lines.

How AI will define the future of healthcare

As more developers introduce AI applications, expertise and understanding around AI-based healthcare technology will grow. AI is still in its infancy, but the desired end goal is a mature concept: AI companionship will define the future of healthcare. This is where AI augments existing workflows across all service lines, acting as an assistant rather than a replacement. To get there, AI applications need to generate the same diagnostic outputs as healthcare professionals. If a radiologist views an x-ray or CT image and detects a potential tumor, the AI must be able to match or exceed these capabilities.

Real-World Data is AI’s equivalent to residency to a physician. It teaches the AI how to operate in line with the developer’s scope of action, essentially training the AI to detect discrepancies and deliver clinical value. In summary, the human-created ground truth and the AI-created model prediction must closely align to achieve regulatory validation—and Real-World Data makes this possible. Real-World Data is also immensely valuable for developers, when considering that AI validation is the final hurdle before go-to-market.

Real-world examples of healthcare leveraging Real-World Data and AI

There are a number of innovative AI applications that are using Real-World Data. GE Healthcare is making it easier to find them on the Edison™ Software Marketplace.[1] These applications represent the future of healthcare innovation, available now.

Improved CT Image Clarity Using AI

TrueFidelity™ CT[2] reduces the radiation dose needed for image clarity using AI reconstruction. Patients are exposed to less radiation, and image clarity improves through an iterative image denoising process based on the DLIR neural engine. The AI model is trained on over 1 million images, only possible with Real-World Datasets.

AI-Fueled Slice Prescription for Consistent MR Imaging

GE Healthcare’s Magnetic Resonance Artificial Intelligence Prescription,[3] MR AIRx™, is a workflow tool that leverages the Intel® OpenVINO™ toolkit. It assists with slice prescription on MR images, validating alignments to ensure consistency and quality of images in addition to reducing rescans and setup time. This AI model is trained on more than 36k images using Real-World Datasets.

Breast Sonograph Risk Assessment With AI

Available on the LOGIQ™ E10 Ultrasound Series,[4] Breast Assistant by Koios, provides automated, AI-based quantitative risk assessments of breast sonogram images. The AI model is trained to align with the BI-RADS®[5] image assessment framework.

All of these AI-enabled applications are made possible with Real-World Data. The more Real-World Data these applications can access, the better the diagnostic results they will generate. The focus for Edison™ Digital Pharma Solutions is to expand access to Real-World Datasets to accelerate the uptake and innovation around AI healthcare technologies. The end goal is to identify disease conditions earlier, enable higher quality patient care, and reduce barriers to AI application adoption for healthcare providers using outcomes data.

Bringing predictive analytics into the mainstream

Patients with rare diseases often see multiple providers many times before reaching the correct diagnosis, which means they are not receiving the needed treatment for these conditions as quickly as possible. The medical burden associated with delayed or undiagnosed disease can significantly reduce quality of life (QoL) for these patients and leaves them at risk of further disease progression.

To drive innovation around data and healthcare, we must bring predictive analytics (machine learning) into the mainstream, enabling the development of innovative new clinical decision support (CDS) tools. This will help healthcare providers identify relevant patient cohorts, in the hope of better diagnosing rare conditions. Furthermore, predictive analytics and CDS tools hold immense promise for facilitating clinical studies of new drugs. Real-World Data important in to generating the evidence needed to source regulatory approvals.

Finally, Real-World Data must be leveraged during the rollout of CDS at scale by implementing change management procedures to modify relevant workflows. These modified workflows will provide an alternative point of view (POV) for physicians, highlighting biomarkers and other indicators, aiming to achieve the right diagnoses.

Real-World Data offers a pathway to the future of healthcare. GE Healthcare Edison™ Digital Pharma Solutions offer the pathway to accessing and leveraging this Real-World Data, in line with data regulations, by augmenting the existing provider-to-developer data supply chain. The end goal with Real-World Data is improving the functionality of AI and driving adoption of these AI technologies among healthcare providers to deliver Precision Health.

Healthcare providers and developers are invited to contact Edison Digital Pharma Solutions to explore these new and exciting data-driven possibilities in healthcare. Click on the link to learn more about Edison™ Digital Pharma solutions.